Atmospheric Research 180 (2016) 189–199

Contents lists available at ScienceDirect

Atmospheric Research

journal homepage: www.elsevier.com/locate/atmosres

Invited review article

Latent cooling and microphysics effects in deep convection

S. Fernández-González a,⁎, P.K. Wang b, E. Gascón c, F. Valero a, J.L. Sánchez c
a
Dpto. Astrofísica y CC. de la Atmósfera, Facultad de CC Físicas, Universidad Complutense de Madrid, Ciudad Universitaria s/n, 28040 Madrid, Spain
b
Academia Sinica, Institute of Earth Sciences, Taipei, Taiwan
c
Atmospheric Physics Group, IMA, University of León, 24071 León, Spain

a r t i c l e i n f o a b s t r a c t

Article history: Water phase changes within a storm are responsible for the enhancement of convection and therefore the elon-
Received 6 April 2016 gation of its lifespan. Specifically, latent cooling absorbed during evaporation, melting and sublimation is consid-
Received in revised form 19 May 2016 ered the main cause of the intensification of downdrafts. In order to know more accurately the consequences of
Accepted 27 May 2016
latent cooling caused by each of these processes (together with microphysical effects that they induce), four sim-
Available online 29 May 2016
ulations were developed with the Wisconsin Dynamical and Microphysical Model (WISCDYMM): one with all
Keywords:
the microphysical processes; other without sublimation; melting was suppressed in the third simulation; and
Cloud resolving model evaporation was disabled in the fourth.
Latent cooling The results show that sublimation cooling is not essential to maintain the vertical currents of the storm. This is
Microphysics demonstrated by the fact that in the simulation without sublimation, maximum updrafts are in the same
range as in the control simulation, and the storm lifespan is similar or even longer. However, melting was of
vital importance. The storm in the simulation without melting dissipated prematurely, demonstrating that melt-
ing is indispensable to the enhancement of downdrafts below the freezing level and for avoiding the collapse of
low level updrafts. Perhaps the most important finding is the crucial influence of evaporative cooling above the
freezing level that maintains and enhances mid-level downdrafts in the storm. It is believed that this latent
cooling comes from the evaporation of supercooled liquid water connected with the Bergeron-Findeisen process.
Therefore, besides its influence at low levels (which was already well known), this evaporative cooling is essen-
tial to strengthen mid-level downdrafts and ultimately achieve a quasi-steady state.
© 2016 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
2. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
3. Initial conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
4. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
4.1. Control simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
4.2. No-sublimation simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
4.3. No-melt simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
4.4. No-evaporation simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
5. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
6. Concluding remarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198

1. Introduction

Severe convection can be defined as the transfer of moisture and

⁎ Corresponding author. heat through vertical currents associated with buoyancy, which can
E-mail address: sefern04@ucm.es (S. Fernández-González). cause meteorological risks such as gale-force wind gusts, lightning,

http://dx.doi.org/10.1016/j.atmosres.2016.05.022
0169-8095/© 2016 Elsevier B.V. All rights reserved.
190 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199
large hail, flash flooding, or even tornadoes (García-Ortega et al., 2014).
Accurate knowledge of severe convective storms is vital because of the
dramatic damage caused each year at mid-latitudes, a result of adverse
phenomena connected with deep convection (Czernecki et al., 2016). In
recent decades, damage caused by severe convection appears to have
increased because of climate change (Lin et al., 2005). According to
projected climate change scenarios, global temperatures may increase
as much as 2 °C by 2050 (IPCC, 2012), which could dramatically increase
damage from convective episodes in midlatitudes during coming de-
cades (Botzen et al., 2010).
Moreover, in the context of global warming, some microphysical
processes may increase in importance because of modification of atmo-
spheric thermodynamic behavior (Wang et al., 2010), making the study
of the contributions of individual microphysical processes critical. For
instance, an increase in altitude of the 0 °C isotherm could augment
the importance of collision-coalescence, which is responsible for the
formation of rain in warm clouds (Lin et al., 2005).
The impacts of severe convective episodes can be minimized by ac-
curate forecasts, even saving lives and preventing economic loss
(Bauer et al., 2015). In the present research, we analyzed the influence
of sublimation, evaporation, and melting processes on a supercell
storm on the High Plains of the USA during 1981. This storm has been
accurately analyzed in previous studies (Johnson et al., 1993; Wang
et al., 2010), which allows us to compare the results of our simulations.
Our investigation is original in that microphysical effects are also evalu-
ated, complimenting several studies of latent cooling effects (e.g., Yang
and Houze, 1995; Wang et al., 2010). The microphysical effects are
assessed using the 3D Wisconsin Dynamical and Microphysical Model
(WISCDYMM), which allows a clear visualization of these effects.
Precipitation originating from convective cells in midlatitudes is
mainly from cold rain processes (Szyrmer and Zawadzki, 1999). The
dynamics within a storm are influenced by water phase changes. For
instance, according to Liu et al. (1997), latent heat absorbed during
melting, evaporation, and sublimation strengthens downdrafts, while
updrafts are fortified by condensation, freezing, and deposition.
These effects are stronger in continental clouds than in maritime ones,
because cold rain processes are more predominant in the former
(Phillips et al., 2007).
Latent heat released and absorbed during water-phase change pro-
cesses can strengthen storms. In warm rain processes, there are only
evaporation-condensation processes. In cold clouds of deep convection,
the processes of melting-freezing and sublimation-deposition also
occur (Li et al., 2013). When warm air rises and water vapor condenses,
latent heat is released into the atmosphere, increasing instability and
strengthening updrafts. The same applies when latent heat is released
during the freezing of liquid droplets. Latent cooling absorbed during
sublimation, evaporation and melting processes can increase negative
buoyancy and strengthen downdrafts, which also increases instability
and can intensify updrafts (Szeto and Stewart, 1997). These phenomena
are why latent heat is considered the main driving force of vertical
currents within a storm.
In addition to dynamic and thermodynamic effects of evaporation,
sublimation and melting, microphysical effects are important to under-
stand the behavior of convective episodes (Kraut, 2015). Aerosols emit-
ted to the atmosphere by natural and anthropogenic sources may act as
CCN or ice nuclei, favoring the formation of cloud drops and ice crystals
(Altaratz et al., 2014). Consequently, a higher concentration of CCN in-
creases the number of cloud droplets but decreases their mean size, re-
ducing the efficiency of collision-coalescence (Loftus and Cotton, 2014).
In addition, environments with high aerosol concentrations may favor
the development of stronger updrafts and higher cloud-top altitudes
(Khain et al., 2005). Moreover, the origin of CCN can alter the behavior
of convection. For instance, continental CCN leads to stronger updrafts
and heavier precipitation than maritime CCN (Seifert and Beheng,
2006). Furthermore, the temperature at which ice nuclei become active
depends on the type of aerosol (Wang, 2013). Changes in the
distribution of aerosols can promote nucleation at relatively high tem-
peratures (near 0 °C) or hinder nucleation even at low temperatures
(−10 °C) (Kumjian et al., 2012). Because of the importance of all factors
mentioned above, there is a desire to improve the accuracy of micro-
physical parameterizations in numerical models (Hazra et al., 2016).
Moreover, the different terminal velocities of distinct hydrometeors
may alter the distance between the principal updraft and downdraft,
which is connected with the fall of precipitation. Therefore, the rate
and type of precipitation can modify the vertical currents. When hydro-
meteors are too heavy, they may descend very close to the updraft and
even cause its collapse (Zeng et al., 2001; Wang et al., 2010).
This paper is organized as follows. In Section 2, there is a brief de-
scription of the WISCDYMM. Section 3 describes the initial conditions
used in the simulations. Results of four simulations are addressed in
Section 4, together with an exhaustive analysis of the thermodynamic
and microphysical causes of the differences between the simulations.
Finally, a discussion of the results and concluding remarks are in
Section 5.
2. Methodology
This study was developed using the WISCDYMM, which is a 3D,
single-moment cloud model used mainly for the study of microphysics
and dynamics of convective clouds. This model was described by Straka
(1989) and subsequently modified by others (Johnson et al., 1994;
Wang, 2003). The model uses a primitive equation, non-hydrostatic,
quasi-compressible system (Anderson et al., 1985). The advection
schemes of finite differences and boundary conditions used by Lin
et al. (2005) were selected, with settings for subgrid-scale features of
the flow defined in Straka (1989). Radiation, the Coriolis force, and to-
pography were ignored in the modeling.
The grid was configured with a horizontal resolution of 1 km and
vertical resolution 200 m. A horizontal domain of 100 × 100 km and
20 km vertically was established. The temporal resolution was 2 s.
Total simulation time was 150 min, with data files saved every 5 min.
The simulation was reinitialized every 30 min to maintain the convec-
tion within the domain, by subtracting the storm translation speed.
This configuration has been selected according to the results of previous
studies using the same model (Johnson et al., 1994; Wang et al., 2010).
There are 38 microphysical processes incorporated in the
WISCDYMM, including nucleation, condensation, evaporation, freezing,
melting, sublimation, deposition, autoconversion, collision-coalescence,
aggregation, and riming. The model is able to predict the three wind
components, potential temperature, turbulent kinetic energy, pressure
deviation, and mixing ratios of water vapor, cloud water, raindrops,
cloud ice, snow, and graupel/hail.
Data assimilation of the WISCDYMM model simulations has two
components. First, it requires data from a pre-storm radiosonde, taken
before the formation of condensate, in which surface pressure is speci-
fied. Radiosonde raw data, which have a variable vertical resolution,
are interpolated linearly (without smoothing) for grid levels established
in the model, in this case every 100 m. Second, the simulation requires
an impulse for initiating the modeled storm. A hot ellipsoidal bubble
at the bottom center of the domain model is considered an initial per-
turbation, with the same relative humidity as the base state. The bubble
had a 10-km radius and 4-km thickness. It was centered 2 km above
ground level, and had an excess of maximum potential temperature at
its center of 3.5 °C. Water vapor mixing ratio of the bubble was adjusted
to maintain relative humidity equal to its base state.
3. Initial conditions
We selected a midlatitude supercell storm on 2 August 1981 that
crossed the Cooperative Convective Precipitation Experiment (CCOPE)
observational network in southeastern Montana (Knight, 1982). This
storm has been described by several authors (Wade, 1982; Miller
 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199
Fig. 1. Sounding at 1746 MDT on 2 August 1981 in Knowlton, Montana. The dashed curve is for dew point, and the solid curve is for temperature (Wang et al., 2010).
et al., 1988; Johnson et al., 1994), exploiting the multitude of instrumen-
tation in the CCOPE network, including seven research aircraft, seven
Doppler radars, six rawinsonde stations, and 123 surface recording
stations.
For model initialization, we used a pre-storm sounding in Knowlton
(Montana) at 1746 MDT on 2 August 1981 (Fig. 1). According to this
sounding, there was a favorable environment for the development of
supercell storms over the northern High Plains of the USA (Johnson
et al., 1994). This case accurately represents a typical High Plains
supercell.
The sounding shows a very unstable air mass, as indicated by a Con-
vective Available Potential Energy (CAPE) of 3286 m2 s−2 (Musil et al.,
1991). Moreover, a potential temperature of 311.5 K and mixing ratio
of 12.0 g kg−1 were calculated on the surface (altitude: 954 m a.s.l.; sur-
face pressure: 910 hPa). The lifting condensation level was about
740 mb, with an equivalent potential temperature of 352 K. The Lifted
Index was −10 °C at 500 mb and −17 °C at 300 mb. The bulk Richard-
son number was estimated at 25, indicating a high risk of supercell
storms. In the lowest 2000 m a.s.l., there was strong veering of the
winds (N120°), which was possibly connected with the development
of the rightward-moving 2 August supercell, as explained Klemp and
Wilhelmson (1978). The 0 °C level in the environment was at
~ 3500 m a.s.l. A wind shear of 0.01 s− 1 was registered in the lowest
4000 m a.s.l.
4. Results
We developed four simulations of the aforementioned storm to
analyze effects of processes involving latent cooling. First, we ran a
control simulation with all microphysical processes. Then,
191
microphysical processes responsible for the absorption of latent heat
were removed individually. Thus, sublimation was disabled in the sec-
ond simulation, disregarding the fusion of hydrometeors into solid
phase. Third, melting was suppressed. Finally, a fourth simulation was
done without considering evaporation. Although these simulations
could be unrealistic, they are useful to reveal the effects of each micro-
physical process.
Then, WISCDYMM model outputs were displayed using Tecplot soft-
ware (www.tecplot.com), which permits the visualization of 3D model
data (Figs. 2–5). R statistical computing and graphics software was later
used for statistical analysis of the results. These results are shown in the
Figs. 6–8, to which we refer in the analysis of results.
4.1. Control simulation
First, we describe the control simulation, in which all microphysical
processes were considered. After comparing the simulation with obser-
vational data from the CCOPE network instrumentation, it was evident
that the model could accurately simulate the storm (Johnson et al.,
1994).
A peak updraft was reached around the 20 min mark, in response to
the initial perturbation. This characteristic was present in all simula-
tions. There was a subsequent decrease in intensity after the initial per-
turbation effects, and the storm began to stabilize in strength because of
internal microphysical processes. A pronounced anvil was observed
60 min after the start of simulation. It was moderately inclined, indicat-
ing the presence of heavier hydrometeors (usually graupel and hail that
have attained considerable size) that tend to precipitate when not
under the influence of updrafts.
192 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199
Fig. 2. Aerial view of estimated liquid and ice water contents from various simulations, at 60, 90, 120 and 150 min after model initialization. Temperature (in K) is superimposed on cloud
structure to clearly show the various cloud top temperatures reached in each simulation.
In Fig. 2, the images for 60 and 90 min show signs of splitting, which
is common in this kind of supercell (Rotunno, 1981). Nevertheless, the
storm did not divide into two cells, as seen in the image after 150 min.
Around 60 min after initialization, the storm appeared to reach a
quasi-steady state, with cloud top height ~ 12,500 m a.s.l. (Fig. 3) and
updrafts in excess of 40 m s−1 (Fig. 6b). However, maximum develop-
ment was reached after 100 min, when the storm cloud top reached
heights N16,000 m a.s.l. and maximum updrafts were ~60 m s−1. Max-
imum downdrafts were 30–40 m s−1, causing the heaviest surface pre-
cipitation, both rain and hail (Fig. 8).
Below the anvil, snowflakes, graupel and hail particles appeared
after reaching precipitable sizes, and begin to fall into the post-updraft
region (Fig. 3). Upon reaching the freezing level, they began to melt.
The frozen particle size, rate of descent, and ambient relative humidity
determined whether fusion was complete or the particles reached the
surface partially in solid phase. Some of these hydrometeors were car-
ried by the wind, reaching storm regions far from the main updraft,
but only those near the main downdraft managed to reach the surface.
In areas far from the main downdraft, the hydrometeors did not reach
the surface because they experienced sublimation and evaporation
after entering unsaturated regions.
At the end of simulation, the storm began to lose intensity, indicated
by a decrease in surface rain and hail, cloud hail content (Fig. 7a), and
intensity of both maximum updraft and downdraft. However, the
anvil reached its maximum extent, with a width ~60 km, because the
amount of ice and snow inside the cloud was still growing. As seen in
Fig. 3, the anvil is nearly horizontal because of the decrease in graupel/
hail amount in the cloud. However, there was an increase in lighter hy-
drometeors such as ice and snow, which were able to remain suspended
above 10,000 m a.s.l. by persistently moderate vertical currents. Liquid
Water Content (LWC) in the cloud is represented in Fig. 4. The temper-
ature (in K) has been superimposed to show the freezing level, which
was ~ 3500 m a.s.l. Most LWC was connected to the main updraft.
Large amounts of supercooled LWC appeared (Fig. 7b) because of the
strong updraft, and slow speed of heterogeneous nucleation processes
(Rogers, 1993; Fernández-González et al., 2014). Indeed, according to
Kumjian et al. (2012), the freezing of large supercooled drops can take
up to several minutes. This supercooled water was present between
 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199
Fig. 3. Front view of liquid (blue) and ice water (white) contents estimated by simulations, at 60, 90, 120 and 150 min after model initialization.
3500 and 9000 m a.s.l., i.e., between the 0 °C and −40 °C isotherms. This
is consistent with Ramachandran et al. (1996), who claimed that drop
freezing occurs at colder temperatures in storms with stronger
updrafts. Above the − 40 °C isotherm, homogeneous nucleation pro-
cesses become increasingly important and prevent the existence of
supercooled liquid water (Wang, 2013). As will be discussed in detail
below, the evaporation of supercooled liquid water is vital in generating
downdrafts at mid-levels of the cell. Below cloud base, LWC appeared in
a wider region, associated with the melting of solid hydrometeors dur-
ing their fall.
Fig. 6a presents the temporal evolution of the distribution of various
hydrometeors (in percentages). It is seen that during the first 5 min,
there are only warm rain processes. In this period, there is clear domi-
nance of cloud droplets, with percentages greater than 90%. It is also
seen that efficiency of the collision-coalescence process in raindrop pro-
duction is very low (raindrops do not reach 5% of the total mass of hy-
drometeors 5 min after the start of the simulation), and precipitation
is not observed at the surface (Fig. 8).
Subsequently, cold cloud processes initiated when the cloud top
reached heights N 12,000 m a.s.l., with a rapid increase in solid phase
hydrometeors. Fig. 6a shows that cloud ice concentration reached
its maximum around 15–20 min. Then, the ice descends and the con-
centration of snow increases by aggregation and the Bergeron-
193
Findeisen mechanism. The maximum hail concentration was also
reached around 20 min, in response to the initial perturbation. As a
result of the melting of hail during its fall, the concentration of rain-
drops increased. Then, the hail concentration stabilized around 40%,
with its main sources the processes of aggregation and riming, and
melting its main sink (Lin et al., 2005). Hail concentration in the
cloud is a good indicator, because the maximum percentage of hail
coincided with peak storm development. In the final periods, there
was a decline in hail concentration (Fig. 6a), matching a decrease
in storm intensity. At this time, the percentage of snow in the cloud
increased, which mainly accumulated in the anvil.
The most impressive characteristic of the mature stage is that ~90%
of hydrometeors in the storm were in solid phase. This large percentage
can be reached in midlatitudes, where there are typically higher
concentrations of ice-phase hydrometeors in comparison with tropical
areas (Lin and Wang, 1997).
Fig. 5 shows that in the control simulation, there was a main updraft
with vertical structure, from the surface to around 13,000–15,000 m
a.s.l. Coinciding with maximum storm development, around
105–135 min, the updraft attained heights ~ 18,000 m a.s.l. Finally, in
the last image, intensities of vertical currents begin to decrease,
confirming that the storm was entering its dissipation stage. In all
panels of Fig. 5, there is a downdraft around the main updraft,
194 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199
Fig. 4. Front view of LWC estimated by various simulations at 60, 90, 120 and 150 min after model initialization. Temperature (in K) is superimposed on cloud structure. The 0 °C isotherm is
highlighted in black to clearly visualize supercooled liquid water that appears above it.
coincident with the fall of hail. According to Johnson et al. (1994),
anticyclonic rotation facilitates separation of the updraft from the
downdraft, allowing the storm to persist. The downdraft is very power-
ful at mid and low levels, possibly associated with characteristic abun-
dant precipitation before the dissipation stage.
4.2. No-sublimation simulation
Storm lifespan is expected to be reduced by disabling the sublima-
tion process, because of reduced latent cooling in the storm. However,
as we describe below, this did not occur in the CCOPE storm simulation.
In fact, the storm appeared to slightly intensify and increase its lifespan.
Without considering sublimation, solid hydrometeors extended
over a wider region, as observed in Figs. 2 and 3. Both the anvil and
cloud region below it occupied larger areas. Similarly, LWC below the
freezing level was greater and more widespread, but above the 0° iso-
therm sublimation did not alter LWC. This finding is supported by Liu
et al. (1997), who claimed that the slow terminal velocity of ice and
snow hydrometeors is the main reason for the formation of widespread
anvils downshear of the storm, because they are transported further
horizontally during their fall. Both ice and snow hydrometeors are
favored in this simulation, because without sublimation their contents
are greater.
The storm continued in its mature stage 150 min from the start of
simulation, showing no signs of entering the dissipation stage. The
cloud top was even slightly higher than those of the control simulation,
indicative of strong instability inside the storm.
According to Yang and Houze (1995), the downdraft is initiated
above the freezing level by sublimation cooling. However, results of
the no-sublimation simulation show that the downdraft above the
freezing level does not decrease its intensity after removing the latent
cooling of sublimation. This outcome suggests that evaporative cooling
above the freezing level is important. The large amounts of supercooled
liquid water in the storm (Fig. 7b) may be the source of this latent
cooling that is responsible for downdraft intensification. Indeed,
with the Bergeron-Findeisen mechanism, when solid and liquid water
phases coexist, cloud droplets are in an unsaturated environment and
tend to evaporate (Castellano et al., 2004). According to Fig. 4, this
mechanism might have operated over the ~ 3500–9000 m a.s.l. layer,
with maximum efficiency ~5000–6000 m a.s.l. because the maximum
difference between water vapor pressure regarding ice and liquid
phase was at − 15 °C (Wang, 2013). However, water vapor molecules
 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199 195

Fig. 5. Front view of updrafts (red) above 2 m s−1 and downdrafts (blue) below −2 m s−1 estimated by the various simulations at 45, 60, 75, 90, 105, 120, 135 and 150 min after model
initialization.

tend to deposit on ice crystals, because saturation with respect to ice is
reached.
With no sublimation, ice crystals and snowflakes can be transported
over greater distances. This allows these particles to grow into snow-
flakes or graupel and hail particles away from the main updraft, without
interfering with that updraft during their descent. Thus, the main up-
draft is stable and persistent. Furthermore, from images of vertical cur-
rents after 100 min (Fig. 5), it is evident that there are more intense
secondary updrafts, further from the main updraft, than in the control
simulation. This scenario is consonant with Wang et al. (2010). Howev-
er, latent cooling from sublimation does not appear relevant in strength-
ening vertical currents, because they are similar to the control run. This
may be because the sublimation takes place in remote regions of the
cloud, far from the main vertical currents. This theory is supported by
Yang and Houze (1995), who claimed that most of the latent cooling
of sublimation is absorbed below the anvil base.
In short, more hydrometeors in solid phase are available when
the sublimation process is disabled. Therefore, fusion and subse-
quent evaporation are enhanced (because more LWC is available
below the freezing level, as seen in Fig. 4) in the downdraft region,
increasing the latent cooling absorbed at mid and low levels of the
cell, strengthening the downdraft. Moreover, there are more solid
phase particles at mid-levels, which increases riming. That process
releases latent heat, and appears to be responsible for the updraft
strengthening in this simulation. These updrafts also occupy larger
regions than in the control run.

Fig. 6. (a) Water contents of each type of hydrometeor in entire domain (in percentages). (b) Maximum updraft and downdraft (m s−1) in the storm.
196 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199

Fig. 7. (a) Hail/graupel content in entire domain (kT). (b) Supercooled LWC in entire domain (kT) over time.

4.3. No-melt simulation
The simulation with a suppressed process of fusion had the greatest
differences from the control run. The behavior after the initial distur-
bance was nearly identical through 20 min. However, when the external
forcing stopped (~20 min), the storm was unable to maintain its devel-
opment, in contrast to the control simulation.
The cloud top did not reach 13,000 m a.s.l. and progressively
descended to below 10,000 m a.s.l. 2 h after simulation start (Fig. 2),
when the storm was in its dissipation stage. The anvil acquired a
quasi-horizontal structure at ~ 90 min, showing decreasing instability
from a very early stage.
Liquid precipitation reaching the surface was virtually nonexis-
tent (Fig. 8a). However, hail was much greater than in the control
run (Fig. 8b) because, without melting, all of it reached the surface
in solid phase. This indicates that precipitation from the storm was
almost entirely from cold cloud microphysical processes, with rain
from collision-coalescence nearly nonexistent. This is consistent
with Lin and Wang (1997), who affirmed that almost all rain at the
surface during convective events comes from the melting of hail
and graupel.
Before analyzing the structure of the vertical currents shown in
Fig. 5, it should be noted that the freezing level was at an altitude
~ 3500 m a.s.l., which is very useful to explain the results. As seen
in the image for 45 min from the beginning of the simulation, vertical
currents above 3500 m a.s.l. are very similar to those of the control
run. However, below this level changes begin to appear, since verti-
cal currents do not reach the surface. These effects are more notice-
able in the subsequent images, which show that the main updraft
began growing to higher altitudes, hovering around 3000 m a.s.l.
after ~ 75 min. This reveals a clear connection to the height of the
0 °C isotherm. After 90 min, the vertical current structure was in-
clined and appeared very fragmented, indicating that the storm
had entered its dissipation stage. Vertical currents at subsequent
model times were residual. According to Zeng et al. (2001), the incli-
nation observed in Fig. 5 in the no-melt simulation after 90 min indi-
cates that stratiform microphysical processes were dominant, after
most of the hail had fallen out. This prevented the formation of
large amounts of hail.
The above results are consistent with those of Barth and Parson
(1996), who claimed that latent cooling from melting of ice-phase hy-
drometeors can intensify or self-sustain deep convection cells. A similar
theory is found in Yang and Houze (1995), who affirmed that melt
cooling does not initiate downdrafts but enhances them below the
freezing level. Moreover, latent cooling from melting strengthens
downdrafts near the surface, intensifying low-level horizontal conver-
gence and increasing instability. Furthermore, there was much less
low-level LWC without melting (Fig. 4), and LWC above the freezing

Fig. 8. Surface precipitation over entire domain (kT) during 150 min of simulation. (a) Precipitation as rain. (b) Hail amount reaching the surface.
 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199
level decreased sharply 90 min after initialization because of updraft
collapse (Fig. 5). This reduced LWC in turn decreased evaporative
cooling.
In addition to thermodynamic causes, microphysics may also help to
explain the storm dissipation in this simulation. The faster terminal ve-
locity of large hail relative to raindrops can cause massive hailfall close
to the main updraft, which can impede its ascent (as seen in Fig. 5).
This may have produced the early dissipation of the storm. This expla-
nation is consistent with Wang et al. (2010).
One possible application of this simulation can be in the forecast-
ing of storms in unsaturated environments, in which melting is
delayed (Phillips et al., 2007). According to our simulation, this
implies a smaller latent cooling rate and thereby earlier storm
dissipation.
4.4. No-evaporation simulation
The simulation without considering the evaporation process yielded
results similar to those of no melting, but with less abrupt effects. For
example, the dissipation stage began later than in the run without
melting.
In the first phase, precipitation reaching the surface as rain was
greater, because there were no evaporation losses and larger raindrops
reached the ground (Fig. 8a). At 110 min after initialization, rainfall was
about equal to that of the control simulation. However, hail precipita-
tion was very slight, especially after 100 min (Fig. 8b), when according
to the vertical currents the storm has entered in the dissipation stage.
At 90–100 min after the simulation start, the storm changed. The in-
tensity of vertical currents weakened (Fig. 5) and the cloud top height
began a decline (Fig. 3), heralding the storm dissipation stage. This
was confirmed by the anvil being largely horizontal at the end of the
simulation.
From the analysis of vertical currents shown in Fig. 5, there were
only minor changes during the first 90 min of simulation. These were
principally in the near-surface layer, where vertical currents began at
a slightly higher level than in the control simulation. This indicates
that latent cooling absorbed from the evaporation of raindrops near
the surface increased downdraft intensity and therefore updrafts in
the boundary layer.
However, in the image at 105 min (Fig. 5), there is a major change
indicated. The main downdraft had no continuity between 3500 and
9000 m a.s.l. This was the time at which intensity in the control simula-
tion increased to a maximum. However, without evaporation this was
not possible, and the storm began to lose its vertical structure. Subse-
quently, the low-level updraft disappeared without forcing by the latent
cooling of the evaporation process, and the storm entered its dissipation
stage. Then, LWC decreased at mid-levels (Fig. 4) as a result of collapse
of the updraft at low levels after 120 min (Fig. 5). Consequently,
evaporative cooling weakened in the central region of the cell, causing
collapse of the downdraft at mid-levels.
In view of these results, it appears that latent cooling from the
evaporation process is responsible for maintaining the main downdraft
at mid-levels, and is essential to sustain a quasi-steady state. According
to Lee and Feingold (2010), the formation of stronger gust fronts is
connected with downdrafts enhanced by evaporative latent cooling.
This seems to be connected with supercooled liquid water in the
convective region of the storm, associated with updrafts. Supercooled
water associated with updrafts is common (Fernández-González et al.,
2015). Because of their coexistence with ice crystals and their lower
saturation vapor pressure, supercooled droplets tend to evaporate
via the Bergeron-Findeisen mechanism because of their unsaturated
environment.
In the same way, the evaporation process is essential to maintain
vertical currents in the near-surface layer. Latent heat absorbed by melt-
ing processes (which keeps the downdraft below the freezing level) and
sublimation (which becomes important above 9000 m a.s.l., coinciding
197
approximately with the − 40 °C isotherm) is insufficient to maintain
continuity of a rotating anticyclonic downdraft.
As shown in Fig. 7a, there was more hail in the no-melt simulation
than in the control at 25–40 min after initialization. Likewise, there
was more hail in the no-evaporation run at 60–90 min after initializa-
tion. It is notable that in both the no-melt and no-evaporation simula-
tions, the storm entered its dissipation stage following this strong
production of hail. As suggested by Zeng et al. (2001), the rapid growth
of hail might shorten the lifespan of an individual storm. Thus, micro-
physics may help to explain the early storm dissipation in both the
no-melt and no-evaporation simulations. In the absence of evaporation,
drops reaching the surface are larger, and so their terminal velocities are
greater. At the same time, precipitation reaching the surface occupies a
larger area (Fig. 3). This increases the chances of raindrops falling near
the main updraft, impeding its rise in lower tropospheric layers, as
noted by Wang et al. (2010). As seen in Fig. 8a, surface rain was greater
in the no-evaporation simulation because of this reason. The negative
buoyancy associated with the descent of large raindrops might have
led to the updraft collapse and cell decay, consistent with Zeng et al.
(2001). This collapse is not as premature as in the no-melt simulation,
because it is caused by large raindrops, whose terminal velocity is
lower than those of large hail.
In all simulations, secondary cloudiness appeared in the rear of the
storm, apparently caused by low-level downdrafts (Charba, 1974). The
associated secondary convection was more developed in the simula-
tions with suppressed melting and evaporation processes, possibly be-
cause the convection was during the dissipation stage.
5. Discussion
We ran several simulations by removing individually one of the
processes responsible for latent cooling in a storm (i.e., sublimation,
melting, and evaporation). We also ran a control simulation in which
all microphysical processes were considered.
First, the storm did not have any notable changes from suppression
of the sublimation process, and it reached a quasi-steady state. There-
fore, the downdraft above the freezing level was not caused solely by
sublimation cooling. The results of the simulations suggest that evapo-
ration of supercooled liquid droplets is the main source of latent cooling
responsible for the intensification of downdrafts above the freezing
level. This evaporation is closely connected with the Bergeron-
Findeisen mechanism, in which saturation with respect to liquid is not
reached when ice crystals and supercooled water coexist, because satu-
ration vapor pressure is lower for ice than for liquid water, causing
evaporation of supercooled droplets. Melt cooling is also very important
for maintaining updrafts at low levels. Obstruction of the main updraft
at low levels is caused by the lack of melt cooling, which produces
early storm dissipation.
Our results agree with those of Yang and Houze (1995), who in-
dicated that latent cooling from evaporation and melting is more
important than that from sublimation in terms of their influence
on vertical currents and storm structure. However, the results of
our study demonstrate that sublimation cooling is not crucial in
downdraft initiation, as a downdraft still appeared above the freez-
ing level in the simulation without sublimation. However, the storm
dissipated prematurely when evaporation or melting processes
were disabled. There are two main hypotheses to explain the chang-
es observed in the simulations without melt or evaporation, as
follows.
The first change is thermodynamic. Latent cooling decreases dra-
matically in the absence of evaporation or melting, weakening the
main downdraft. Therefore, warm, moist air is not forced to rise and re-
activate the storm. This is the hypothesis of Barth and Parson (1996),
who stated that latent cooling during melt and evaporation might in-
crease or self-sustain convection. A similar theory was advanced by
Phillips et al. (2007), who explained that latent cooling from these
198 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199
processes intensifies the downdraft, helping to maintain or even inten-
sify updrafts and, thereby, the storm.
Our results show that the low-level downdraft was driven by melt-
ing and evaporative cooling, consistent with Liu et al. (1997). As claimed
by Braun and Houze (1995), strong melting in the convective region in-
creases instability within the storm. When latent cooling decreases be-
cause of the absence of melting or evaporation, instability weakens and
the storm tends to dissipate prematurely.
Upward (downward) vertical velocity is directly linked with the rate
of latent heat released (absorbed) during water-phase change (Li et al.,
2013). This explains why vertical currents weaken and, in response, the
storm dissipates prematurely, given less latent cooling because of sup-
pression of evaporation or melting. Furthermore, latent cooling mod-
ifies the vertical lapse rate, increasing atmospheric instability (Liu
et al., 1997). In the absence of melting or evaporation, latent cooling is
reduced, stabilizing the atmosphere and promoting storm dissipation.
The storm dissipated earlier in the no-melt simulation than in the
other simulations. This indicates that latent cooling generated by the
melting process is essential for strengthening of the downdraft below
the 0 °C isotherm, which in turn increases instability near the surface
and supports a powerful updraft. Although the latent heats of sublima-
tion (2.8 MJ kg−1) and evaporation (2.5 MJ kg−1) are one order of mag-
nitude greater than that of melting (0.3 MJ kg−1), the latent cooling
from these two processes are inadequate to ensure maintenance of
the vertical currents and reach a quasi-steady state. This suggests the
importance of not only the amount of heat absorbed but also the loca-
tions where these processes operate.
Several authors (Tao et al., 1991; Liu et al., 1997; Wang et al., 2010)
have claimed that the introduction of ice microphysics lengthens the
lifespan of storms. One of the causes of short lifespans in the absence
of ice microphysics is the suppression of melting. This is because that
the process causes latent cooling, which destabilizes the atmosphere
(Rauber et al., 2000), favoring storm intensification. Such a process
highlights the need to improve the accuracy of numerical models
(Halder et al., 2015).
The other explanation focuses on microphysical factors. When melt-
ing is disabled, descending hail does not melt and falls very close to the
main updraft. This may hinder updraft ascent and upstream convection,
which may be an additional cause of the early dissipation storm. Al-
though raindrops have a much faster terminal velocity than ice particles
of the same mass (Liu et al., 1997), graupel and hail particles lose weight
during melt because of the water they shed. Furthermore, drops larger
than 5 mm usually break up into smaller drops before reaching the
surface. According to the observations of Gunn and Kinzer (1949), the
terminal velocity of a 5-mm raindrop is ~9 m s−1. However, large hail
with diameters of 5 cm (which are not uncommon during deep convec-
tion in the High Plains of the USA) can reach a terminal velocity of
35 m s−1 (Wang et al., 2015). This indicates that the terminal velocity
of hailstones are much faster than those of raindrops. Therefore, during
the fallout of those stones, they are transported less horizontally than
raindrops, falling very close to the updraft and eventually causing its
collapse. This theory is consistent with Zeng et al. (2001), who claimed
that cells producing large hail have on average a shorter lifespan than
cells producing only rain, because heavy falling particles at lower levels
tends to destroy the main updraft.
Similarly, when sublimation is suppressed, lighter hydrometeors
(which disappear by sublimation in unsaturated regions of the cloud)
are transported over longer distances before precipitating. As described
in the results section, this allows the storm to have a greater extent and
the emergence of secondary currents. This enables better organization
of the vertical currents, intensifying the storm at the end of the simula-
tion and lengthening its lifespan.
Both hypotheses agree with the results of Wang et al. (2010),
who claimed that when precipitation falls very close to the main up-
draft, there is rapid weakening of the storm. However, when hydro-
meteors are transported over long distances, the descending branch
of the storm circulation is downstream, far from the ascending in-
flow of unstable air, facilitating feedback mechanisms between
them.
6. Concluding remarks
In conclusion, latent cooling caused by evaporation and melting
contribute substantially to strengthening of the downdraft. Interaction
between the cold descending and warm ascending air currents result
in the development of a positive feedback mechanism through a rotor
circulation. Therefore, latent cooling contributes to intensification of
the main updraft. In the studied storm, latent cooling from melting
was more important than that from evaporation, but this can vary
with environmental humidity. Depending on the humidity in the
study area, melt cooling will predominate in a saturated atmosphere,
whereas evaporation cooling is dominant in a dry environment. When
there is not evaporation or melting, the terminal velocity of hydrome-
teors and surface precipitation are greater. Because of this, hydrome-
teors fall more vertically, interfering with the updraft and likely
contributing to its collapse. This leads to dissipation of the storm.
Acknowledgements
This paper was supported by the following grants: TEcoAgua
(CEN20091028), METEORISK PROJECT (RTC-2014-1872-5), Granimetro
(CGL2010-15930) and MINECO (CGL2011-25327, RTC-2014-1872-5
and ESP2013-47816-C4-4P), and LE220A11-2 and LE003B009 awarded
by the Junta de Castilla and León. Special thanks to Steven Hunter and
Analisa Weston. S. Fernández-González acknowledges grant support
from the FPU program (AP 2010-2093) and Academia Sinica in Taipei
for facilities provided during a Ph.D. stay.
References
Altaratz, O., Koren, I., Remer, L.A., Hirsch, E., 2014. Review: cloud invigoration by aerosols-
coupling between microphysics and dynamics. Atmos. Res. 140-141, 38–60. http://
dx.doi.org/10.1016/j.atmosres.2014.01.009.
Anderson, J.R., Droegemeier, K.K., Wilhelmson, R.B., 1985. Simulation of the thunderstorm
subcloud environment. Preprints, 14th Conf. On Severe Local Storms. Indianapolis, IN.
Am. Meteorol. Soc. 147–150.
Barth, M.C., Parson, D.B., 1996. Microphysical processes associated with intense frontal
rainbands and the effect of evaporation and melting on frontal dynamics. J. Atmos.
Sci. 53, 1569–1586.
Bauer, P., Thorpe, A., Brunet, G., 2015. The quiet revolution of numerical weather predic-
tion. Nature 525, 47–55. http://dx.doi.org/10.1038/nature14956.
Botzen, W.J.W., Bouwer, L.M., van den Bergh, J.C.J.M., 2010. Climate change and hailstorm
damage: empirical evidence and implications for agriculture and insurance. Resour.
Energy Econ. 32, 341–362. http://dx.doi.org/10.1016/j.reseneeco.2009.10.004.
Braun, S.A., Houze, R.A., 1995. Melting and freezing in a mesoscale convective system. Q.
J. R. Meteorol. Soc. 121, 55–77.
Castellano, N.E., Avila, E.E., Saunders, C.P.R., 2004. Theoretical model of the Bergeron-
Findeisen mechanism of ice crystal growth in clouds. Atmos. Environ. 38,
6751–6761. http://dx.doi.org/10.1016/j.atmosenv.2004.09.003.
Charba, J., 1974. Applications of the gravity current model to analysis of squall-line gust
fronts. Mon. Weather Rev. 102, 140–156.
Czernecki, B., Taszarek, M., Kolendowicz, L., Konarski, J., 2016. Relationship between human
observations of thunderstorms and the PERUN lightning detection network in Poland.
Atmos. Res. 167, 118–128. http://dx.doi.org/10.1016/j.atmosres.2015.08.003.
Fernández-González, S., Valero, F., Sánchez, J.L., Gascón, E., López, L., García-Ortega, E.,
Merino, A., 2014. Observation of a freezing drizzle episode: a case study. Atmos.
Res. 149, 244–254. http://dx.doi.org/10.1016/j.atmosres.2014.06.014.
Fernández-González, S., Valero, F., Sánchez, J.L., Gascón, E., López, L., García-Ortega, E.,
Merino, A., 2015. Analysis of a seeder-feeder and freezing drizzle event. J. Geophys.
Res. 120 (9), 3984–3999. http://dx.doi.org/10.1002/2014JD022916.
García-Ortega, E., Hermida, L., Hierro, R., Merino, A., Gascón, E., Fernández-González, S.,
Sánchez, J.L., López, L., 2014. Anomalies, trends and variability in atmospheric
fields related to hailstorms in north-eastern Spain. Int. J. Climatol. 34, 3251–3263.
http://dx.doi.org/10.1002/joc.3910.
Gunn, R., Kinzer, G.D., 1949. The terminal velocity of fall for water drops in stagnant air.
J. Meteorol. 6, 243–248.
Halder, M., Hazra, A., Mukhopadhyay, P., Siingh, D., 2015. Effect of the better representa-
tion of the cloud ice-nucleation in WRF microphysics schemes: a case study of a se-
vere storm in India. Atmos. Res. 154, 155–174. http://dx.doi.org/10.1016/j.atmosres.
2014.10.022.
Hazra, A., Padmakumari, B., Maheskumar, R.S., Chen, J.-P., 2016. The effect of mineral dust
and soot aerosols on ice microphysics near the foothills of the himalayas: a numerical
 S. Fernández-González et al. / Atmospheric Research 180 (2016) 189–199
investigation. Atmos. Res. 171, 41–55. http://dx.doi.org/10.1016/j.atmosres.2015.12.
005.
IPCC, 2012. Managing the risks of extreme events and disasters to advance climate change
adaptation. In: Field, C.B., Barros, V., Stocker, T.F., Qin, D., Dokken, D.J., Ebi, K.L.,
Mastrandrea, M.D., Mach, K.J., Plattner, G.-K., Allen, S.K., Tignor, M., Midgley, P.M.
(Eds.), A Special Report of Working Groups I and II of the Intergovernmental Panel on Cli-
mate Change. Cambridge University Press, Cambridge, UK, and New York, NY, p. 582.
Johnson, D.E., Wang, P.K., Straka, J.M., 1993. Numerical simulation of the 2 August 1981
CCOPE supercell storm with and without ice microphysics. J. Appl. Meteorol. 32,
745–759.
Johnson, D.E., Wang, P.K., Straka, J.M., 1994. A study of microphysical processes in the 2
August 1981 CCOPE supercell storm. Atmos. Res. 33, 93–123.
Khain, A., Rosenfeld, D., PoKrovsky, A., 2005. Aerosol impact on the dynamics and
microphysics of deep convective clouds. Q. J. R. Meteorol. Soc. 131 (611), 2639–2663.
Klemp, J.B., Wilhelmson, R.B., 1978. Simulations of right- and left-moving storms
produced through storm splitting. J. Atmos. Sci. 35, 1097–1110.
Knight, C.A., 1982. The cooperative convective precipitation experiment (CCOPE), 18
May–7 August 1981. Bull. Am. Meteorol. Soc. 63, 386–398.
Kraut, I., 2015. Separating the Aerosol Effect in Case of a “Medicane”, 170 pp. KIT Scientific
Publishing, Karlsruhe.
Kumjian, M.R., Ganson, S.M., Ryzhkov, A.V., 2012. Freezing of raindrops in deep convec-
tive updrafts: a microphysical and polarimetric model. J. Atmos. Sci. 69, 3471–3490.
http://dx.doi.org/10.1175/JAS-D-12-067.1.
Lee, S.-S., Feingold, G., 2010. Precipitating cloud-system response to aerosol perturbations.
Geophys. Res. Lett. 37 (L23806). http://dx.doi.org/10.1029/2010GL045596.
Li, R., Min, Q., Wu, X., Fu, Y., 2013. Retrieving latent heating vertical structure from cloud
and precipitation profiles—part II: deep convective and stratiform rain processes.
J. Quant. Spectrosc. Radiat. Transf. 122, 47–63. http://dx.doi.org/10.1016/j.jqsrt.
2012.11.029.
Lin, H.-M., Wang, P.K., 1997. A numerical study of microphysical processes in the 21 June
1991 Northern Taiwan mesoscale precipitation system. Terr. Atmos. Ocean. Sci. 8,
385–404.
Lin, H.-M., Wang, P.K., Schlesinger, R.E., 2005. Three dimensional non hydrostatic simula-
tions of summer thunderstorms in the humid subtropics versus High Plains. Atmos.
Res. 78, 103–145.
Liu, C., Moncrieff, M.W., Zipser, E.J., 1997. Dynamical influence on microphysics in tropical
squall lines: a numerical study. Mon. Weather Rev. 125, 2190–2210.
Loftus, A.M., Cotton, W.R., 2014. Examination of CCN impacts on hail in a simulated
supercell storm with triple-moment hail bulk microphysics. Atmos. Res. 147-148,
183–204. http://dx.doi.org/10.1016/j.atmosres.2014.04.017.
Miller, L.J., Tuttle, J.D., Knight, C.A., 1988. Airflow and hail growth in a severe northern
high plain supercell. J. Atmos. Sci. 45, 736–762.
Musil, D.J., Christopher, S.A., Deola, R.A., Smith, P.L., 1991. Some interior observations of
southeastern Montana hailstorms. J. Appl. Meteorol. 30, 1596–1612.
Phillips, A.T.J., Pokrovsky, A., Khain, A., 2007. The influence of time-dependent melting on
the dynamics and precipitation production in maritime and continental storm clouds.
J. Atmos. Sci. 64, 338–359.
199
Ramachandran, R., Detwiler, A., Helsdon Jr., J., Smith, P.L., Bringi, V.N., 1996. Precipitation
development and electrification in Florida thunderstorm cells during convection and
precipitation/electrification project. J. Geophys. Res. 101 (D1), 1599–1620.
Rauber, R.M., Olthoff, L.S., Ramamurthy, M.K., Kunkel, K.E., 2000. The relative importance
of warm rain and melting processes in freezing precipitation events. J. Appl. Meteorol.
39, 1185–1195.
Rogers, D.C., 1993. Measurements of natural ice nuclei with a continuous flow diffusion
chamber. Atmos. Res. 29, 209–228.
Rotunno, R., 1981. On the evolution of thunderstorm rotation. Mon. Weather Rev. 109,
577–586.
Seifert, A., Beheng, K.D., 2006. A two-moment cloud microphysics parameterization for
mixed-phase clouds. Part 2: maritime vs. continental deep convective storms.
Meteorog. Atmos. Phys. 92, 67–82. http://dx.doi.org/10.1007/s00703-005-0113-3.
Straka, J.M., 1989. Hail Growth in a highly Glaciated Central High Plains Multi-Cellular
Hailstorm. Ph.D. Diss., Dept. Meteorology. Universitiy of Wisconsin, Madison (413
pp).
Szeto, K.K., Stewart, R.E., 1997. Effects of melting on frontogenesis. J. Atmos. Sci. 54,
689–702.
Szyrmer, W., Zawadzki, I., 1999. Modeling of the melting layer. Part I: dynamics and
microphysics. J. Atmos. Sci. 56, 3573–3592.
Tao, W.K., Simpson, J., Soong, S.T., 1991. Numerical simulation of a subtropical squall line
over the Taiwan Strait. Mon. Weather Rev. 119, 2699–2723.
Wade, C.G., 1982. A preliminary study of an intense thunderstorm which moved across
the CCOPE research network in southeastern Montana. Proc. Ninth Conf. On Weather
Forecasting and Analysis. Am. Meteor. Soc, Boston, MA, Seattle, WA, pp. 388–395.
Wang, P.K., 2003. Moisture plumes above thunderstorm anvils and their contributions to
cross tropopause transport of water vapor in midlatitudes. J. Geophys. Res. 108 (D6),
4194. http://dx.doi.org/10.1029/2003JD002581.
Wang, P.K., 2013. Physics and Dynamics of Clouds and Precipitation, 467 pp. Cambridge
Univ, Press, New York.
Wang, P.K., Lin, H.-M., Su, S.-H., 2010. The impact of ice microphysical processes on the life
span of a midlatitude supercell storm. Atmos. Res. 97, 450–461. http://dx.doi.org/10.
1016/j.atmosres.2010.05.006.
Wang, P.K., Chueh, C.-C., Wang, C.–.K., 2015. A numerical study of flow fields of lobed
hailstones falling in air. Atmos. Res. 160, 1–14. http://dx.doi.org/10.1016/j.atmosres.
2015.02.013.
Yang, M.-H., Houze Jr., R.A., 1995. Sensitivity of squall-line rear inflow to ice microphysics
and environmental humidity. Mon. Weather Rev. 123, 3175–3193. http://dx.doi.org/
10.1175/15200493(1995)123b3175:SOSLRIN2.0.CO;2.
Zeng, Z., Yuter, S.E., Houze, R.A., Kingsmill, D.E., 2001. Microphysics of the rapid develop-
ment of heavy convective precipitation. Mon. Weather Rev. 129 (8), 1882–1904.
http://dx.doi.org/10.1175/1520-0493(2001)129b1882:MOTRDON2.0.CO;2.