3392 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 71

Graupel and Hail Terminal Velocities: Does a ‘‘Supercritical’’ Reynolds

Number Apply?

ANDREW HEYMSFIELD
NCAR,* Boulder, Colorado

ROBERT WRIGHT
Robert L. Wright & Associates, Inc., Houston, Texas

(Manuscript received 1 March 2014, in final form 13 April 2014)

ABSTRACT

This study characterizes the terminal velocities of heavily rimed ice crystals and aggregates, graupel, and
hail using a combination of recent drag coefficient and particle bulk density observations. Based on a non-
dimensional Reynolds number (Re)–Best number (X) approach that applies to atmospheric temperatures
and pressures where these particles develop and fall, the authors develop a relationship that spans a wide
range of particle sizes. The Re–X relationship can be used to derive the terminal velocities of rimed particles
for many applications. Earlier observations suggest that a ‘‘supercritical’’ Reynolds number is reached where
the drag coefficient for large spherical ice—hail—drops precipitously and the terminal velocities increase
rapidly. The authors draw on observations and model simulations for slightly roughened large ice particles
that suggest that the critical Reynolds number is dampened and that the rapid increase in the terminal velocity
of smooth spherical ice particles rarely occurs for natural hailstones.

1. Introduction in 2012. Crop–hail losses for 1949–98, as measured by

Hail produces considerable monetary damage in the
United States and throughout many parts of the world,
including Italy, South Africa, and Russia. Damage is
mostly to crops and buildings [see Changnon (2009) and
articles cited therein]. One primary source of hail in-
formation in the United States is the historical records of
the crop–hail insurance industry, kept since 1948. The
national annual values of insured crop–hail losses have
shown ever-increasing amounts (2014 dollars), from $20
million in 1948 to $174 million in 1974, jumping to $358
million in 1980, approaching $550 million in the early
1990s, and with continuing increases since. In 2010, there
were 468 000 claims for all types of hail damage filed in
the United States. That number increased to 862 000
* The National Center for Atmospheric Research is sponsored
by the National Science Foundation.
Corresponding author address: Andrew Heymsfield, NCAR,
3450 Mitchell Lane, Boulder, CO 80301.
E-mail: heyms1@ucar.edu
insurance data, averaged $575 million (2000 dollars) per
year (Changnon and Hewings 2001). The intensity of
hail at a point or over an area is a function of the fre-
quency of hail, the size of the hailstones as a result of
their fall speeds, their total cross-sectional area, and, of
course, a function of the hail particle size distribution
(their number concentration as a function of size) and
the horizontal wind speed. A search of the literature
reveals that there is a wide spread in estimates of graupel
and hail terminal velocities as a function of their size
(maximum dimension). The goal of our study is to im-
prove these estimates, thereby leading to improved
representations for use in climate models and hail
damage predictions.
Accurate estimates of the terminal velocities Vt of
graupel and hail are also important for improving the
representation of precipitation development in weather
forecast models. For example, using hail with a bulk
density of 0.9 g cm23 rather than graupel with a bulk
density of 0.4 g cm23 as the rimed ice species, with ap-
propriate adjustments to the Vt, a hail-producing squall-
line simulation resulted in a much narrower convective
line and slightly smaller stratiform region (Bryan and

DOI: 10.1175/JAS-D-14-0034.1

Ó 2014 American Meteorological Society

Unauthenticated | Downloaded 12/17/22 11:01 PM UTC
SEPTEMBER 2014 HEYMSFIELD AND WRIGHT 3393

TABLE 1: Cited rimed crystals/aggregates, graupel, and hail studies.

Researcher
Achenbach (1972)
A74
Auer et al. (1970)
BR37
Cheng (2009)
Cheng and Wang (2013)
Heymsfield and
Westbrook (2010)
Kajikawa (1975)
Knight and Heymsfield
(1983)
List et al. (1970)
Locatelli and Hobbs (1974)
Millikan and Klein (1933)
Roos (1972)
Roos and Carte (1973)
Takahashi and Fukuta (1988)
Wang and Kubicek (2013)
Willis et al. (1964)
Sizes
Study (cm; approximate) Comments
Drag coefficient of smooth sphere 6.0 Laboratory
Drag coefficient of rough sphere 6.0 Laboratory
Graupel/hail density, terminal velocity as 0.1–2.4 Field measurements
function of size
Drag coefficient of smooth ice spheres 4–14 Towed by airplane; data
taken from Millikan and
Klein (1933)
Drag coefficient of smooth spheres over a ,0.01–4
wide range of Re based on observations
Drag coefficients of hail — Model study
Best number–Reynolds number for — Laboratory, empirical,
particles of different shapes theoretical
Graupel, masses, terminal velocities 0.05–0.89 Field measurements
Graupel (hail) density, terminal velocity 0.85–1.6 Natural ice particles
(measured in laboratory)
Drag coefficient–Reynolds number for 12 Laboratory
smooth through roughened spheroids
Graupel (hail) mass, terminal velocity 0.05–2.0 Field measurements
Drag coefficient of hail 11.5 Towed behind airplane
Dropped replicas of largest hailstone 17.8 900-m drops, also wind
collected to date tunnel
Drag coefficient of hail replicas 1.3–11.8 Wind tunnels, mine shaft
Terminal velocities of rimed crystals 0.006–0.49 Vertical wind tunnel
Drag coefficients of graupel — Theoretical
Drag coefficients of melting roughened 5.1, 7.0 Dropped from balloon and
ice spheres tracked by radar

Morrison 2012). Comparing graupel and hail simula-
tions with similar adjustments, Milbrandt and Morrison
(2013) and van Weverberg et al. (2012) found that,
because of the higher Vt of this hail than graupel, the
rate of precipitation fallout was larger than for graupel,
and melting extended to the surface rather than not
far from the melting layer. However, most schemes
do not explicitly consider density but use a single Vt –D
relationship.
Terminal velocity depends on the drag coefficient Cd
and the particle mass m, which in turn depends on the
bulk density of the particles rb (the mass of the particle
divided by the volume of a sphere of the same maximum
dimension D). For a particle at its equilibrium velocity
at a given temperature T and pressure P,
Vt 5 (2mg/rf Cd A)0:5 5 [2(p/6)D3 rb g/rf Cd A]0:5 ,
where rf is the density of air, g is the gravitational ac-
celeration constant, and A is the cross-sectional area
normal to the airflow. Note that rb actually considers
two aspects: a reduction in density from that of a sphere
due to a nonspherical shape (e.g., spherical versus con-
ical graupel), and a reduction in density due to air
pockets and voids inside of the particle. Defined as in
(1), rb considers both effects.
Because Vt } Cd, accurate estimates of Vt must be
based on reliable Cd. For certain sizes of graupel and
especially hail, however, there is a sudden drop and
flattening of Cd followed by an increase at larger sizes.
This anomaly in Cd may be best illustrated by the ear-
liest measurements of particles simulating hailstones. To
‘‘at least furnish a rough idea of the order of magnitude
of the terminal velocity actually occurring in nature,’’
Bilham and Relf (1937, hereafter BR37, p. 150) reported
on the form of the relationship between terminal ve-
locity and diameter ‘‘deduced from values of the Cd
obtained from observations on 4–14-cm spheres towed
by airplanes’’ (see Table 1). They reported a rapid
dropoff in Cd for Reynolds numbers 300 000 , Re ,
400 000 and a leveling off of Cd for Re . 4000 (Fig. 1a).
The Reynolds number is given by
(1)
Re 5 Vt D/n , (2)
where n is the kinematic viscosity of air. Note that the
actual data in the BR37 study were not reported, only
the form of the relationship. The actual data used to
derive the curve they show in their article were derived
from a pioneering study by Millikan and Klein (1933),
although this data source was not acknowledged in their
article and took hunting to find.

Unauthenticated | Downloaded 12/17/22 11:01 PM UTC

3394 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 71

FIG. 1. (a) Drag coefficient as a function of Reynolds number as derived by towing spheres
behind an airplane (BR37) and (b) calculated terminal velocities from those data, assuming
that the particles are solid ice spheres. Data points are derived from their Fig. 2.

With points taken from their Fig. 2 (and as shown in
their Table 1) converted to appropriate units, Vt can be
readily derived assuming the hail particles are solid
ice spheres and falling at standard sea level conditions
(Fig. 1b). Note that the BR37 definition of the Cd is one-
half of the conventional definition, but we have adjusted
their values upward by a factor of 2. Rather than
showing a monotonic increase in Vt with size from 0 to
above 15 cm and a constant Cd of 0.4, which is approxi-
mately what is usually assumed (dashed line, Fig. 1b),
the resulting Vt show a rather large increase for sizes
beginning at 4 cm and below 10 cm, associated with the
decrease in Cd. The resulting Vt is due to the trend of Cd
with Re and the interplay of Cd and D and suggests that
there may be an unexpected change in the Vt–D
relationship. BR37 specifically note that ‘‘Over a certain
range of diameters, depending on the densities of air and
ice and the kinematic viscosity of air, three values of the
terminal velocity would therefore appear possible.’’ The
dip in the Cd–Re relationship above Re ; 2 3 105 has
been referred to as a ‘‘supercritical Reynolds number’’
regime (Willis et al. 1964).
Following in the footsteps of Laurie (1960), the
building industry has widely used the BR37 Cd–Re and/
or the Vt–D relationships to assess and infer hail impact
damage. Therefore, we need to ask: does this anomaly
apply to real hailstones? Are there other unexpected
changes in graupel and hail fall speeds that should be
addressed? Can the values of Cd–Re based on smooth
hail be applied to natural hailstones?

Unauthenticated | Downloaded 12/17/22 11:01 PM UTC

SEPTEMBER 2014 HEYMSFIELD AND WRIGHT 3395

FIG. 2. Symbols showing the diameters of particles referred to as graupel in the studies noted
in the legend as a function of the measured mass. Associated bulk densities are plotted as lines
of constant slope. The black filled circles are the mean masses as a function of diameter, and the
standard deviations, from the four datasets plotted.

To answer these questions, this article uses a general-
ized, nondimensional Reynolds number–Best number
(X) approach often used to derive Vt that depend on
characteristic properties of the particles and the local
temperature and pressure (see, e.g., Heymsfield and
Westbrook 2010). In what follows, we relate Cd to Re and
Re to X, so that for a given ice particle and environ-
mental conditions, X can be computed directly, Re can
be estimated from X, and thus the fall speed can be
derived directly.
In section 2, we present many of the observations of
rimed crystals or aggregates, graupel, and hail drag co-
efficients and associated masses collected to date. Sec-
tion 3 attempts to explain the observations and draw on
theoretical models, resulting in the development of the
nondimensional relationships for deriving the terminal
velocities. The results are summarized and conclusions
drawn in section 4.
A word on the nomenclature used in this article:
Drawing on the American Meteorological Society Glos-
sary of Meteorology (available online at http://glossary.
ametsoc.org/wiki/Main_Page), hail is precipitation in the
form of balls or irregular lumps of ice greater than 5 mm;
small hail is hail less than 6.4 mm. Graupel are heavily
rimed snow particles, often indistinguishable from very
small soft hail, except for the size convention that these
particles are less than 5 mm. Heavily rimed particles and
aggregates are noted where applicable.
2. Observations of graupel and hail
This section presents many of the earlier Cd–Re ob-
servations and estimations from numerical simulations
of the drag coefficients of rimed crystals and aggregates,
graupel, and hail. The primary datasets from which we
draw are summarized in Table 1.
According to (1), terminal velocity is proportional to
particle mass. Particle mass is proportional to the par-
ticle bulk density, which is a highly variable and un-
certain parameter. The graupel mass has a major effect
on its fall speed, and there have been relatively few
measurements of graupel rb. As shown in Fig. 2, graupel
masses and the associated rb average about 0.1 g m–3.
The rb values from the Knight and Heymsfield (1983)
study (Fig. 2) are above the large end of the graupel size
designation and are within the small end of the hail size
range—small hail—although the rb values are relatively
low. They derived the actual densities through volu-
metric measurements, thereby considering the ice-free
volume inside the particle. Densities so derived were an
average of 0.44 g cm23 for the actual particles and an
average of 0.31 g cm23 for the corresponding values of rb,
or about 50% larger. Their rb values are higher than for
most of the graupel studies to date and may suggest that
the rb values for hail are higher than those for graupel.
When graupel are riming in relatively high–liquid
water content regions and at subfreezing temperatures

Unauthenticated | Downloaded 12/17/22 11:01 PM UTC

3396 JOURNAL OF THE ATMOSPHERIC SCIENCES
FIG. 3. Drag coefficients as a function of the Reynolds number from observations of graupel
and small hail (symbols) and empirically and theoretically derived values.
but close to the 08C level, rb values are probably higher
than those in Fig. 2. Heymsfield and Pflaum (1985)
measured the densities of graupel beginning on frozen
drops with diameters of 0.3 and 0.6 mm and growing in
a vertical wind tunnel with air temperatures from 238 to
2158C and liquid water contents from 0.5 to 2.6 g cm23.
The relationship that they developed was directly pro-
portional to the graupel density and the ‘‘impact veloc-
ity’’ of the cloud droplets on the graupel (which is
approximately equal to the graupel terminal velocity)
and inversely proportional to the surface temperature of
the graupel. The density of the deposited rime at the
warmest temperatures was about 0.6 and 0.25 g cm23 for
most temperatures and liquid water contents. Cober and
List (1993) have also quantified the density of rime from
laboratory measurements. Based on these experiments
and the observations of Knight and Heymsfield (1983), it
is reasonable to assume that, in convection with rela-
tively high liquid water content and warmer cloud-base
conditions than found in most of the collections used in
Fig. 2 and most of the experiments reported in the
Heymsfield and Pflaum (1985) study, graupel densities
are closer to 0.3 g cm23.
A curve fit to the graupel mass versus diameter data in
Fig. 2, plotted as a line and listed in the figure, indicates
that on average, the graupel density increases with size
(exponent in the mass power law greater than 3). As-
suming for simplicity that the graupel are approximately
spherical rather than conical, one can derive an associated
power-law density–diameter relationship for graupel:
VOLUME 71
r 5 0:18D0:33 , (3)
where D is in centimeters and r is in grams per cubic
centimeters. Note that diverse datasets are included in
Fig. 2 and that the graupel density for one type of scenario
(e.g., graupel developed from frozen drops in deep con-
vection) may not conform to the results shown in Fig. 2.
The few measurements of rb of hail suggest that they
are often close to those of solid ice density. Prodi (1970)
used an x-ray absorption technique to measure the in-
ternal density (e.g., cavities and hollows) of natural and
artificial hailstones, obtaining values between 0.82 and
0.87 g cm23. List et al. (1970) measured rb of naturally
occurring 2.54–4.82-cm-diameter hailstones to be ap-
proximately 0.9 g cm23. By comparison, Knight and
Heymsfield (1983) found rb from 0.31 to 0.61 g cm23
with an average of 0.44 g cm23 for hail from 0.63 to
1.54 cm (see Fig. 2). These particles developed in
a springtime Colorado convective storm with surface
temperatures of 128C, implying that little melting had
occurred and providing an indication of how much in-
terior volume is free of ice.
No major sudden decreases have been found in the
drag coefficients and the resulting terminal velocities of
rimed crystals, rimed aggregates, and graupel particles.
Variable parameters for Vt include the graupel di-
ameter, shape (conical or spherical), fall orientation,
and thus the cross-sectional area, mass, and associated
bulk density. Figure 3 shows observations of Cd for
particles identified as heavily rimed crystals, rimed
Unauthenticated | Downloaded 12/17/22 11:01 PM UTC
SEPTEMBER 2014 HEYMSFIELD AND WRIGHT
aggregates, graupel, and small hail are shown as a func-
tion of their derived Reynolds numbers (Locatelli and
Hobbs 1974; Kajikawa 1975; Knight and Heymsfield
1983; Takahashi and Fukuta 1988; see Table 1 for
identification of the various studies). With increasing Re
or size, the Cd decrease by almost an order of magnitude,
signifying that all else being equal, Vt would increase by
about a factor of 3 rather than remaining constant.
Drawing on fluid dynamic tank studies and observa-
tions, Heymsfield and Westbrook (2010) developed an
empirical relationship between Cd and Re for graupel as
a function of its area ratio Ar, defined as the ratio of the
graupel cross-sectional area normal to the airflow rela-
tive to that of a circle with the same maximum diameter.
This formulation predicts that Cd decrease from about
4 to 0.7 with increasing Re over the typical range of
typical graupel Re (Fig. 3). Their formulation yields
Cd that are comparable to those derived from a more
recent theoretical study (Wang and Kubicek 2013),
where in Fig. 3a we plot both formulations assuming
Ar 5 1. Both the Heymsfield and Westbrook (2010) and
Wang and Kubicek (2013) formulations give a relation-
ship between Cd and Re that agrees well with those
derived from direct measurements (Fig. 3, from studies
reported in Table 1). The main differences between the
observations, those from tank experiments, and the
theoretical Cd (Re) are differences in Cd at the lower Re,
possibly because of the graupel shape (spherical versus
conical).
Observations of the Cd for hailstone-size particles
indicate that the particle’s external structure has a major
effect on the resulting relationship between Cd and Re.
Observations of the Cd for spherical, smooth ice are
plotted in Fig. 4a. The rapid dropoff in Cd with Re noted
in the observations of BR37 (Fig. 4a, curve 1) has since
been corroborated by Achenbach (1972; curve 2 and
others). However, as the roughness of the ice sphere
(hailstone) increases, the dip in Cd with Re becomes
increasingly flattened [cf. filled and open circles, labeled
11 and 12, from Achenbach (1974, hereafter A74) and
curves 7 and 8 from List et al. (1973); Fig. 4b]. As illus-
trated schematically in Figs. 4b and 5b, the amount of
roughening is quantified by d/D, where d is the diameter
or height of the nodules on the surface of the hailstone
(visualized as the reverse of the diameter of the dimples
on a golf ball). In the A74 experiments, the rough sur-
face was obtained by pasting glass spheres of diameter
d onto the surface or by abrading the surface with coarse
emery paper. Indeed, for the more typical ‘‘spiky’’ giant
hailstones and from theoretical considerations (curves 7
and 13 of Fig. 4b), the supercritical Reynolds number
and the very low values of drag coefficients observed for
smooth ice spheres are not observed.
3397
Roughening damps out the supercritical Reynolds
number observed for the smooth ice spheres. To show
this effect more clearly, Figs. 5a and 5b plot Cd versus
Re for smooth spheres and lightly through very rough
spheres, respectively. The Re where Cd dips de-
creases with increasing roughening and is clearly
damped out.
A question remains as to what happens to the Cd of
hail as it falls through the melting layer and partially
melts. In an ingenious experiment, Willis et al. (1964)
lofted ice spheres to subfreezing temperatures on bal-
loons, and when the balloons popped, the spheres de-
scended through the melting level. A 5.1-cm slightly
roughened ice sphere with a roughness parameter (ratio
of the diameter of the roughened surface to the diameter
of the sphere) less than 0.02 initially had a drag co-
efficient of 0.26 and Re of 1.3 3 105. After melting, Cd
increased to 0.56 and Re dropped by about 30%. For
a 7-cm lightly roughened sphere with an initial Cd of 0.24
and Re of about 1.6 3 105 and just below the critical
Reynolds number, the drag coefficient remained con-
stant during melting, presumably because the ice sphere
diameter changed little during the melting process. Be-
cause of the effect of roughening on the fall speeds of the
simulated hailstones, they concluded that the rapid in-
crease in the terminal velocity observed for smooth ice
spheres may only be ‘‘attained rather infrequently in
nature’’ (Willis et al. 1964, p. 107).
In summary, graupel rb values increase with size or
Re, and Cd decreases with increasing Re or size in
a systematic fashion that is reasonably well predicted
from fluid tank (empirical) data and theoretical con-
siderations. Hail of sizes 2 cm and above have densities
close to those of solid ice. Experimental data indicate
that there is a critical Reynolds number for smooth ice
spheres, but this rapid decrease in the Cd becomes more
damped and the dip occurs at increasingly lower Re as
the ice sphere becomes more roughened. Given this
observation and the observations of Browning and
Beimers (1967), whose data suggest that most hailstones
are likely to be oblate and spiky, it can be concluded that
only in occasional instances would a critical Reynolds
number be reached by a naturally occurring hailstone.
3. Synthesis of results leading to the calculation of
graupel and hail fall speeds
For calculation of terminal velocities, it is convenient
to use the nondimensional Davies or Best number,
X 5 Cd Re2 , (4)
where
Unauthenticated | Downloaded 12/17/22 11:01 PM UTC
3398 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 71

FIG. 4. Drag coefficient data from observations and theoretical considerations as a function
of the Reynolds number for (a) smooth ice spheres and (b) ice spheres with differing degrees of
surface roughness, defined as the ratio of d/D, where d is magnified in the figure. The sizes of the
symbols are for plotting purposes only, not for visualizing the relative ratios of d/D.

X 5 2mD2 g/(rf y 2 A) .
In other words, X may be calculated from the known
properties of the particle and the air temperature and
pressure. The objective of this section is to relate X to
Re, so that for a given ice particle and environmental
condition, X can be computed directly, Re can be esti-
mated from X, and thus the fall speed can be derived
using (2).
Drawing on the rimed crystal, graupel, and small-hail
datasets, and empirical and theoretical relationships
identified in Table 1 and used in Fig. 3, Fig. 6 shows the
relationship between the Best and Reynolds numbers
(5) for rimed through small hail particles. There is a mono-
tonic increase in Re with X and no surprising deviations
across the range of Re. The theoretical (Wang and
Kubicek 2013) and the Heymsfield and Westbrook (2010)
relationships agree quite closely except at the lower
Re values, and both provide a reasonably good match with
the observations. The Heymsfield and Westbrook (2010)
data provide a reasonably good match to the data across
the full range of approximately 1 , Re , 105.
The Reynolds number–Best number relationships for
smooth and roughened spheres from the observations
and for oblate spheroids based on theoretical consider-
ations are compared in Fig. 7. The dotted straight lines in

Unauthenticated | Downloaded 12/17/22 11:01 PM UTC

SEPTEMBER 2014 HEYMSFIELD AND WRIGHT 3399

FIG. 5. The relationship between drag coefficient and Reynolds numbers for spheres with
increasing roughness as defined by d/D. The data are from BR37, A74, and Young and
Browning (1967, hereafter YB67). The numbers shown in the lower left are the approximate
diameter of the spheres (cm).

the two panels show the corresponding Cd. For smooth
spheres, the discontinuity in the Re–X relationship as X
increases above 1011 clearly reflects the abrupt decrease
in Cd that would be associated with a rapid increase in Vt
(Fig. 7a). This effect is appreciably muted or nonexistent
for lightly to heavily roughened ice spheres and, where
noted, occurs at much lower values of X (Fig. 7b). A
similar effect is noted from dimples on a golf ball (Choi
et al. 2006), where the dimples cause local flow separa-
tion and trigger the shear-layer instability along the
separating shear layer, resulting in the generation of
large turbulence intensity. In any event, the theoretical

FIG. 6. Reynolds number–Best number relationship for rimed particles, graupel, and small
hail based on observations of natural particles and from wind tunnel, theoretical, and fluid tank
experiments.

Unauthenticated | Downloaded 12/17/22 11:01 PM UTC

3400 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 71

FIG. 7. Reynolds number vs Best number for (a) smooth spheres and (b) light- to heavily
roughened ice spheres. Dotted lines indicate constant drag coefficients of 0.4, 0.5, and 1.0. A
curve based on theoretical considerations for oblate spheroids is shown.

Re–X curve in Fig. 7 falls considerably below the data The Re–X relationship developed here conforms well
for roughened spheres. to the one developed by Knight and Heymsfield (1983)
To prevent discontinuities in the fall speeds of graupel for small hail and hail of sizes 0.63–1.54 cm (Fig. 8).
and hail for modeling studies and other applications, it is Because their relationship was developed for the actual
desirable to develop a continuous relationship across all particles both as they fell and then after soaking them in
values of the Best number. A power-law Re–X curve, oil that fills the particle, it is probably applicable for both
fitted over different ranges of X with coefficient of fits of small hail as it falls through the melting level and fol-
0.95 and 0.99, respectively, are lowing partial melting to fill void cavities. For rimed
crystals and graupel, with sizes below those they con-
Re 5 0:106X 0:693 , X , 6:77 3 104 (6a) sidered, their relationship would overestimate Re, and
for large hail at sizes above those they considered, it
and would underestimate Re.
We now apply the new Re–X relationship to derive
Re 5 0:55X 0:545 , X . 6:77 3 104 . (6b) graupel and hail Vt. For simplicity, we assume a pressure

Unauthenticated | Downloaded 12/17/22 11:01 PM UTC

SEPTEMBER 2014 HEYMSFIELD AND WRIGHT
FIG. 8. Reynolds number vs Best number for graupel and hail combined. Large symbols show
median values for the graupel dataset and for the hail for roughened ice spheres with a best fit to
these median values plotted and listed in the figure.
level of 1000 hPa and temperature of 258C. To illustrate
the dependence of Vt on D, two densities are assumed:
0.2 and 0.9 g cm23
. The results indicate that the logarithm
of Vt is nearly a linear function of the logarithm of D,
indicating that they can be represented by the following
power laws:
Vt 5 523D0:65 for a density of 0:2 g m23, (7a)
and
Vt 5 1207D0:64 for a density of 0:9 g m23, (7b)
where D is in centimeters and Vt is in centimeters per
second. From (6), these relationships change with P by
the ratio of (1000/P)0.55.
It is also straightforward to estimate the effect of the
particle density on Vt. The density–diameter relation-
ship developed for graupel and small hail in Fig. 2 [see
(3)] is also applied in Fig. 9. The Vt–D relationship
becomes
Vt 5 488D0:84 ; (8)
thus, with a steeper slope comes an expected result.
Note that this equation would only apply to particles
that originate as rimed crystals.
Our Vt–D relationship, for the same particle proper-
ties, is quite similar to those of Milbrandt and Morrison
3401
(2013) for about the same bulk densities (Fig. 9). Their
relationships are developed from a prognostic graupel
density scheme that uses the Reynolds number–Best
number approach of Mitchell and Heymsfield (2005)
assuming that particles are spherical. Also shown is a fit
to the terminal velocities derived from the Mitchell and
Heymsfield (2005) representation for graupel, assuming
solid ice spheres with an area ratio of 1.0. These conform
nicely to our new relationship out to 0.1 cm and then
diverge from it. Our new Re–X relationship, fitted
across the full range of Re–X for rimed particles through
hail, extends across a wide range of diameters and can
be readily used with variable densities (e.g., see light
blue solid curve; Fig. 9). Also shown in Fig. 9 are Vt–D
relationships used in weather forecast modeling by
Thompson et al. (2008) and Lim and Hong (2010). Al-
though not intended to be used at such large diameters,
those relationships do produce Vt that are only somewhat
higher than those developed here. The range of diameters
considered, although too large, is plotted because other
modeling studies may apply them in this way.
As suggested by the experiments of Willis et al. (1964),
melting will lead to a decrease in the graupel/hail di-
ameter, more so for graupel than hail, with the net result
being that X, Re, and hence Vt, will decrease. From those
experiments with 5.1- and 7-cm hail, there is no evidence
that there is an abrupt change in the form of the Re–X
relationship. Soaking of graupel or hail with densities
below those of solid ice (Knight and Heymsfield 1983)
Unauthenticated | Downloaded 12/17/22 11:01 PM UTC
3402 JOURNAL OF THE ATMOSPHERIC SCIENCES
FIG. 9. Terminal velocities derived from the Re–X relationship, two assumed constant ice
densities and one variable [from (3)], with an area ratio of 1.0, for a pressure level of 1000 hPa
and temperature of 258C. Also shown are representations for graupel and hail developed by
other researchers.
may change the mass but not the diameter, and this in-
creased mass could increase the terminal velocity.
4. Summary and conclusions
The goal of this study is to develop accurate expressions
for the terminal velocities of rimed crystals, graupel, and
hail over size ranges from several tenths of a millimeter to
above 10 cm in diameter that can be used in cloud mod-
els or engineering-related applications and are general
enough to be used for most atmospheric pressures and
temperatures. Using observations of the masses and
terminal velocities of these particles reported in the liter-
ature, we developed a nondimensional Reynolds number–
Best number relationship that spans a broad range of Re
and corresponding rimed crystal–graupel–hail sizes.
Necessary inputs into the development of this re-
lationship are the particle masses or bulk densities and
drag coefficients. For heavily rimed crystals or aggregates
and graupel, rb values are found to average about
0.1 g cm23. In convection with relatively high liquid water
contents and warmer cloud-base conditions than most of
the data used in our study, graupel densities are likely to
be closer to 0.3 g cm23. The interior volume of graupel
undergoing melting will fill with water, leading to densities
of solid ice or above. Also, nonspherical graupel will be-
come more spherical, with higher masses for a given size.
The few observations of the rb of hail are closer to those of
solid ice, about 0.5–0.9 g cm23, and possibly higher for
VOLUME 71
low-density hail soaked with liquid water during a melting
period. Also, because hail is often oblate and not spheri-
cal, most hail will have smaller masses for a given size,
leading to lower terminal velocities. Detailed calculations
of rb using models with explicit consideration of the
density of accreted water (e.g., Mansell et al. 2010;
Milbrandt and Morrison 2013) are needed to further
quantify particle densities under different scenarios.
Using rimed crystal and graupel mass and terminal
velocity data, it is found that for increasing Re, Cd of
heavily rimed crystals through graupel particles de-
crease by about an order of magnitude, from about 10 to
1. Drag coefficients developed recently based on theo-
retical treatments and laboratory tank data for graupel
are somewhat lower than the observations. The Cd
values of hailstones of about 0.4–0.45 are somewhat
lower than the often-assumed value of about 0.5. The
so-called hail ‘‘critical Reynolds number’’ observed ex-
perimentally and theoretically for smooth ice spheres
above 10 cm in diameter is damped and occurs at in-
creasingly smaller diameters as the particle surface
becomes rougher. It can be concluded that only in oc-
casional instances would a critical Reynolds number be
reached by naturally occurring hailstones.
Acknowledgments. The authors thank Hugh Morrison
and Jason Milbrandt for their helpful comments. We
especially thank Meg Miller for her editorial assistance.
This research was partially supported by NCAR.
Unauthenticated | Downloaded 12/17/22 11:01 PM UTC
SEPTEMBER 2014 HEYMSFIELD AND WRIGHT
REFERENCES
Achenbach, E., 1972: Experiments on the flow past spheres at
very high Reynolds numbers. J. Fluid Mech., 54, 565–575,
doi:10.1017/S0022112072000874.
——, 1974: The effects of surface roughness and tunnel blockage on
the flow past spheres. J. Fluid Mech., 65, 113–125, doi:10.1017/
S0022112074001285.
Auer, A. H., Jr., J. D. Marwitz, G. Vali, and D. L. Veal, 1970: Final
report to the National Science Foundation in fulfillment of
grants GA-1527 and GA-19105, Department of Atmospheric
Resources Rep., University of Wyoming, 92 pp.
Bilham, E. G., and E. F. Relf, 1937: The dynamics of large hail-
stones. Quart. J. Roy. Meteor. Soc., 63, 149–162, doi:10.1002/
qj.49706326904.
Browning, K. A., and J. G. D. Beimers, 1967: The oblateness of
large hailstones. J. Appl. Meteor., 6, 1075–1081, doi:10.1175/
1520-0450(1967)006,1075:TOOLH.2.0.CO;2.
Bryan, G. H., and H. Morrison, 2012: Sensitivity of a simulated
squall line to horizontal resolution and parameterization of
microphysics. Mon. Wea. Rev., 140, 202–225, doi:10.1175/
MWR-D-11-00046.1.
Changnon, S. A., 2009: Increasing major hail losses in the U.S.
Climatic Change, 96, 161–166, doi:10.1007/s10584-009-9597-z.
——, and G. J. Hewings, 2001: Losses from weather extremes in the
United States. Nat. Hazards Rev., 2, 113–123, doi:10.1061/
(ASCE)1527-6988(2001)2:3(113).
Cheng, K.-Y., and P. K. Wang, 2013: A numerical study of the flow
fields around falling hails. Atmos. Res., 132–133, 253–263,
doi:10.1016/j.atmosres.2013.05.016.
Cheng, N.-S., 2009: Comparison of formulas for drag coefficient
and settling velocity of spherical particles. Powder Technol.,
189, 395–398, doi:10.1016/j.powtec.2008.07.006.
Choi, J., W-P Jeon, and H. Choi, 2006: Mechanism of drag re-
duction by dimples on a sphere. Phys. Fluids, 18, 041702––,
doi:10.1063/1.2191848.
Cober, S. G., and R. List, 1993: Measurements of the heat and mass
transfer parameters characterizing conical graupel growth. J. At-
mos. Sci., 50, 1591–1609, doi:10.1175/1520-0469(1993)050,1591:
MOTHAM.2.0.CO;2.
Heymsfield, A. J., and J. C. Pflaum, 1985: A quantitative as-
sessment of the accuracy of techniques for calculating
graupel growth. J. Atmos. Sci., 42, 2264–2274, doi:10.1175/
1520-0469(1985)042,2264:AQAOTA.2.0.CO;2.
——, and C. D. Westbrook, 2010: Advances in the estimation of ice
particle fall speeds using laboratory and field measurements.
J. Atmos. Sci., 67, 2469–2482, doi:10.1175/2010JAS3379.1.
Kajikawa, J., 1975: Measurement of falling velocity of individual
graupel particles. J. Meteor. Soc. Japan, 53, 476–481.
Knight, N. C., and A. J. Heymsfield, 1983: Measurement and in-
terpretation of hailstone density and terminal velocity. J. Atmos.
Sci., 40, 1510–1516, doi:10.1175/1520-0469(1983)040,1510:
MAIOHD.2.0.CO;2.
Laurie, J. A. P., 1960: Hail and its effects on buildings. National
Building Research Institute Council for Scientific and In-
dustrial Research Rep. 176, 12 pp.
Lim, K.-S. S., and S.-Y. Hong, 2010: Development of an effective
double-moment cloud microphysics scheme with prognostic
3403
cloud condensation nuclei (CCN) for weather and climate
models. Mon. Wea. Rev., 138, 1587–1612, doi:10.1175/
2009MWR2968.1.
List, R., P. R. Kry, and U. W. Rentsch, 1970: On the tumbling of
spheroidal hailstones. Proc. Conf. on Cloud Physics, Fort
Collins, CO, Amer. Meteor. Soc., 67–68.
——, U. W. Rentsch, A. C. Byram, and E. P. Lozowski, 1973: On
the aerodynamics of spheroidal hailstone models. J. Atmos.
Sci., 30, 653–661, doi:10.1175/1520-0469(1973)030,0653:
OTAOSH.2.0.CO;2.
Locatelli, J. D., and P. V. Hobbs, 1974: Fall speeds and masses of
solid precipitation particles. J. Geophys. Res., 79, 2185–2197,
doi:10.1029/JC079i015p02185.
Mansell, E. R., C. L. Ziegler, and E. C. Bruning, 2010: Simulated
electrification of a small thunderstorm with two-moment
bulk microphysics. J. Atmos. Sci., 67, 171–194, doi:10.1175/
2009JAS2965.1.
Milbrandt, J. A., and H. Morrison, 2013: Prediction of graupel
density in a bulk microphysics scheme. J. Atmos. Sci., 70, 410–
429, doi:10.1175/JAS-D-12-0204.1.
Millikan, C. B., and A. L. Klein, 1933: The effect of turbulence: An
investigation of maximum lift coefficient and turbulence in
wind tunnels and in flight. Aircr. Eng. Aerosp. Technol., 5,
169–174, doi:10.1108/eb029703.
Mitchell, D. L., and A. J. Heymsfield, 2005: Refinements in the
treatment of ice particle terminal velocities, highlighting ag-
gregates. J. Atmos. Sci., 62, 1637–1644, doi:10.1175/JAS3413.1.
Prodi, F., 1970: Measurements of local density in artificial and
natural hailstones. J. Appl. Meteor., 9, 903–910, doi:10.1175/
1520-0450(1970)009,0903:MOLDIA.2.0.CO;2.
Roos, D. S., 1972: A giant hailstone from Kansas in free fall. J. Appl.
Meteor., 11, 1008–1011, doi:10.1175/1520-0450(1972)011,1008:
AGHFKI.2.0.CO;2.
——, and A. E. Carte, 1973: The falling behavior of oblate and
spiky hailstones. J. Rech. Atmos., 7, 39–52.
Takahashi, T., and N. Fukuta, 1988: Observations of the embryos
of graupel. J. Atmos. Sci., 45, 3288–3297, doi:10.1175/
1520-0469(1988)045,3288:OOTEOG.2.0.CO;2.
Thompson, G., P. R. Field, R. M. Rasmussen, and W. D. Hall, 2008:
Implicit forecasts of winter precipitation using an improved
bulk microphysics scheme. Part II: Implementation of a new
snow parameterization. Mon. Wea. Rev., 136, 5095–5115,
doi:10.1175/2008MWR2387.1.
van Weverberg, K., A. M. Vogelmann, H. Morrison, and J. A.
Milbrandt, 2012: Sensitivity of idealized squall-line simula-
tions to the level of complexity used in two-moment bulk
microphysics schemes. Mon. Wea. Rev., 140, 1883–1907,
doi:10.1175/MWR-D-11-00120.1.
Wang, P. K., and A. Kubicek, 2013: Flow fields of graupel falling in air.
Atmos. Res., 124, 158–169, doi:10.1016/j.atmosres.2013.01.003.
Willis, J. T., K. A. Browning, and D. Atlas, 1964: Radar observations
of ice spheres in free fall. J. Atmos. Sci., 21, 103–108, doi:10.1175/
1520-0469(1964)021,0103:ROOISI.2.0.CO;2.
Young, R. G. E., and K. A. Browning, 1967: Wind tunnel tests of
simulated spherical hailstones with variable roughness. J. At-
mos. Sci., 24, 58–62, doi:10.1175/1520-0469(1967)024,0058:
WTTOSS.2.0.CO;2.
Unauthenticated | Downloaded 12/17/22 11:01 PM UTC