Вы находитесь на странице: 1из 22

Silver iodide seeding impact on the microphysics and dynamics of

convective clouds in the high plains


Baojun Chen
a,
, Hui Xiao
b
a
Key Laboratory of Mesoscale Severe Weather/MOE, School of Atmospheric Sciences, Nanjing University, Nanjing 210093, China
b
Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China
a r t i c l e i n f o a b s t r a c t
Article history:
Accepted 6 April 2009
The concept of dynamic seedingis a physicallyplausible hypothesis but has not yet beenconrmed
by observations or numerical simulations. To verify the hypothesis of dynamic seeding, a three-
dimensional nonhydrostatic cloud model with two-moment bulk microphysics scheme has been
used to investigate the effects of silver iodide seeding on cloud microphysics, dynamics and
precipitation of convective storms. Eight species of water are included in the model: vapor, cloud
water, rain, cloud ice, snow, graupel, frozendrops and hail. AHighPlains hailstormcase developed
on 1 August 1981 during the Cooperative Convective Precipitation Experiment season is used for
the simulations.
The simulated cloud system consists of two cloud cycles during the period of integration. The
second-cycle clouds are the dominant precipitation producer of the simulation, contributes
about 80% of the total surface precipitation, which are caused by interactions of the downdraft
outowinduced by falling precipitation fromthe rst cycle cloud in the boundary layer and the
southeasterly relative inow at low level.
The model results show that the cold microphysical processes dominate the hydrometeor
production in the simulated storms. The ice hydrometeors account for ~70% of the total
hydrometeor mass. Accretion of cloud water is the dominant growth mechanism for precipitating
ice hydrometeors. Melting of graupel and accretion of cloud water by rain are the major sources of
rain water. Conversion of graupel is the largest source of hail formation, contributing about 80%.
Four seeding tests have been carried out to investigate the effects of seeding at a different release
mode (instantaneous or continuous, one grid point or several grid points), and with different
amounts of the seeding agent. All of cases are seeded in the region of the strongest updraft when
the model cloud top was passing the 10 C level at 10 min, and produce signicant effects. The
cloud seeding results in substantial increases in accumulated precipitation at the surface in all
seeded cases (by 2030%). Moreover, both rainfall and hailfall have increased due to seeding. The
most important contribution to the increase in hail is due to conversion of graupel to hail and
accretion of cloud water by hail. Increase of graupel melting and subsequent accretion of cloud
water by rain contribute mostly to rain enhancement.
The seedingenhances the unloadingeffect of precipitationmass mainly inthe formof graupel, leads
to a stronger downdraft outow and enhanced convergence in the boundary layer, further causes
the secondaryclouds toformearlier andgrowlarger. The enhancedupdraft increases the inowand
causes the cloud to process more water vapor and thereby cloud water, resulting in increase of
accretional growth of cloud water by precipitating particles, nally the precipitation enhancement.
These results indicate that silver iodide seeding could signicantly inuence the cloud
dynamics, microphysics and further precipitation of convective storms in the High Plains. The
simulation not only supports the hypothesis of dynamic seeding, but also demonstrates that the
convective cloud with a cold base but a long lifetime has dynamic seeding potential as well.
2009 Elsevier B.V. All rights reserved.
Keywords:
Convective clouds
Precipitation enhancement
Dynamic seeding
Weather modication
High Plains hailstorm
Atmospheric Research 96 (2010) 186207
Corresponding author. School of Atmospheric Sciences, Nanjing University, 22 Hankou Road, Nanjing 210093, China. Tel.: +86 25 83592575.
E-mail address: bjchen@nju.edu.cn (B. Chen).
0169-8095/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.atmosres.2009.04.001
Contents lists available at ScienceDirect
Atmospheric Research
j our nal homepage: www. el sevi er. com/ l ocat e/ at mos
1. Introduction
In many places around the world, cloud seeding has been
widely used for rain enhancement and hail suppression.
Among precipitating cloud systems, convective clouds are
some of the major targets for cloud seeding. There are two
major methods of convective cloud seeding: glaciogenic
seeding and hygroscopic seeding. The scientic basis for
increasing precipitation fromconvective clouds byglaciogenic
seeding rests on two concepts, namely, the static and dynamic
seeding concepts (Braham, 1986).
The fundamental concept of the static seeding is to
increase the efciency of precipitation formation in clouds
by introducing an optimum concentration of ice crystal by
cloud seeding. While the main object of the dynamic seeding
is to enhance the vertical motion in clouds and thereby
vertically process more water through the clouds resulting in
increased precipitation.
The hypothesized chain of physical responses to dynamic
seeding in earlier experiments has been summarized by
Woodley et al. (1982). But few of the hypothesized steps in
the chain of events have been measured in experiments or
veried and validated by numerical models (Orville, 1996). To
explain the less-than-expected increases in cloud-top heights,
Rosenfeld and Woodley (1993, hereafter RW93) proposed a
modied conceptual model of dynamic seeding. The revised
conceptual model gives much more attention to microphysi-
cal processes than before. It involves the production and
sustenance of greater precipitation mass at and above the
seeded region, which allows more time for continued growth
of the cloud. The subsequent unloading of this enhanced
water mass increases the downdraft and precipitation while
at the same time allowing for additional growth in the region
that retains some of the previously released latent heat.
Furthermore, RW93 note that the modied conceptual model
applies to convective clouds in which the coalescence process
is active to produce rain drops in the supercooled region.
However, this is also a proposed hypothesis of dynamic
seeding that has not been veried.
Although the concept of dynamic seeding is a physically
plausible hypothesis that offers the opportunity to increase
rainfall by much larger amounts than simply enhancing the
precipitation efciency of a cloud, this method remains as yet
an unproven technology for increasing rainfall for water
resources (Bruintjes, 1999; Silverman, 2001). It is still a critical
issue in weather modication that the hypothesis of dynamic
seeding needs to be validated and supported (NRC, 2003).
Numerical cloud models are important tool for weather
modication research (Orville, 1996; Garstang et al., 2005).
Orville (1996) provided a comprehensive overview of the use
of cloud models in the eld of weather modication. During
the past decades, great progress has been made in the eld of
cloud modeling. Guo and Huang (2002) simulated the hail
formation and growth mechanism in a multicellular storm
using a three-dimensional hail category model. Farley et al.
(2004a,b) examined the relative inuence of warm rain
process on precipitation development and hail formation in
thunderstorms and seeding effects using a two-dimensional
hail category model. Lin et al. (2005) investigated the
differences in microphysical structures of summer thunder-
storms in the humid subtropics versus High Plains using a
three-dimensional cloudmodel. Kuhlmanet al. (2006) made a
successful simulation of the 29 June 2000 tornadic supercell
stormusing a three-dimensional cloud model with electrica-
tion mechanisms. uri et al. (2006, 2007, 2008) investigated
the dispersal of seeding agent within convective cloud and the
effects of silver-iodide seeding on precipitation using a cloud-
resolving mesoscale model. And their simulations showed
that a silver-iodide seeding resulted in enhanced precipitation
over 100 km downwind from the initial point of release.
Despite signicant advances in numerical simulation of
convective cloud structures and seeding effects, modeling
studies with respect to verifying the hypothesis of dynamic
seeding in three-dimensional cloud model have not be done
yet.
In this work we investigate the effect of Silver Iodide (AgI)
seeding on the cloud microphysics, dynamics and precipita-
tion of convective clouds in the High Plains using a three-
dimensional cloud model with bulk microphysical scheme.
The primary purpose of this study is 1) to examine whether
cloud seeding can increase the precipitation of convective
clouds with a cold base in the high plains region; 2) to verify
and validate the hypothesis of dynamic seeding. Generally,
the demand for fresh water resources is very urgent in some
high plains region around the world, for instance, in the west
of the United States and northwest of China. Previous studies
indicated that the High Plains convective clouds might be not
suitable for static-mode seeding due to the limited lifetime
and low liquid water content of clouds (Cooper and Lawson,
1984; Smith and Coauthors, 1984). Such clouds may also have
no dynamic seeding potential because of the lack of abundant
supercooled raindrops as pointed out by RW93. Note that the
retention of increased precipitation mass in the cloud is an
important aspect of RW93 conceptual model. If the cloud
lifetime is long enough to allow the seeding-increased ice
particles growing in the supercooled region longer, resulting
in the retention of precipitation mass in the cloud, can the
dynamic seeding effect be accomplished?
The paper is organized as follows. Section 2 describes the
cloud model used for the present study. The results of
numerical simulation including the unseeded and seeded
runs are presented in Section 3. Discussions of the results and
conclusions are presented in Sections 4 and 5, respectively.
2. Model description
2.1. Convective cloud model
A three-dimensional compressible nonhydrostatic cloud
model with two-moment bulk microphysics scheme (Kong et
al., 1990; Hong and Fan, 1999; Xiao et al., 2004) was used to
simulate the hailstorm. The nineteen predictive variables
include the three Cartesian velocity components, the Exner
function, potential temperature, mixing ratios of water vapor
(v), cloud water (c), rain water (r), cloud ice (i), snow (s),
graupel (g), frozen drop (f) and hail (h), and number
concentrations of raindrop, cloud ice, snow, graupel, frozen
drop and hail. The basic equations in standard Cartesian
coordinates (x, y, z) are:
du
dt
+ C
p

v
AV
Ax
= D
u
1
187 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
dv
dt
+ C
p

v
AV
Ay
= D
v
2
dw
dt
+ C
p

v
AV
Az
= g
V

+ 0:608qV
v
q
c
+ q
r
+ q
i
+ q
s
+ q
g
+ q
f
+ q
h
_ _
+ D
w
3
dV
dt
+
C
2
C
p

2
v
A
v
u
j
Ax
j
=
R
d
C
v
V
Au
j
Ax
j
+
C
2
C
p

2
v
d
v
dt
+ D

4
d
dt
= Q
lv
+ Q
il
+ Q
vi
+ D

5
dq
x
dt
= S
qx
+ D
qx
+
1

A
Az
q
x
V
x
6
dN
x
dt
= S
Nx
+ D
Nx
+
1

A
Az
N
x
V
x
7
dX
s
dt
= D
X
s
+ source + sink 8
Where u, v, w, ', ,
v
and X
s
are the three velocity
components, the perturbation Exner function, potential
temperature, virtual potential temperature and the mixing
ratio of the seeding agent, respectively. q
x
is one of mixing
ratio of water vapor, cloud water, rain water, cloud ice, snow,
graupel, frozen drop and hail. N
x
is number concentration of
raindrop, cloud ice, snow, graupel, frozen drop and hail,
respectively. D
u
, D
v
, D
w
, D

, D

, D
qx
, D
Nx
and D
Xs
are the
turbulent uxes of u, v, w, ', , q
x
, N
x
and X
s
. Q
lv
, Q
il
and Q
vi
are the latent heating/cooling terms due to condensation/
evaporation, freezing/melting, and sublimation/deposition
produced by microphysical processes, respectively. V
x
is the
terminal velocity of hydrometeor x but omitted for water
vapor and cloud water. S
qx
and S
Nx
are denoted as cloud
microphysical processes. source in Eq. (8) represents the
seeding rate of seeding agent added at the seeding time. The
total sink term for the seeding agent is given by
sink = S
bc
+ S
ic
+ S
br
+ S
ir
+ S
dv
; 9
where S
bc
, S
ic
, S
br
and S
ir
are Brownian and inertial impact
collection rates due to cloud droplets and raindrops, respectively.
S
dv
is the activated agent particles which work as deposition
nuclei. These sink terms represent the effects of contact freezing
and condensationfreezing/deposition nucleation.
The model domain is on a standard spatially staggered
mesh system. A conventional time-splitting integration
technique, the same as that proposed by Klemp and
Wilhelmson (1978), is also used in this model. The large
time step is 5 s, while the small time step is 1 s. The spatial
difference terms are of second-order accuracy except for the
advection term that has fourth-order accuracy. All other
derivatives are evaluated with second-order centered differ-
ences. The radiation boundary conditions of Klemp and
Wilhelmson (1978) are used for the lateral boundaries
while the top and bottom boundaries are assumed as a rigid
wall. A Rayleigh friction zone is also used to absorb vertically
propagating gravity waves near the top of the domain. The
model includes a conventional rst-order closure for subgrid
turbulence and a diagnostic surface boundary layer based on
Monin-Obukhov similarity theory.
2.2. Model microphysics
Cloud ice and rain are assumed to follow gamma size
distribution. While all precipitating ice particles including
snow, graupel, frozen drops and hail, are assumed to follow
inverse exponential size distributions (Hong and Fan, 1999).
Cloud droplet is initiated by condensation and assumed to
be monodisperse. The collision and coalescence of cloud
droplets to form raindrops is parameterized following Lin
et al. (1983), but the relative dispersion is calculated from the
prescribed concentration of cloud droplets, following Grabow-
ski (1999). Herein, the number concentration of cloud droplets
is set to 1000 cm
3
, a typical value for cold-based convective
clouds of the High Plains region (e.g., Kubesh et al., 1988).
The natural cloud ice is produced through depositional
nucleation and homogeneous freezing of cloud droplets below
40 C. Once formed, cloud ice grows via the depositional and
riming processes. Ice multiplication, or the secondary ice
generation mechanism, for riming of snow and graupel/frozen
drop at temperatures between 3 C and 8 C is based on
Hallet and Mossop (1974) and is parameterized following Hu
and He (1988).
In the model, two categories of hail embryos are
simulated: frozen drops and graupel. Frozen drop can be
initiated by probabilistic freezing of raindrops, collisions
between rain and cloud ice, snow or active AgI particles
only when the raindrop diameter is greater than 1 mm. If the
raindrop diameter is smaller than 1 mm, frozen raindrop is
converted to graupel. Graupel may be also created via a
parameterized form of the Bergeron process, or by aggrega-
tion of ice crystals and snowakes. The autoconversion rate
coefcient for cloud ice to formsnow, snowto formgraupel, is
based on Lin et al. (1983).
Graupel and frozen drop convert to hail when their
diameters are greater or equal to 5 mm via autoconversion
(Hu and He, 1988). All of frozen drop, graupel and hail grow
by accretion of cloud water, cloud ice, rain and snow, or can be
melted or sublimated.
Snow can be formed by the BergeronFindeisen process
and autoconversion of cloud ice to snow. Production terms for
snowinclude various accretion terms (collisions of snowwith
cloud ice, cloud water, raindrops, and hail/graupel/frozen
drop), snow melting and sublimation/deposition.
Rainwater can be initiated by autoconversion of cloud
droplets to raindrops, melting of precipitating ice, or shedding
of excess water drops accreted by hail embryo and hail in the
wet growth regime.
Microphysical processes in the model are presented in
Appendix B. A complete description on microphysical para-
meterization can be found in Hong and Fan (1999) and Xiao
et al. (2004).
2.3. Silver iodide seeding
In the model, the possible mechanisms by which the silver
iodide can produce the ice phase include condensation
freezing/deposition nucleation and contact freezing nucleation.
188 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
Contact nucleation mechanisms are limited to inertial impact
and Brownian collection. The seeding processes included in the
model are calculated as follows (Hsie et al., 1980).
(1) Collections due to cloud droplets and raindrops are
S
bc
=
X
S
t
= 4D
S
X
S
N
c
R
c
= 9:657 10
17
X
S
N
c
T;
10
S
ic
=
X
S
t
= X
S
N
c
R
2
c
E
cs
V
C
= 3:142 10
16
X
S
N
c
;
11
S
br
=
X
S
t
= 4D
S
X
S
_

0
1
2
D
r
N
0r
exp
r
D
r
dD
r
= 12D
S
X
S
N
0r

4
r
;
12
S
ir
=
X
S
t
= X
S
E
rs
_

0

4
D
2
r
V
r
D
r
N
r
D
r
dD
r
= 3:6914 10
4

3=2
X
S
N
0r
A
vr

1=2

5:5
r
;
13
where N
c
is the number concentration of cloud droplets, N
0r
is number of raindrops per unit diameter,
r
is the slope
parameter in rain distribution (cm
1
), R
c
and V
c
are the
radius and terminal velocity of a cloud droplet (assumed to
be 10 m and 1.0 cm s
1
, respectively), E
cs
and E
rs
are the
collection efciency of cloud droplets and raindrops for the
seeding agents (assumed to be 10
4
and 0.510
4
,
respectively), and D
s
is the diffusivity of the seeding agent,
given by D
s
=kTB, B can be expressed as
B =
1 + aVd= R
S

6R
S
; 14
where d, the mean free path, is assumed to be 0.1 m, a =
0.9, R
s
, the radius of the seeding agent, is assumed to be
0.1 m. k(=1.3810
23
J K
1
) is Boltzmann's constant.
(=1.8110
5
kg m
1
s
1
) is the dynamic viscosity of air.
(2) The activated seeding agent as deposition nuclei under
saturation with respect to water is
S
DV
= m
s
dN
aD
T
dt
= w
A X
S
N
a
T
Az
= N
a
20-C ; 5 V T b 20-C
15
S
DV
= m
s
N
aD
T = X
S
N
a
T = N
a
20-C ; T z 20-C 16
where m
s
is the mass of a seeding agent. N
aD
, the number of
the seeding agents active as deposition nuclei under the
supercooling T (=T
0
T, and T
0
=0 C) condition, is
calculated as
N
aD
T = X
S
N
a
T
N
a
20-C
_ _
= m
s
; 17
where N
a
(T) is the number of nuclei active at the super-
cooling T. The activation curve of AgI is taken in agreement
with Hsie et al. (1980)
N
a
T =
exp 0:022 T
2
+ 0:88 T 3:8
_ _
; 5-C V T b 20-C
160; T z 20-C
_
18
where N
a
(T) is in unit of L
1
.
The interaction of the seeding agents with a cloud is
considered to follow the contact and deposition nucleation
processes, and only inertial impact and Brownian collection
are considered as possible mechanism for contact nucleation.
The processes included in the model are as follows.
(1) Contact freezing nucleation
NU
csi
= Q
c
N
ac
T
tN
c
= S
bc
+ S
ic

N
a
T
N
a
20-C
Q
c
N
1
c
m
1
s
;
20
NU
rse
= Q
r
N
ar
T
tN
r
= S
br
+ S
ir

N
a
T
N
a
20-C
Q
r
N
1
r
m
1
s
;
21
Where, NU
csi
and NU
rse
are the rate of cloud water
transformation to cloud ice and the rate of rain water
Fig. 1. Environmental characteristics from1330 MDT Miles City sounding for 1
August 1981: (a) temperature (solid line) and dewpoint temperature (dashed
line), and (b) wind components (m s
-1
) at the heights (km) indicated.
189 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
transformation to graupel or frozen raindrops due to contact
nucleation, respectively. t is the timestep for numerical
integration. N
r
is the number concentration of raindrops. N
ac
(T) and N
ar
(T) are the number concentrations of active ice
nuclei captured by cloud droplets and rain drops at a
supercooling T, respectively.
(2) Deposition nucleation
This process is only considered under the condition of
saturation with respect to ice. The rate of deposition
nucleation is calculated as NU
vsi
NU
vsi
= m
i0
S
dv
m
1
s
; 22
where m
i0
is the initial cloud ice mass formed by seeding
agent and assumed to be 10
12
kg.
The major assumptions used in model related to AgI
seeding are given as follows:
(i) The size distribution is assumed to be monodispersed
for AgI particles, and the radius and mass are assumed
to be 0.1 m and 2.3810
17
kg, respectively.
(ii) Only one active ice nucleus is captured by one liquid
drop for contact freezing nucleation.
(iii) The terminal velocity of AgI particles is assumed small
enough to be ignored.
(iv) The collection rates of ice hydrometeors for AgI particle
droplets are not considered.
2.4. Model initiation
The model was initialized based on the rawinsonde
sounding shown in Fig. 1 taken from Miles City, Montana on
1 August 1981 during the Cooperative Convective Precipita-
tion Experiment (CCOPE) season. The sounding shows a small
water vapor supplies with mixing ratio of 12 g kg
1
and a
cold cloud base with temperature of slightly more than 10 C.
The wind hodograph showed moderate wind shear in the
lowest 6 km (510
3
s
1
) but little shear above the 6 km
level. Note that a strongly southeasterly relative inow exists
in sub-cloud layer.
All simulations were integrated to 6600 s using a
horizontal grid spacing of 1 kmover a 36-km36-kmdomain
and a vertical grid spacing of 0.5 km over a 19-km depth.
Convection was initiated by a warm thermal bubble of 16 km
wide and 5 km deep, which was centered at 2.5 km above
ground level in a horizontally homogeneous environment.
The maximumthermal perturbation was 1.5 K in the center of
the bubble. A domain moving method that the grids translate
with the center of total hydrometeor mass is used to keep the
simulated storm within the computational domain.
3. Results
3.1. Simulation of the reference (unseeded) case
3.1.1. Temporal and spatial patterns of dynamical and
microphysical structures
The simulatedstormis strong andlong-lived. Fig. 2shows the
time series of the maximum updraft velocity in the model
domain at each level during each 1-min interval for the 110 min
integration. The natural cloud systemconsists of two maincycles
during the period of integration. The rst cycle (050 min) is
produced by a model perturbation, and the second-cycle clouds
(50110 min) are initiated by downdraft outow induced by
precipitation in the boundary layer. Note that the second-cycle
cloud also consists of two cells. The cores of the updraft for both
cycles are at about 8 km altitude and 30 to 40 C level. And
the maximumvalues for each cycle are 4246 ms
1
. That value
is similar to the 47 ms
1
derived by Doppler radar (Miller et al.,
1990).
The ice hydrometeors play a dominant role in the total
hydrometeor mass. The mass percentage of the hydrometeor,
including bothnon-precipitating (cloudwater andcloudice) and
precipitating (rain, snow, graupel, frozen drop and hail),
integrated over the entire domain with respect to time is
Fig. 2. Time vs. height sections of maximum updraft (m s
1
) and mean
temperature (C) in the cloud.
Fig. 3. Mass percentage of total condensate, integrated over the entire
simulation domain (353518.5 km
3
), for both non-precipitating hydro-
meteors (cloud water and cloud ice) and precipitating ones (rain, snow,
graupel, frozen drops and hail).
190 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
Fig. 4. Time vs. height sections of maximumcontent (g m
3
) for (a) cloud water, (b) rainwater, (c) graupel, (d) frozen drops and (e) hail in the model domain. The
mean temperature in the cloud also presents.
191 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
shown in Fig. 3. One can see that the liquid phase (cloud water
and rainwater) is far less abundant than the ice phase (cloud ice,
snow, graupel, frozen drops and hail). Overall, the percentages of
the hydrometeor in the domain for the whole integration time
are 25% for graupel, 22% for snow, 16% for cloud ice, 5% for frozen
drop and 4% for hail, while 20% for cloud water and 8% for rain.
The ice hydrometeors account for 72% of the total hydrometeor
mass.
The supercooled liquid water (SLW) accounts for 56% of total
liquid water content. But only a small fraction of the SLW is rain
water. The liquid water elds including the cloud water and rain
water are presented in Fig. 4a and b, respectively. Most of the
cloud water is above 4.0 km, at temperatures below 0 C.
Maximumcloud water content is 4.6 g m
3
at 20 min, located at
~9 km. After 40 min, the maximum value of the supercooled
cloud water decreases to 3.8 g m
3
. Rain water exists primarily
below the melting level except for the early stage of cloud
development. In detail, nearly all rain water consists of super-
cooledwater before 24min, while warmrainappears during 25
50min. After 50min, most rainis locatedbelowthemeltinglevel.
Based onthe time-averaged domain-integratedmass percentage
statistics, more than 95% of the SLW is the cloud water in the
simulation. The simulated maximumcontent of SLWat z=6 km
is 3.5 g m
3
after 50 min for the second-cycle clouds. That value
is similar to the 4 g m
3
detected by the T-28 aircraft (Kubesh
et al., 1988).
The graupel, frozen drops and hail elds are presented in
Fig. 4ce. One can see that the frozen drops and hail have two
clear cycles corresponding to the updraft. Most of graupel, frozen
drops andhail are locatedabove 4 kmlevel, especially during the
rst 22 min. After that time, some of these particles fall through
the 0 C level and melt into rain or reach to the ground. Peak
Fig. 5. Production of precipitating hydrometeors for the simulated storm: (a) frozen drops, (b) graupel, (c) hail and (d) rain. The curves are the result of the various
rates being summed over the entire domain and accumulated to the indicated time.
192 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
values are 3.3 g m
3
for graupel, 3.8 g m
3
for frozen drops and
2.4 g m
3
for hail, respectively.
Inthe simulation, theaccumulatedprecipitationpartitioning
at the surface is roughly 90.3% for rain, 8.0% for hail, 1.4% for
graupel and 0.3% for frozen drops, so that the rain precipitation
signicantly dominates the total precipitation at the surface.
Previous studies have shown that melting of graupel/hail and
shedding of rain during wet growth are the largest sources of the
rainproductionfor theHighPlains storm(e.g., Kubeshet al., 1988;
Lin et al., 2005). In this study, we have similar results as well.
Detailed microphysical processes will be presented in the
following sections.
3.1.2. Microphysical productions of precipitating hydrometeors
In order to better understand the precipitation micro-
physical processes and the effects of cloud seeding, Fig. 5
presents various sources of the precipitating hydrometeors
including frozen drops, graupel, hail and rain. Because no
snowfall exists on the ground in the simulation, the micro-
physical processes of snow are not presented herein.
The production terms for frozen drops are illustrated in
Fig. 5a. Note the largest source is accretion of cloud water (CLcf),
accounting for 47% of the total frozen drop production. Freezing
of raindrops (NUrf), contributing 17%, is the second largest
source. And accretion of rain (CLrf), is the third largest source,
averaging an 14% contribution. This three processes account for
78% of the total frozen drop production.
Similar as thesituationof thefrozendrops, thelargest source
for graupel is alsoaccretionof cloudwater (CLcg), accountingfor
60% of the total graupel production. The second largest is
accretion of snow(CLsg), contributing 18%. Deposition of water
vapor (VDvg) is the third largest source, contributing 10%.
Accretion of rain (CLrg) is signicantly less than above three
sources, contributing only 4%, because less supercooled rain is
produced.
In the simulation, the number concentrations of frozen
drops are signicantly smaller than that of graupel, maximum
values averaging 1530 m
3
for frozen drops and 7300 m
3
for
graupel. Thus, conversion of graupel is the largest source of the
hail formation, contributing 80%, while conversion of frozen
drops accounts for only 20%. The ratio of graupel to frozendrops
in the hail embryos is similar to the observations (Rasmussen
and Heymsfeild, 1987). The maximum hail concentration is
about 5 m
3
, a little higher than the observed 3 m
3
(Kubesh
et al., 1988), but both in the same order of magnitude. After
formation, growth by accretion of cloud water (CLch) rapidly
becomes a dominant mechanism for hail production (Fig. 5c).
The production terms for rain are shown in Fig. 5d. One can
see that warm microphysics including autoconversion of cloud
water (CNcr) and accretion of cloud water (CLcr) dominate the
rst 30 min when the ice-phase hydrometeors are still at the
middle and upper levels. After that time, the rain source is
gradually dominated by the melting of graupel (MLgr), frozen
drops (MLfr) and hail (MLhr). Overall, the source percentage of
the rainproduction accounts 58% for melting of graupel, 15%for
melting of frozen drops, 12% for melting of hail, while less than
1% for autoconversion of cloud water and 14% of accretion of
cloudwater. Thus, the meltingof graupel, frozendropandhail is
thelargest sources for rain, together accountingfor about 85%of
thetotal rainproduction. The accretionof cloudwater alsoplays
a very important role in the rain production.
In summary, the primary growth mechanism of the
precipitating hydrometeor in the simulated stormis accretion
of cloud water, while the primary generation mechanism of
the rain is melting of precipitating ice, especially, the melting
of graupel.
3.2. Cloud seeding simulation
3.2.1. Methods
In this subsection, four sensitivity tests have been carried
out rstly to investigate the effects of seeding at a different
release mode (instantaneous or continuous, one grid point or
several grid points), and with different amounts of the
seeding agent. Among these seeding scenarios, we further
select one that leads to signicantly enhanced precipitation to
analyze the dynamical and microphysical effects of seeding.
All of the seeded cases are seeded in the region of the
strongest updraft when the model cloud top was passing the
10 C level at 10 min. The center point of the initial seeding
agent is centered at 4 km level (~0 C). Except case A1 that is
limited to one grid point, all other three cases extend
Table 1
Seeding parameters and the differences in rain, hail and total precipitation
accumulated on the ground at 110 min with respect to the unseeded case.
Cases Seeding
time (min)
Xs0
(g/g)
Seeding
amounts(g)
TR
(kt)
TH
(kt)
TP
(kt)
Unseeded / / / 7968.0 706.6 8827.3
A1 10 1.110
10
240 1450.7
(18%)
353.7
(50%)
1934.4
(22%)
A2 10 1.610
11
245 1064.8
(13%)
298.5
(42%)
1509.6
(17%)
A3 10 1.610
10
2450 2003.3
(25%)
815.0
(115%)
3166.3
(36%)
A4 1015 1.710
11
2540 1192.7
(15%)
603.1
(85%)
2021.5
(23%)
Here, Xs0 is maximum value of the initial seeding agent mixing ratio; TR, TH
and TP are rain, hail and total precipitation, respectively. The numbers in the
parentheses are the percentage increase compared to the unseeded cases.
Fig. 6. The percentage of seeding agent in the domain versus time for case A3.
193 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
vertically and horizontally over three grid points. For cases A2,
A3 and A4, the initial distribution of the seeding agent, similar
to Hsie et al. (1980) but extended over three dimensions,
which forms a cubic block, has maximumvalues in the center
and decreases with distance outward. Cases A2 and A3 are
used to test the effects of seeding with different amounts of
agent. For cases A1, A2 and A3, the seeding agents are
assumed to be released instantaneously within 30 s. Case A4
is used to simulate the effects of continuous seeding for 300 s.
Some features of sensitivity tests are shown in Table 1.
3.2.2. General description of sensitivity tests
For all cases the accumulated precipitation at the surface
has been increased due to seeding. The greatest magnitude of
precipitation enhancement is about 36%, occurring in the case
A3 with a larger amount of seeding agent.
Case A2 is seeded at the same time and region as case A3, but
withone-tenthof theseedingagent amount whichis usedincase
A3. One can see that the total accumulated precipitation at the
surface shows a 17% increase compared to the unseeded case.
Case A1 is seeded at the same time and with almost the
same seeding agent amount as case A2, but only at one grid
point. The results indicate a 22% increase in the accumulated
precipitation due to seeding. Compared with the increase
between case A2 and the unseeded, case A1 is more effective,
Fig. 7. Accumulated precipitation on the ground as a function of time for the unseeded (solid lines) and seeded (dashed lines) runs: (a) rain, (b) frozen drops,
(c) graupel and (d) hail.
Fig. 8. Spatial distribution of total precipitation rate (mm/h) on the ground
for the unseeded (solid lines) and the seeded (dashed lines) runs, with
contour intervals of 10 mm/h starting at 10 mm/h.
194 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
Fig. 9. The differences in the hydrometeor elds (units in g m
3
) for domain averaged between the seeded and unseeded runs: (a) cloud ice, (b) cloud water,
(c) snow, (d) graupel, (e) frozen drop, (f) hail and (g) rain.
195 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
which is due to the more effective transport of the seeding
agent into a favorable region.
For case A4 the seeding agent is released continuously for
5 min at the same region as case A3. The results indicate a 23%
increase in the surface accumulated precipitation in the
seeded case. The magnitude of precipitation augmentation is
smaller than that of case A3 though with a little larger amount
of seeding agent in case A4.
These results demonstrate that the effectiveness of
seeding for precipitation enhancement is affected signi-
cantly by the intensity of seeding.
Note that the seeding increase both rain and hail
precipitation in each seeded case. The percentage increases
of rain and hail are 1325% and 42115%, respectively. In next
section, we shall answer the question what particular
microphysical and dynamical mechanisms mostly to differ-
ences between seeded and unseeded cases.
3.2.3. Case A3: analysis of cloud seeding effects
Hereafter we select test A3 for our analysis because it
simulates the greatest increase in precipitation amount
compared to other sensitivity tests.
3.2.3.1. Sinks and conservation of seeding agent. At the time of
seeding, the cloudtophas already reachedthe 14Clevel. The
seeding region is located above 4 km, the level of the updraft
core with the maximum value of about 10 m s
1
. After being
injected into the domain, the seeding agent is advected into the
supercooled region very rapidly. Fig. 6 shows clearly that most
of the seeding agent has been activated by 16 min (within
6 min). Deposition nucleation or condensationfreezing
nucleation is the most important mechanism responsible for
cloud ice production by seeding, accounting for 99.5% of the
total sink of the seeding agent. The contact nuclei are captured
primarily by cloud droplets, mainly through the Brownian
motion mechanism, although this amounts to less than 0.5%.
However, the seeding agent does not work effectively in case
A3. At the end of simulation, the total consumption of seeding
agent is 1624 g, accounting for 67% of the initial seeding
amounts. This indicates that morethan30%of theseedingagent
is decreased due to strong advection through the boundaries,
because of less than 0.2% of seeding material remaining in the
domain after 110 min. A similar situation occurs in case A2 and
A4. Compared to other three cases, conservation of the seeding
agent in case A1 is very good. The total mass of seeding agent in
the domain is only decreased about 0.1%. Therefore, compared
with the increase in surface precipitation between case A2 and
the unseeded run, case A1 is more effective, although with the
same seeding amount. This result shows that the effectiveness
of seeding is affected signicantly by the release place.
3.2.3.2. Surface precipitation comparison. The seeding even-
tually increases all kinds of precipitation amounts, although
the precipitation has decreased within scores of minutes
after seeding (Fig. 7). The seeded run at 110 min has
produced 9971.3 kt of rain, 67.1 kt of frozen drops, 433.7 kt
of graupel and 1521.6 kt of hail, while 7968.0 kt of rain, 27.5 kt
of frozen drops, 125.1 kt of graupel and 706.6 kt of hail in the
unseeded run. The accumulated precipitation on the ground
in the seeded run are increased by 25% for rain, 143% for
frozen drops, 247% for graupel and 115% for hail. The
convection in the seeded case is much stronger than that in
the unseeded case, as indicated by twice the amount of total
ice-phase precipitation, especially hail precipitation in the
seeded compared to the unseeded case.
The distribution of total precipitation rates on the ground
in the unseeded and seeded runs is shown in Fig. 8. Although
two modes of precipitation are clearly seen in both two runs,
the precipitation area in the seeded run is larger than that in
the unseeded run, especially for intense precipitation. More-
over, the precipitation on the ground has been redistributed
due to seeding, and more tends to shift southwardly. The
seeding slows down the simulated storm motion.
Note that the rain and graupel precipitation at the surface
are delayed about 1 min by seeding, this is due to compete
benecially for the available supercooled water resulting in
decreasing of graupel fallout and melting.
3.2.3.3. Variations in the hydrometeor elds and microphysical
processes. The differences inhydrometeor content for domain
averagedbetweenthe seededandunseededruns inthe time vs.
height sections are shown in Fig. 9. And the differences in the
accumulated production terms for hydrometeor vs. time are
shown in Fig. 10. Herein, contrastively insignicant terms are
not included.
Silver iodide seeding directly affects the production of cloud
ice. The amount of cloud ice has increased except for the earlier
times in the upper level of the seeded cloud (Fig. 9a). And
signicant enhancement in cloud ice appears after 50 min. The
increase in cloud ice content is mainly due to the increase in
homogenous freezing of cloud water (HFZci) and depositional
growth(VDvi). Inthe simulation, the roleof HMmultiplication
is less signicant to the cloud ice generation because the
average droplet diameter in the clouds is only about 12 m,
whichis typical of continental droplet spectra, simulatedherein
and observed by aircraft (Kubesh et al., 1988). That value is
smaller than that required in the HMmechanismwith droplet
diameter larger than 24 m (Pruppacher and Klett, 1997).
The seeded case has less cloud water content than the
unseeded case within 10 min after seeding because the
Fig. 9 (continued).
196 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
Fig. 10. The differences in accumulated production terms for hydrometeor vs. time between the seeded and unseeded runs: (a) cloud ice, (b) snow, (c) graupel,
(d) frozen drop, (e) hail and (f) rain.
197 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
increased ice crystals grow at the expense of the supercooled
cloudwater (Fig. 9b). After 50min, theamount of cloudwater at
lowlevel andmiddle level has signicantly increased, especially
at 7090 min; this is a direct result of enhanced supply of water
vapor by seeding. After 100 min, the cloud water content is
decreased slightly.
The snow content is decreased by seeding before 60 min
(Fig. 9c). This is primarily due to the loss of increasing in
accretion by graupel (CLsg, Fig. 10c). After that time, the snow
content has increased in the seeded run because of the increase
in cloud ice content, resulting in the increase in accretion and
conversion of cloud ice (CNis). However, the difference in snow
elds to produce surface precipitation is insignicant.
The differences between the seeded and unseeded runs in
graupel content and various production terms are presented in
Figs. 9d and 10c, respectively. Due to seeding, the number
concentration of graupel has increased signicantly (Fig. 11a).
But in the early seeding stage, riming growth of the graupel is
suppressed as a result of competition benecially for the
available supercooled cloud water, resulting in decrease in the
graupel content in the mid-lower level. Furthermore, the
average graupel diameter is decreased. While the graupel
content in upper level is increased because of the increases in
depositional growth and accretion of cloud ice and snow. After
50 min, the increase in accretionof cloud water directly leads to
the increase in the content and diameter of graupel. Abundant
amounts of graupel are present in the lower regions at the later
times. This increased amount of graupel in the lower portion is
favorable for rain formation because melting of graupel is the
largest source of the rain production.
Although the number concentration of frozen drop has
increased in the upper portion of the seeded run due to the
increase inaccretionof rainbycloudice as a result of the increase
in amount of the cloud ice, the frozen drop content has
signicantly decreased for the almost entire domain in the
seeded run before 80 min (Fig. 9e). This is because the decreases
in probabilistic freezing of rain and accretion of cloud water
(Fig. 10d). After 80 min, the increase in accretional growth of rain
and cloud water leads to the increase in the frozen drop content.
Both hail and rain content are less in the seeded run than in
the unseeded run before 50 min (Fig. 9fg). Due to seeding,
more graupel and frozen drops are formed. These seeding-
induced more numerous hail embryos compete benecially for
the available supercooled cloud water which has already been
reduced by the generation and growth of additional cloud ice
producedby seeding at the earlier times. As showninFig. 11, the
relative increase in total number are greater than in total mass
of the graupel and frozen drops before 50 min, resulting in
reduced size of embryo particles and the formation of less hail
in the seeded run (Fig. 10e). Most of these numerous small
embryos remain in upper parts of the cloud by the strong
updraft and less fall through the 0 C level, resulting in reduced
amounts of rainwater due to the decrease in melting of graupel
and frozen drop (Fig. 10f).
After 50 min, the hail and rainwater content have increased
in the seeded run. The increase in the hail content is mainly due
to the increase in autoconversion of graupel and accretion of
cloudwater by hail. As showninFig. 9b, the cloudwater content
Fig. 12. Time evolution of updraft and downdraft velocity maxima over the
model domain in the seeded and unseeded cases.
Fig. 11. Relative increase in domain totals of mass and number concentration
by seeding for the (a) graupel and (b) frozen drop.
198 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
is signicantly increasedafter 50 min. The earlier formationand
increased amounts of graupel in the seeded runcan growin the
plentiful supercooled cloud water region, resulting in the
formation of more hail. Subsequently, these small hailstones
grow in the plentiful supercooled cloud water region, resulting
in increased amounts of the hail. While the fallout and melting
of the enhanced amounts of graupel, hail and subsequent
accretion of cloud water result in the increase in rain content at
the later times. Farley et al. (2004a,b) reports similar results
that cloud seeding can increase both rain and hail if the clouds
process more supercooled cloud water.
In summary, the increased amounts of cloud water in the
seededcaseat the later times enhance the accretional growthof
cloud water by precipitating hydrometeors, that is, graupel,
frozen drop, hail and rain, and nally lead to the increase in
surface precipitation.
Fig. 13. Surface horizontal divergence (solid) and convergence (dashed) eld for the unseeded case at (a) 44 min and (b) 50 min, and the seeded case at (c) 44 min
and (d) 50 min.
199 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
3.2.3.4. Dynamical effects of seeding on clouds. The seeding
has a signicant effect on the cloud dynamics. Fig. 12 shows the
timeevolutionof maximumvertical velocities for theseededand
unseeded runs. The difference in the maximum vertical
velocities between the seeded and unseeded runs is not
signicant until 44 min. The average increases in maximum
updraft and downdraft caused by seeding are about 0.5 m s
1
and 2.3 ms
1
, respectively. At certain time, the increase is great.
For example, themaximumupdraft anddowndraft areincreased
by 11.3 m s
1
at 86 min and 15.7 m s
1
at 89 min, respectively.
Exclusive of the early peak from the initial perturbation, the
updraft maximaintheseededrunis 51ms
1
versus 45ms
1
in
the unseeded run, while the downdraft maxima in the seeded is
32ms
1
versus 20ms
1
intheunseededrun. Fig. 12alsoshows
that the seeding leads tomore rapidformationanddevelopment
of the convection and thereby hastens the cloud system
evolution. After 100 min, the convection in the seeded case is
much weaker than that in the unseeded case. The results
indicate that the life time of the cloud system is shortened by
seeding. Subsequently, the duration of precipitation has been
shortened, whichcanbe drawnfromthe variations of maximum
precipitation rate vs. time (gure omitted).
In this case, the loading effect of increasing precipitation
mass caused by seeding is very crucial to the secondary-cloud
growth. Compared to this, the effect of release of latent heat of
fusionon clouddynamics is negligible, due to the lowefciency
of freezing via accretion of liquid water by ice particles and
throughtheBergeroneffect at middle-lowlevel. That is why the
maximum updraft velocity has not increased signicantly
within about 30 min after seeding.
The seeded cloud has more precipitation mass than the
unseeded cloud. Subsequent unloading of the enhanced
precipitation mass causes more intense downdrafts and
interactions with the environment, which strengthens conver-
gence in the boundary layer and simulates more active
secondary convection. Fig. 13a and b shows that the seeded
cloud has much stronger divergence at the surface than the
unseeded cloud at 44 min. Note that two main regions
coexisting (denoted by D1 and D2) in both unseeded and
seeded cases, caused by the precipitation-induced downdrafts
(Fig. 14a and c). Under environment of strongly southeasterly
relative inowat lowlevel, the enhancement of the downdraft
by seeding increases the convergence at its gust front, as
illustrated in Fig. 13b and d. The enhanced convergence
simulates more active secondary cloud growth. Compared to
the unseeded case, the new cell C2 in the seeded case has
stronger updraft and produces more precipitating hydrometer,
mainly in the form of rain water. Moreover, as illustrated in
Fig. 13d, this secondary cloudgives rise to squall-line formation.
Whereas the secondary cloud formed in the unseeded case, is
much weaker and does not develop into a squall line through
the whole life time, still being the form of several isolated cells
(denotedby C1, C2andC3). Theresults indicate that the seeding
increases cell merging and cause the cloud to grow larger.
The seeding increases the inow and causes more water
vapor into the cloud. Fig. 15 shows the difference inwater vapor
mass transported upward through a given altitude in an
updraft, which is computed from
a
wq
v
dA, and
a
is air
density, w is updraft magnitude and A is area. One can see that
the total water vapor mass have increased between 50 and
100 min due to seeding, particularly in the mid-low level.
Furthermore, signicant increase is present in the lowest 4 km.
Two major ux cores formed at 64 min with a peak of
2.710
6
kg s
1
, and 92 min with a peak of 4.310
6
kg s
1
,
respectively, both centered at about 2.5 km. Fig. 15 also shows
that the water vapor mass entering the base of the cloud has
increased inthe seeded case. The augmented water vapor mass,
together with stronger updraft, causes more condensation and
leads to increased cloud water contents at the later times.
The second-cycle cloud is the dominant precipitation
producer of the simulation, contributes 78% of the total surface
precipitation for the unseeded case, and 87% for the seeded
case. Althoughthe cell C2 inboth cases caused by the rst-cycle
precipitation-induced downdrafts, the early formation of
strong convection in the seeded case leads to the earlier
formation of precipitating particles. These precipitating parti-
cles subsequently grow in the abundant cloud water available
viarimingandaccretion, resultinginmore precipitationmass in
the seeded cloud and greater developemt of the cloud, as
illustrated in Fig. 16. And the earlier formation and increased
amounts of graupel particles in the seeded case grow in the
plentiful supercooled water region resulting in the formation of
more hail in the seeded run.
4. Discussion
The results obtained here support the hypotheses that
glaciogenic seeding of convective clouds would enhance the
vertical air motion in clouds and thereby vertically process
more water vapor inux through the clouds, resulting in the
surface precipitationenhancement. These results are consistent
with the fundamental concept of dynamic cloud seeding.
Our results conrmthe signicant role of the retention of the
seeding-increased graupel mass in the cloud on the cloud
dynamics and precipitation. In the simulation, the graupel plays
adominant roleintheproductionof rainandhail. Autoconversion
andmeltingof large graupel particles are the dominant sources of
hail and rain, respectively. Although the domain-integrated
graupel mass is increased due to seeding, the size of graupel
particles has reduced within 40 min after seeding as a result of
competition for the available supercooled water, deduced from
Fig. 11. These smaller graupel particles stay in upper level of the
cloud longer by strong updraft, resulting in retention of the
graupel. Note that if the cloud is short-lived, these particles have
no sufcient time for continuing growth and further falling
through the melting level; as a result, the precipitation would be
decreased, just as showninFig. 7. Theresult afrms thendings of
Cooper and Lawson(1984) that the short-livedconvective clouds
in the high plains did have no seeding potential for increased
precipitation. For clouds in this study, with a lifetime more than
110 min long, the seeding-increased amounts of small-sized ice
particles allow the particles to reside in the supercooled region
longer and achieve greater size resulting in the increased graupel
mass (Fig. 9d). Although other precipitating ice mass has
Fig. 14. Comparison of total precipitating hydrometeor (not including cloud water and cloud ice) content (shaded, unit in g m
3
), vertical velocity (contour, in ms
1
)
and stream lines in XZ (west-east) vertical cross-sections in the unseeded case at (a) 44 min and (b) 50 min, and the seeded case at (c) 44 min and (d) 50 min. The
shaded denotes the total water content greater than 1 g m
3
.
200 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
201 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
decreasedinthe seededrunwithin40minafter seeding, the total
precipitationmass has increased as a result of signicant increase
in graupel. The eventual unloading of the increased precipitation
mass enhances the downdraft and promotes the second-cycle
clouds development rapidly, leading to stronger updraft and
more low-level water vapor entering the cloud.
The results show that the convective cloud with a cold
base but a strong updraft also has dynamic seeding potential.
Rosenfeld and Woodley (1993) suggest that their conceptual
model applies to warm-based convective clouds in which the
coalescence process is active to produce raindrops in the
supercooled region. But in the high plains region, convective
clouds are decient in supercooled raindrops; the supercooled
liquid water consists of cloud water. Therefore, the seeding
cannot rapidly produce the bulk of the fusion heat release in the
seeded region; as a result, the cloud dynamics response to the
seedingis veryslow. As showninFig. 12, it is clear that maximum
vertical velocities did not signicantly change until 30 min after
seeding. Although a small amount of raindrops exist in the
supercooled region at the early times, their contribution to the
seeding-induced change in cloud dynamics is negligible. It is the
retention of the increased precipitation mass high in the cloud
that eventually causes the cloud to produce signicant dynami-
cal effects, even though it takes a long time.
It should be noted that this study has examined only the
effects of cloudseedingonone convective cloudwithcoldbase in
the high plains. In future works we are going to investigate cloud
seeding effects on the development of convective clouds with
warmbase and abundant supercooledrainwater. Inaddition, the
hail precipitation at the surface has increased due to seeding in
the simulation. The schemes which produce positive effects for
rain enhancement and hail suppression by cloud seeding need
further study. Notwithstanding its limitation, this study can
clearly indicate that the dynamic seeding concept is a feasible
technology for increasing precipitation for water resources.
5. Conclusions
In this study we investigate the effects of silver iodide
seeding on a High Plains long-lived convective stormby using
a three-dimensional nonhydrostatic cloud model. From the
results obtained herein one can conclude the following.
1) The simulated cloud system consists of two cloud cycles
during the period of integration. The rst cloud cycle is
produced by a model initial perturbation, and the second-
cycle clouds are caused by interactions of the downdraft
outow induced by falling precipitation from the rst cycle
cloud in the boundary layer and the southeasterly relative
inow at low level. The second-cycle clouds are the
dominant precipitation producer of the simulation, con-
tributes about 80% of the total surface precipitation.
2) The results of the numerical simulation further conrmthat
the cold microphysical processes dominate the hydrome-
teor productioninthe HighPlains hailstorms. The simulated
clouds have slightly supercooled rain but abundant super-
cooled cloud water. The main growth mechanisms for
precipitating ice particle are accretion of cloud water. The
dominant rain sources are the melting of graupel and
accretion of cloud water.
3) The cloud seeding results in substantial increases in
accumulated precipitation at the surface in all seeded
cases (by 2030%). The results show that the effectiveness
of cloud seeding for augment precipitation is closely related
to the intensity of the seeding.
4) The seeding increases both rain and hail precipitation. The
most important contribution to the increase in hail is due to
conversion of graupel to hail and accretion of cloud water by
hail. Increase of graupel melting and subsequent accretionof
cloud water by rain contribute mostly to rain enhancement.
5) Numerical simulation herein support the hypothesis of
dynamic seeding that glaciogenic seeding of convective
clouds would enhance the vertical air motion in clouds and
thereby vertically process more water vapor inux through
the clouds, resulting in the precipitation enhancement. The
results also afrm that unloading of the seeding-increased
precipitation mass, mainly in the form of graupel, and
interactions with favorable environment in the boundary
layer did play very important role in the dynamic seeding
conceptual model.
6) This study indicates that the convective cloud with a cold
base and a strong updraft also has dynamic seeding
potential. But the cloud lifetime must be long enough to
remain the seeding-increased ice particles in the super-
cooled water region longer and achieve precipitating size.
Otherwise, the dynamical effect of cloud seeding could not
be accomplished.
Acknowledgments
We thank Dr. Yan Yin and Dr. Zhaoxia Hu for their helpful
discussions. Thanks also due to the reviewers for their critical
and constructive comments. This work was partially supported
by the National Natural Science Foundation of China (Grant No.
40775005, 40675005, 40875080 and 40830958) and Ministry
of Science and Technology of China (Grant No. 2006BAC12B07).
Fig. 15. Time vs. height section of the difference in areally integrated vertical
water vapor ux (10
6
kg s
1
) between the seeded and unseeded cases. The
thick solid line represents the mean cloud-base height for the seeded case,
which is made using a cloud water content threshold of 10
25
g m
3
.
202 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
F
i
g
.
1
6
.
A
s
i
n
F
i
g
.
1
4
,
e
x
c
e
p
t
f
o
r
t
h
e
u
n
s
e
e
d
e
d
c
a
s
e
a
t
(
a
)
5
6
m
i
n
,
(
b
)
6
6
m
i
n
,
(
c
)
7
6
m
i
n
a
n
d
(
d
)
8
6
m
i
n
,
a
n
d
t
h
e
s
e
e
d
e
d
c
a
s
e
a
t
(
e
)
5
6
m
i
n
,
(
f
)
6
6
m
i
n
,
(
g
)
7
6
m
i
n
a
n
d
(
h
)
8
6
m
i
n
.
203 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
F
i
g
.
1
6
(
c
o
n
t
i
n
u
e
d
)
.
204 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
Appendix A. Source and sink terms
The microphysical source/sink terms of the continuity equations of the proposed scheme are listed below.
The source and sink terms for the mass mixing ratios are
S
qv
= QVD
rv
+ QVD
hv
QVD
vc
QVD
vi
QVD
vs
QVD
vg
QNU
vi
QNU
vai
A1
S
qc
= QVD
vc
QCN
cr
QCL
cr
QHFZ
ci
QCL
ci
QCL
cs
QCL
cg
QCL
cf
QCL
ch
QNU
cai
A2
S
qr
=
QCN
cr
+ QCL
cr
+ QML
ir
+ QML
sr
+ QML
gr
+ QML
fr
+ QML
hr
QCL
ri
QCL
rs
QCL
rg
QCL
rf
QCL
rh
QNU
rg
QNU
rf
QNU
rag
QNU
raf
QVD
rv
_
A3
S
qi
=
QNU
vi
+ QHFZ
ci
+ QVD
vi
+ QIM
si
+ QIM
gi
+ QIM
fi
+ QNU
cai
+ QCL
ci
QCN
is
QCN
ig
QCL
ii
QCL
is
QCL
ig
QCL
if
QCL
ih
QCL
ri
QML
ir
_
A4
S
qs
=
QCN
is
+ QCL
ii
+ QVD
vs
+ QCL
cs
+ QCL
is
QCN
sg
QCL
sg
QCL
sf
QCL
sh
QCL
rs
QIM
si
QML
sr
_
A5
S
qg
=
QCN
ig
+ QCN
sg
+ QNU
rg
+ QNU
rag
+ QVD
vg
+ QCL
cg
+ QCL
rg
+ QCL
ig
+ QCL
sg
+ QCL
rig
+ QCL
rsg
QCN
gh
QML
gr
QIM
gi
_
A6
S
qf
=
QNU
rf
+ QNU
raf
+ QVD
vf
+ QCL
cf
+ QCL
rf
+ QCL
if
+ QCL
sf
+ QCL
rif
+ QCL
rsf
QCN
fh
QML
fr
QIM
fi
_
A7
S
qh
= QCN
gh
+ QCN
fh
+ QCL
ch
+ QCL
rh
+ QCL
ih
+ QCL
sh
QVD
hv
QML
hr
A8
The source and sink terms for the total number concentrations are
S
Nr
=
NCN
cr
+ NML
ir
+ NML
sr
+ NML
gr
+ NML
fr
+ NML
hr
+ NSH
hr
+ NSH
fr
+ NSH
gr
NVD
rv
NCL
rg
NCL
rf
NCL
rh
NCL
ri
NCL
rs
NNU
rg
NNU
rf
NNU
rag
NNU
raf
_
_
_
A9
S
Ni
=
NNU
vi
+ NHFZ
ci
+ NIM
si
+ NIM
gi
+ NIM
fi
+ NNU
vai
+ NNU
cai
NCN
is
NCN
ig
NVD
vi
NCL
ii
NCL
ri
NCL
is
NCL
ig
NCL
if
NCL
ih
NML
ir
_
A10
S
Ns
= NCN
is
+ NCL
ii
NCL
ss
NVD
vs
NCL
sg
NCL
sf
NCL
sh
NCL
rs
NCN
sg
NML
sr
A11
S
Ng
=
NCN
ig
+ NCN
sg
+ NNU
rg
+ NCL
rig
+ NCL
rsg
+ NNU
rag
NIM
gi
NCN
gh
NVD
vg
NML
gr
_
A12
S
Nf
= NNU
rf
+ NCL
rif
+ NCL
rsf
+ NNU
raf
NIM
fi
NCN
fh
NVD
vf
NML
fr
A13
S
Nh
= NCN
gh
+ NCN
fh
NVD
hv
NML
hr
A14
The source and sink terms for the temperature are
S

=
L
f
QIM
si
+ QIM
gi
+ QIM
fi
+ QNU
rg
+ QNU
rf
+ QHFZ
ci
+ QNU
rag
+ QNU
raf
+ QNU
cai
+ QCL
ci
+ QCL
cs
+ QCL
cg
+ QCL
cf
+ QCL
ch
+ QCL
ri
+ QCL
rs
+ QCL
rg
+ QCL
rf
+ QCL
rh
QML
ir
QML
sr
QML
gr
QML
fr
QML
hr
_
_
_
_
+ L
s
QNU
vi
+ QVD
vi
+ QVD
vs
+ QVD
vg
+ QVD
vf
QVD
hv
+ QNU
vai
_ _
+ L
v
QVD
vc
QVD
rv

_

_
A15
where L
v
, L
s
and L
f
are latent heat of condensation, sublimation and fusion, respectively.
Appendix B. Denition of acronyms for microphysical processes
Acronym Process
VDvc Condensation/Evaporation to/from cloud water
VDrv Evaporation of rain
CNcr Autoconversion of cloud water to rain
CLcr Accretion of cloud water by rain
NUvi Nucleation of cloud ice
(continued on next page)
205 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
References
Braham Jr., R.R., 1986. Precipitation enhancementa scientic challenge.
Precipitation EnhancementA Scientic Challenge, Meteor. Monogr.,
No. 43, Amer. Meteor. Soc., pp. 15.
Bruintjes, R.T., 1999. A review of cloud seeding experiments to enhance
precipitation and some new prospects. Bull. Am. Meteorol. Soc. 80, 805820.
Cooper, W.A., Lawson, R.P., 1984. Physical interpretation of results from the
HIPLEX-1 experiment. J. Clim. Appl. Meteorol. 23, 523540.
uri, M., Janc, D., Vukovi, V., 2006. Seeding agent dispersion within
convective cloud as simulated by a 3-D numerical model. Meteorol.
Atmos. Phys. 92, 205216.
uri, M., Janc, D., Vukovi, V., 2007. Cloud seeding impact on precipitation
as revealed by cloud-resolving mesoscale model. Meteorol. Atmos. Phys.
95, 179193.
uri, M., Janc, D., Vukovi, V., 2008. Precipitation change from a
cumulonimbus cloud downwind of a seeded target area. J. Geophys.
Res. 113, D11215. doi:10.1029/2007JD009483.
Farley, R.D., Wu, T., Orville, H.D., Hjelmfelt, M.R., 2004a. Numerical simulation
of hail formation in the 28 June 1989 Bismarck thunderstorm: Part I.
Studies related to hail production. Atmos. Res. 71, 5179.
Farley, R.D., Chen, H., Orville, H.D., Hjelmfelt, M.R., 2004b. Numerical
simulation of hail formation in the 28 June 1989 Bismarck thunderstorm:
Part II. Cloud seeding results. Atmos. Res. 71, 81113.
Garstang, M., Bruintjes, R., Seran, R., Orville, H., Boe, B., Cotton, W.,
Warburton, J., 2005. Weather modication: nding common ground.
Bull. Am. Meteorol. Soc. 86, 647655.
Grabowski, W.W., 1999. Aparameterizationof cloudmicrophysics for long-term
cloud-resolving modeling of tropical convection. Atmos. Res. 52, 1741.
Guo, X.L., Huang, M.Y., 2002. Hail formation and growth in a 3D cloud model
with hail-bin microphysics. Atmos. Res. 63, 5999.
Hallet, J., Mossop, S.C., 1974. Production of secondary ice particles during the
riming process. Nature 249, 2628.
Hong, Y.C., Fan, P., 1999. Numerical simulation study of hail cloud. Part I: The
numerical model. Acta Meteorol. Sin. 13, 188199.
Hu, Z., He, G., 1988. Numerical simulation of microphysical processes in
cumulonimbus. Part I: Microphysical model. Acta Meteorol. Sin. 2, 471489.
Hsie, E.Y., Farley, R.D., Orville, H.D., 1980. Numerical simulation of ice-phase
convective cloud seeding. J. Appl. Meteorol. 19, 950977.
Klemp, J.B., Wilhelmson, R.B., 1978. The simulation of three-dimensional
convective storm dynamics. J. Atmos. Sci. 35, 10701096.
Kong, F.Y., Huang, M.Y., Xu, H.Y., 1990. Three-dimensional numerical
simulation of ice phase microphysics in cumulus clouds. Part I: Model
establishment and ice phase parameterization. Chin. J. Atmos. Sci. 14,
441453.
Kubesh, R.J., Musil, D.J., Farley, R.D., Orville, H.D., 1988. The 1 August 1981
CCOPE storm: observations and modeling results. J. Appl. Meteorol. 27,
216243.
(continued)
Acronym Process
IMsi Secondary production of cloud ice from snow
IMgi Secondary production of cloud ice from graupel
IM Secondary production of cloud ice from frozen drop
HFZci Homogeneous freezing of cloud water to cloud ice
VDvi Vapor Deposition/Sublimation to/from cloud ice
VDvs Vapor Deposition/Sublimation to/from snow
VDvg Vapor Deposition/Sublimation to/from graupel
VDvf Vapor Deposition/Sublimation to/from frozen drop
VDhv Vapor sublimation from hail
MLir Melting of cloud ice to rain water
MLsr Melting of snow to rain water
MLgr Melting of graupel to rain water
MLfr Melting of frozen drop to rain water
MLhr Melting of hail to rain water
CLci Accretion of cloud water by cloud ice
CLcs Accretion of cloud water by snow
CLcg Accretion of cloud water by graupel
CLcf Accretion of cloud water by frozen drop
CLch Accretion of cloud water by hail
NUrg Probabilistic freezing of rain to form graupel
NUrf Probabilistic freezing of rain to form frozen drop
CLri(g,f) Accretion of rain by cloud ice to form graupel or frozen drop
CLrh Accretion of rain by hail
CLrf Accretion of rain by frozen drop
CLrg Accretion of rain by graupel
CLrs(g,f) Accretion of rain by snow to form graupel or frozen drop
CLii Accretion of cloud ice by cloud ice to snow
CLis Accretion of cloud ice by snow
CLig Accretion of cloud ice by graupel
CLif Accretion of cloud ice by frozen drop
CLih Accretion of cloud ice by hail
CLss Aggregation of snow
CLsg Accretion of snow by graupel
CLsf Accretion of snow by frozen drop
CLsh Accretion of snow by hail
SHgr Rain water shed from graupel
SHfr Rain water shed from frozen drop
SHhr Rain water shed from hail
CNis Autoconversion of cloud ice to snow
CNig Autoconversion of cloud ice to graupel
CNsg Autoconversion of snow to graupel
CNgh Autoconversion of graupel to hail
CNfh Autoconversion of frozen drop to hail
NUvai Vapor deposition to cloud ice on AgI particles
NUcai Cloud water transformation to cloud ice due to contact freezing with AgI particles
NUrag(f) Rain water transformation to graupel (frozen drop) due to contact freezing with AgI particles
Appendix B. (continued)
206 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207
Kuhlman, K.M., Ziegler, C.L., Mansell, E.R., Macgorman, D.R., Straka, J.M., 2006.
Numerically simulated electrication and lightning of the 29 June 2000
STEPS supercell storm. Mon. Weather Rev. 134, 27342757.
Lin, H., Farley, R.D., Orville, H.D., 1983. Bulk parameterization of the snoweld
in a cloud model. J. Clim. Appl. Meteorol. 22, 10651092.
Lin, H., Wang, P.K., Schlesinger, R.E., 2005. Three-dimensional nonhydrostatic
simulations of summer thunderstorms in the humid subtropics versus
High Plains. Atmos. Res. 78, 103145.
Miller, L.J., Tuttle, J.D., Foote, G.B., 1990. Precipitation production in a large
Montana hailstorm: airow and particle growth trajectories. J. Atmos.
Sci. 47, 16191646.
NRC (National Research Council), 2003. Critical Issues in Weather Modica-
tion Research. National Academy Press, Washington, D. C. 123 pp.
Orville, H.D., 1996. A review of cloud modeling in weather modication. Bull.
Am. Meteorol. Soc. 77, 15351555.
Pruppacher, H.R., Klett, J.D., 1997. Microphysics of Clouds and Precipitation.
Reidel, Dordrecht, p. 954.
Rasmussen, R.M., Heymsfeild, A.J., 1987. Melting and shedding of graupel and
hail. Part III: Investigation of the role of shed drops as hail embryos in the
1 August CCOPE severe storm. J. Atmos. Sci. 44, 27832803.
Rosenfeld, D., Woodley, W.L., 1993. Effects of cloud seeding in West Texas:
additional results and new insights. J. Appl. Meteorol. 32, 18481866.
Silverman, B.A., 2001. A critical assessment of glaciogenic seeding of
convective clouds for rain enhancement. Bull. Am. Meteorol. Soc. 82,
903924.
Smith, P.L., Coauthors, 1984. HIPLEX-1: experimental design and response
variables. J. Clim. Appl. Meteorol. 23, 497512.
Woodley, W.L., Jordan, J., Simpson, J., Biondini, R., Flueck, J.A., Barnston, A.,
1982. Rainfall results of the Florida Area Cumulus Experiment, 1970
1976. J. Appl. Meteorol. 21, 139164.
Xiao, H., Wang, X.B., Zhou, F.F., Hong, Y.C., Huang, M.Y., 2004. A three-
dimensional numerical simulation on microphysical processes of
torrential rainstorms (in Chinese). Chin. J. Atmos. Sci. 28, 385404.
207 B. Chen, H. Xiao / Atmospheric Research 96 (2010) 186207

Вам также может понравиться