PDF - American Meteorological Society

JANUARY 2009
61
HONG ET AL.
Sensitivity Study of Cloud-Resolving Convective Simulations with WRF Using Two
Bulk Microphysical Parameterizations: Ice-Phase Microphysics versus
Sedimentation Effects
SONG-YOU HONG, KYO-SUN SUNNY LIM, JU-HYE KIM,
AND JEONG-OCK JADE
LIM*
Department of Atmospheric Sciences and Global Environment Laboratory, Yonsei University, Seoul, South Korea
JIMY DUDHIA
Mesoscale and Microscale Meteorology Division, National Center for Atmospheric Research,1 Boulder, Colorado
(Manuscript received 12 February 2008, in final form 27 June 2008)
ABSTRACT
This study examines the relative importance of ice-phase microphysics and sedimentation velocity for
hydrometeors in bulk microphysics schemes. The two bulk microphysics schemes having the same number
of prognostic water substances, the Weather Research and Forecasting (WRF) Single-Moment 6-Class
Microphysics Scheme (WSM6) and the Purdue–Lin scheme (PLIN), are evaluated for a 2D idealized storm
case and for a 3D heavy rainfall event over Korea. The relative importance of microphysics and sedimentation velocity for ice particles is illuminated by the additional experiments that exchange the sedimentation
velocity formula for graupel in the two schemes. In a 2D idealized storm simulation test bed, it is found that,
relative to the PLIN scheme, the WSM6 scheme develops the storm late with weakened intensity because of a
slower sedimentation velocity for graupel. Such a weakened intensity of precipitation also appears in a 3D
model framework when the WSM6 scheme is used, in conjunction with the overall distribution of the precipitation band southward toward what was observed. The major reason is found to be the ice-phase microphysics of the WSM6 and related ice-cloud–radiation feedback, rather than the smaller terminal velocity
for graupel in the WSM6 than in the PLIN scheme.
1. Introduction
It is well known that the simulations of many individual phenomena, ranging from tropical and extratropical cyclones to the climate variability, are sensitive
to the way convection is represented. It has also been
recognized that the water vapor content of large parts
of the atmosphere is strongly controlled by cloud and
precipitation processes. In numerical modeling of the
atmosphere, the precipitation physics component plays
* Current affiliation: Numerical Weather Prediction Center,
Korea Meteorological Administration, Seoul, South Korea.
1 The National Center for Atmospheric Research is sponsored
by the National Science Foundation.
Corresponding author address: Song-You Hong, Dept. of Atmospheric Sciences, College of Science, Yonsei University, Seoul
120-749, South Korea.
E-mail: [email protected]
DOI: 10.1175/2008JAMC1960.1
Ó 2009 American Meteorological Society
a central role in predicting weather phenomena in numerical weather prediction (NWP) and the climate signal in general circulation models (GCMs). Also, of the
many physical processes that must be represented in
numerical models of the atmosphere, precipitation
physics is generally regarded to be the most complex
and challenging task.
In the NWP and GCM areas, the precipitation from
an explicit representation of cloud and its precipitation
processes is regarded as a grid-resolvable precipitation
process, and the subgrid-scale precipitation is due to
parameterized cloud and precipitation processes from
the cumulus parameterization scheme. The cumulus parameterization scheme assumes that the model gridvolume variables represent an average of many convective cells, and that the influence of a convective cell is
confined to a grid column, which is not valid at high
resolution. The grid resolution cutoff for cumulus
schemes is unclear (Noda and Niino 2003; Bryan et al.
2003). Despite the uncertainties in the precipitation
62
JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY
physics in high-resolution grids, it is clear that accurate
representation of clouds and precipitation physics in
the grid-resolvable precipitation algorithm is a critical
factor for the improvement of precipitation forecasts in
high-resolution models.
The grid-resolvable precipitation algorithm in NWP
models commonly uses bulk parameterization methods
because of their economic treatments versus explicit
bin-resolving cloud models (Lin et al. 1983; Rutledge
and Hobbs 1984; Ferrier et al. 1995; Meyers et al. 1997;
Reisner et al. 1998). Lin et al. (1983) and Rutledge and
Hobbs (1984) have been a core part of bulk microphysical methods in representing clouds and precipitation
processes, and these reduce the number of prognostic
variables by assuming the hydrometeor size spectra to
follow a prescribed exponential (Kessler 1969) or
gamma distribution (Walko et al. 1995).
In the Advanced Research Weather Research and
Forecasting (WRF) Model (ARW; Skamarock et al.
2007) as of April 2006, bulk microphysics schemes include the Kessler, Purdue–Lin (Chen and Sun 2002),
Ferrier (new Eta), WRF-Single-Moment-Microphysics
(WSMMPs; Hong et al. 2004, hereinafter referred to as
H04; Hong and Lim 2006), and Thompson (Thompson
et al. 2004). The Kessler scheme is called a simple warm
rain scheme because it has no ice phase. The Eta Ferrier scheme predicts changes in water vapor and condensate in the forms of cloud water, rain, cloud ice, and
precipitation ice. An improved Goddard bulk microphysics parameterization having hail species as a prognostic water substance (Tao et al. 2003; Lang et al.
2007) has recently been implemented into the WRF,
version 3.0.
The WSMMPs include the WSM3, WSM5, and
WSM6 schemes, having a revised ice process treatment
of H04. The numbers at the end of WSM, that is, 3, 5
and 6, refer to the number of categories of water species, including vapor, predicted by the scheme. H04
evaluated two categories of the WSMMPs, namely, (i)
three class (WSM3) with the prognostic water substance variables of water vapor, cloud water/ice, and
rain/snow, and (ii) five class (WSM5) with water vapor,
cloud, ice, rain, and snow. In the WRF physics options,
the WSM3 and WSM5 are the revised versions of the
National Centers for Environmental Prediction
(NCEP) cloud 3 and 5, respectively. H04 concluded
that together with the sedimentation of cloud ice, the
new microphysics scheme reveals a significant improvement in the high cloud amount, surface precipitation,
and large-scale mean temperature through a better representation of the ice-cloud–radiation feedback. Further, Hong and Lim (2006) showed that the amount of
VOLUME 48
rainfall increases and the peak intensity becomes stronger as the number of hydrometeors classes increase.
At the National Center for Atmospheric Research
(NCAR), real-time forecasting experiments with a
4-km grid mesh over the central United States employed the WSM6 scheme with graupel that replaced
the Purdue–Lin scheme (PLIN) in early 2005. Both
schemes have the same amount of prognostic water
substance including graupel. Although some preliminary reports identified the overall superiority of the
WSM6 to the PLIN scheme in resolving precipitating
convective systems (e.g., Klemp 2006; Kuo 2006), reasons for the different behaviors have not been clarified.
The goal of this research is to understand the importance of microphysics, especially ice-phase microphysics processes in the bulk parameterization of clouds and
precipitation. The performance of the WSM6 microphysics will be evaluated, relative to that of the PLIN
scheme in 2D idealized and 3D real-case experiment
platforms, focusing on the major differences in the
treatment of ice properties and their sedimentation velocity. Regarding the sedimentation velocity versus microphysics in previous studies, McCumber et al. (1991),
in their simulations of tropical squall lines, found that
within a particular type of scheme, the greatest sensitivity was due to the physical parameters of hydrometeors (i.e., graupel terminal fall velocity rather than the
particular processes within a scheme), whereas several
other studies (e.g., Gilmore et al. 2004) stressed the
importance of a parameter setting in ice microphysics in
the same scheme. In this study, additional experiments
that exchange the sedimentation velocity formula for
graupel in the two schemes (PLIN versus WSM6) are
designed to further investigate the relative importance
of microphysical parameterization of ice processes and
sedimentation velocity.
Section 2 provides overall differences between the
WSM6 and PLIN schemes. In section 3, the numerical
experiments conducted in this study are described, with
their results being discussed in section 4. Concluding
remarks appear in the final section.
2. Comparison of the WSM6 and PLIN schemes
The WSM6 scheme was developed by adding additional processes related to graupel species onto the
WSM5 scheme (Hong and Lim 2006), whereas the
PLIN scheme is based on the Lin et al. with some modifications (Chen and Sun 2002). In both schemes, the
six-class prognostic water substance includes the mixing
ratios of water vapor (qV ), cloud water (qC ), cloud ice
(qI ), snow (qS ), rain (qR ), and graupel (qG ). A detailed
description of the WSM6 scheme including the produc-
JANUARY 2009
HONG ET AL.
63
FIG. 1. Flowcharts of the microphysics processes in the (a) WSM6 and (b) PLIN schemes. The terms in red (blue) are activated when
the temperature is above (below) 08C, whereas the terms in black are in the entire regime of temperature. Note that the source term
for graupel by Psacw in Hong and Lim (2006) should be corrected as for snow. The different processes between the schemes are circled
in green.
tion terms in Fig. 1 and the computational procedures
are given in Hong and Lim (2006). The names of the
processes in Fig. 1 are included in the appendix. Note
that Psacw in Hong and Lim was a source term for
graupel instead of snow in this study (see Fig. 1). This
change for Psacw introduced in this study leads to the
increase of snow and decrease of graupel but not significantly. Further, the differences do not appear to
have any impact on the results presented in this paper.
The most important difference in the two schemes is
the treatment of ice-phase microphysical processes
(Table 1). The WSM6 scheme treats the ice crystal
number concentration (N I ) as a function of cloud ice
amount (rqI ), and the ice nuclei number concentration
(N I0 ) is separated from N I , whereas the PLIN scheme
uses the formula of Fletcher (1962) for both N I and N I0 .
The Fletcher formula produces a concentration increase of a factor of 10 for about every 48C cooling. The
snow intercept parameter is a function of temperature
in the WSM6 scheme (Houze et al. 1979). Related
changes for the ice-phase microphysics are described in
H04.
In addition to the distinguishing differences in ice
microphysics devised by H04, the production and generation terms for the water substances in the two
schemes differ (Fig. 1). For example, the initial generation for ice crystals (Pigen) and the heterogeneous
freezing of cloud water (Pihtf), are absent in the PLIN,
whereas the reduction of cloud ice (Psfi) and water
(Pidw and Psfw) by Bergeron process is absent in the
WSM6 (green circles in Fig. 1). Further, as with WSM3
and WSM5, the saturation adjustment in the WSM6
scheme follows Dudhia (1989) in separately treating ice
and water saturation processes, rather than a combined
saturation such as the PLIN scheme. Also, freezing–
melting processes in the WSM6 scheme are computed
during the fall-term substeps to increase accuracy in the
vertical heating profile of these processes, whereas they
are computed on the regular time step in the PLIN
scheme.
Another apparent difference is the treatment of the
snow and graupel sedimentation. As in Hong and Lim
(2006), the mass-weighted terminal velocity for graupel
in the WSM6 scheme, V G , is given by
aG Gð4 1 bG Þ r0 1=2 1
1
VG m s
;
ð1Þ
5
6
r
lbGG
where aG and bG are the empirical coefficients for terminal velocity, lG is the slope parameter, r is the density of air, and r0 is the density of air at reference state.
The PLIN scheme also employs the same formula, but
with different coefficients, aG and bG (see Table 1).
The terminal velocity for snow takes the same formula
as in (1), but different empirical coefficients for the two
schemes, aS and bS . It is seen that the mass-weighted
terminal velocity for graupel, V G , is about twice as fast
in the PLIN scheme than in the WSM6 scheme (Fig.
2a). The terminal velocity for snow, V S , is also different,
but not significantly (Fig. 2b). It can be seen that V S for
the WSM6 scheme is relatively slow (fast) at the small
(large) mass of snow, in comparison with that in the PLIN
scheme.
64
JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY
VOLUME 48
TABLE 1. Major differences of the microphysics parameterization between the WSM6 and PLIN schemes.
WSM6
3
Ice number concentration, N I ðm Þ
Ice nuclei number, N 10 ðm3 Þ
Snow intercept parameter, n0S ðm4 Þ
Density of graupel, rG ðkg m3 Þ
Constant aG
Constant bG
Constant aS
Constant bS
7
0:75
5:38 3 10 ðrqI Þ
103 exp½0:1ðT 0 TÞ
2 3 106 exp½0:12ðT 0 TÞ
500
330
0.8
11.72
0.41
Thus, major differences in the WSM6 and PLIN
schemes can be categorized by the 1) ice-phase microphysics based on H04 and 2) terminal velocity for graupel. The relative importance of the two components in
the WSM6 and PLIN schemes will be investigated.
FIG. 2. Comparison of the mass-weighted terminal velocity for
(a) graupel and (b) snow in the WSM6 (solid) and PLIN (dashed)
schemes as a function of graupel and snow amount with the assumption that the temperature is 08C.
PLIN
2
10 exp½0:5ðT 0 TÞ
102 exp½0:5ðT 0 TÞ
3 3 106
400
82.5
0.5
4.836
0.25
3. Numerical experimental setup
The WRF is a next-generation mesoscale numerical
weather prediction system designed to serve both operational forecasting and atmospheric research needs
(http://www.wrf-model.org). It offers numerous physics
options, thus tapping into the experience of the broad
modeling community. WRF is suitable for a broad spectrum of applications across scales ranging from meters
to thousands of kilometers. The model used in this
study is the Advanced Research WRF, version 2.1.2,
(Skamarock et al. 2007), which was released in January
2006. Two sets of experiments were carried out: an idealized 2D thunderstorm case and a 3D real-data simulation of a heavy rainfall event over Korea. Readers are
referred to Hong and Lim (2006) for a more detailed
description of the experimental design, and the experimental setup will be briefly discussed below.
The 2D idealized thunderstorm experiment was designed to systematically distinguish the intrinsic differences between the WSM6 and PLIN schemes by virtue
of fixed initial conditions and the absence of other nonmicrophysical processes, which in turn would help us to
understand the impact of the changes in the microphysics in the 3D framework. Although it is recognized that
a 3D idealized simulation would be more beneficial to
examine the convective storm dynamics interacting
with the microphysics than a 2D framework, our major
concern in an idealized test bed is to clarify the direct
effects of the microphysical processes on the simulated
storm, rather than the interaction between the microphysics and storm dynamics. The idealized thunderstorm simulation is a present option for the WRF. We
chose a 2D domain in the x direction. The grid in this
direction comprised 201 points with a 250-m grid spacing. The model was integrated for 60 min with a time
step of 3 s. The initial condition included a warm bubble
with a 4-km radius and a maximum perturbation of 3 K
at the center of the domain. Open boundary conditions
were applied, and there was no Coriolis force or friction. The only physical parameterization was the microphysics scheme, and other physical processes including
JANUARY 2009
HONG ET AL.
65
FIG. 4. Model terrain contoured every 200 m for the 48-km
resolution grid (D01). The resolutions of inner domains (D02 and
D03) are 12 and 3 km, respectively. Terrain heights .1000 m are
shaded.
FIG. 3. (a) Observed 24-h accumulated precipitation (mm) valid
at 0000 UTC 15 Jul 2001 and (b) radar image of rain rate (mm
h21) at 1700 UTC 14 Jul 2001 when the maximum precipitation
intensity is observed. Shading in (a) indicates where precipitation
is .90 mm.
radiation, vertical diffusion, land surface, and deep convection due to the cumulus parameterization scheme
were turned off.
The second test case was a real-data example. A significant amount of precipitation was recorded in Korea
on 15 July 2001, with a local maximum of approximately 371.5 mm near Seoul (Fig. 3a). Most of the rainfall was observed during the 12-h period from 1200
UTC 14 July to 0000 UTC 15 July 2001, and the maximum rainfall intensity was 99.5 mm h21 (Fig. 3b). In
this study, the physics packages other than the microphysics include the Kain and Fritsch (1993) cumulus
parameterization scheme, the Noah land surface model
(Chen and Dudhia 2001), the Yonsei University planetary boundary layer (PBL; Hong et al. 2006), a simple
cloud-interactive shortwave radiation scheme (Dudhia
1989), and Rapid Radiative Transfer Model (RRTM)
longwave radiation (Mlawer et al. 1997) scheme. The
model configuration consisted of a nested domain defined on a Lambert conformal projection. A 3-km
model covering the Korean Peninsula (domain 3, 317 3
293) was surrounded by a 12-km grid model (domain 2,
141 3 137), which in turn was surrounded by a 48-km
grid model (domain 1, 76 3 76) by a one-way interaction (Fig. 4). The experiments were carried out for 24 h,
from 0000 UTC 14 July to 0000 UTC 15 July 2001.
No cumulus parameterization was used in the 3-km
grid model since at that resolution, updrafts may be
resolved sufficiently so as to result in explicit convective
vertical transports. A 3-km grid spacing may not be
really sufficient to resolve the updrafts as shown by
Bryan et al. (2003), but the grid resolution cutoff for the
usage of the cumulus parameterization scheme can vary
depending on the characteristics of a system. For example, Hong (2004) selected this heavy rainfall case
66
JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY
VOLUME 48
TABLE 2. A summary of experiments conducted in this study
Expt
Description
WSM6
WSM6_vg
PLIN
PLIN_vg
WSM6_nora
Employing the WRF Single-Moment 6-class microphysics scheme
Replacing the VG in the WSM6 scheme by that in the PLIN scheme; the density of graupel is also replaced by the
value for the PLIN
Employing the Purdue–Lin microphysics scheme
Replacing the VG in the PLIN scheme by that in the WSM6 scheme; the density of graupel is also replaced by the
value for the WSM6
Excluding the effects of clouds on the radiation computation in the WSM6 scheme
over Korea for identifying differences in mechanisms
responsible for heavy rainfall occurring over geographically different regions. From the modeling studies with
different precipitation physics, Hong (2004) demon-
strated that the removal of the convective instability by
the cumulus parameterization schemes is an essential
process for heavy rainfall over the United States,
whereas it plays an insignificant role in reproducing
FIG. 5. Isolines of the condensation fields for (a) cloud ice (dotted) and water (solid) non-precipitating particles
and (b) rain (solid), snow (dotted), and graupel (dashed) precipitating particles from the WSM6 experiment. (c),
(d) Corresponding plots from the PLIN experiment. Contour lines are at 0.01, 0.02, 0.04, 0.08, 0.16, 1.28, 2.56, 5.12,
and 10.24 g kg21. All figures are 60-min snapshots. Light and dark shadings in (a) and (c) indicate the vertical velocity
of 1 and 7 m s21.
JANUARY 2009
HONG ET AL.
heavy rainfall over Korea where climatologically the
Korean Peninsula is characterized as thermodynamically neutral in contrast to large convective available
potential energy (CAPE) over the United States.
In addition to the WSM6 and PLIN experiments, another set of sensitivity experiments is carried out to
determine the relative importance of two factors in the
scheme: 1) ice-phase microphysics based on H04 and 2)
terminal velocity for graupel. Realizing the fact that
even small changes in just one of the parameters like
the hydrometeor densities or the mean size diameter
can have significant impacts on the surface precipitation amounts, and the hydrometeor distribution in the
same scheme (e.g., Gilmore et al. 2004; van den Heever
and Cotton 2004), it is very difficult to assess how the
individual parameters in the two schemes influence the
different simulation results. Therefore, the WSM6_vg
(PLIN_vg) experiment replacing the VG in the WSM6
(PLIN) scheme by that in the PLIN (WSM6) scheme is
designed to identify a major difference between the
WSM6 and PLIN schemes. The density of graupel is
also replaced by the value for the PLIN (WSM6)
scheme in the WSM6_vg (PLIN_vg) experiment. Although fall speed may be considered one aspect of microphysics, we are distinguishing the differences in the
microphysics from the differences in the fall speed. The
impact of the difference in Vs in both schemes was
found not to be significant and will not be further discussed in this study. An additional experiment,
WSM6_nora, which excludes the effects of clouds on
the radiation computation is designed to further investigate the ice-cloud–radiation feedback. A summary of
all the experiments appears in Table 2.
4. Results
a. Idealized experiments
Figure 5 compares the condensate fields from the
experiments with the WSM6 and PLIN schemes after
60-min integration, which is the mature stage of this
idealized storm in terms of the distribution of hydrometeors. The maximum updraft velocities appeared at
about 30 min with 26.4 and 25.6 m s21 for the WSM6
and PLIN runs, at 8 km in the vertical and 2 km in
negative x axis (not shown). The general structure of
the thunderstorm, such as the ice water in the updraft
region near the storm center and anvil clouds, is well
simulated with both schemes; however, a distinct difference in hydrometeors from the WSM6 scheme is less
ice substance in the PLIN scheme, especially for the
cloud ice and graupel species (Fig. 6). A possible reason
for these differences in hydrometeors between the two
67
FIG. 6. Vertical distribution of the differences in the timedomain-averaged water quantities (PLIN 2 WSM6). All fields
represent 60-min integration averages. Units: g kg21 for rain, snow,
and graupel; and 10 3 g kg21 for cloud ice and cloud.
schemes will be explained by analyzing the results of
some sensitivity experiments.
The time series of domain-averaged precipitation
and hydrometeor path is plotted in Fig. 7. The hydrometeor path is defined by the vertical integral of the
sum
R z of all condensates with respect to the height
(5 0 top rqtotal dz). It can be seen that the WSM6 scheme
develops the mature stage of the storm about 10 min
later, as compared with the PLIN scheme, although
initial development before 25 min is as fast (Fig. 7a).
The maximum intensity of precipitation is also weakened in the WSM6 experiment. To be consistent with
the evolution of surface precipitation, the amount of
hydrometeors in the atmosphere is larger in the WSM6
scheme than in the PLIN scheme after 25 min (Fig. 7b).
Meanwhile, the results from the WSM6_vg and PLIN_vg
experiments identify that both the evolution of surface
precipitation and hydrometeors are significantly affected
by the magnitude of sedimentation of graupel, rather
than differences in the ice-phase microphysics. A possible
reason for this relative importance is described below.
The immediate impact of the different sedimentation
velocity would appear in the distribution of hydrome-
68
JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY
VOLUME 48
FIG. 7. Time series of the (a) surface precipitation rate [mm (5 min)21] and (b) hydrometeor path amount (kg
m22) averaged over the domain, resulting from the WSM6 (solid), PLIN (dotted), WSM6_vg (dashed), and
PLIN_vg (dotted–dashed) experiments.
teors (Fig. 8a). Relative to the results from the
WSM6 run, the reduction of graupel above the freezing
level with the maximum at 8 km is evident in the
WSM6_vg run. The reduction of other hydrometeors is
not as distinct as in the graupel, but still visible, and can
be attributed to enhanced accretion of them by graupel.
A detailed analysis of each source–sink term in the
WSM6 scheme identified that the accretion process of
cloud water by graupel (Pgacw) is the dominant process
for graupel formation in the convective system when
the temperature is below 08C, consistent with the results of Wang et al. (2007). Thus, a relative cooling
above the freezing level in the case of the WSM6_vg run
(Fig. 8c) can be attributed to the reduction of latent
heat release from freezing in the Pgacw term at the
later stages of storm development because of the reduced graupel aloft.
It is also seen that in the case of the WSM6_vg run the
increase of raindrops below the freezing level is distinct
(Fig. 8a). Faster sedimentation of graupel would reduce
the time available for sublimation, which in turn increases its melting below. Analysis also shows that
given the same amount of mass flux, the faster sedimentation of graupel enhances accretion of other hydrometeors since graupel is carried to lower levels more
rapidly, which results in the reduction of hydrometeors
aloft. The increase of cloud water below the freezing
level may be due to the fact that increased surface cooling and enhanced surface rainfall from more rain increase the gust-front lifting to produce more clouds and
condensation (Fig. 8a; also compare Figs. 5a,c), result-
ing in heating and drying around 2 km above the
ground level (Fig. 8c). The increase of cloud water in
the case of the WSM6_vg over the WSM6 in Fig. 8a was
found to be due to enhanced cloud water formation
after 50 min, as can be seen in the snap shot in Fig. 8.
The cooling and moistening near the surface (Fig. 8c)
can be attributed to the evaporation effects of the
larger amounts of falling raindrops.
Relative to the impact of the sedimentation velocity for graupel, differences in the ice-phase microphysics between the WSM6 and PLIN schemes do not affect
significantly the storm evolution in terms of the hydrometeors (cf. Figs. 8a,b), whereas the changes in the temperature and moisture are comparable (cf. Figs. 8c,d).
Relative to the WSM6 physics, the increase of cloud
ice at colder temperatures (;11 km) and its reduction
at warmer temperatures (;9 km) are prominent in the
PLIN physics run, together with the reduction of snow
amount (Fig. 8b), which reflects the typical characteristics of H04 ice-phase microphysics. The decrease of
graupel can be attributed to the weakening of accretion
due to the reduced amount of ice and snow. The enhanced heating with the maximum at 10 km perhaps
reflects the increase of liquid hydrometeors at that
height in the PLIN physics, whereas the decrease of
them below that level may be related to relative cooling
and moistening due to less cloud water to be frozen and
the saturation profile of PLIN that is weighted by ice
and water content (Fig. 8d). The increase of cloud water in the PLIN scheme in the 8–10-km layer (Fig. 8b;
see also Fig. 5) indicates that the saturation adjustment
JANUARY 2009
HONG ET AL.
69
FIG. 8. Differences in the vertical profiles of (a) hydrometeors (units: g kg21 for rain, snow, and graupel; and
10 3 g kg21 for cloud ice and cloud waters) and (c) temperature and relative humidity (WSM6_vg 2 WSM6). (b),
(d) The differences of PLIN_vg and WSM6 experiments (PLIN_vg 2 WSM6). Units are 0.18C for temperature (T)
and percent for relative humidity (RH). All fields are obtained from domain-averaged values during the 60-min
integration period.
for ice generation in the PLIN scheme is not as efficient
as the generation of ice (Pigen) in the WSM6 scheme.
The relative warming and drying seen near the surface
may reflect a difference in the precipitation evaporation between the two schemes.
From Figs. 7 and 8, it is found that the impact of the
sedimentation velocity for graupel overwhelms the effect of microphysics on the storm evolution in terms of
surface precipitation and hydrometeors aloft. A relatively short integration time in the 2D run could be a
reason why the impact of ice-phase microphysics is relatively insignificant. One may argue that a longer integration could clarify the reason for the differences between the experiments, but integrating the model
longer than 1 h is less meaningful in this idealized test
bed with no other physics and with periodic lateral
70
JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY
VOLUME 48
FIG. 9. The 24-h accumulated rainfall (mm) ending at 0000 UTC 15 Jul 2001, from the 48-km resolution experiments with the (a)
WSM6 and (b) PLIN schemes; (c) their differences (WSM6 2 PLIN); (d)–(f) The corresponding 3-km resolution results.
boundary conditions. Indeed, the extended run longer
than 1 h did not reveal a realistic evolution of the storm
(not shown). The relative importance of the ice microphysics and sedimentation velocity for graupel can be
better seen in the case of the longer 3D simulation with
full physics, as will be shown in the following subsection.
b. Heavy rainfall event
Figure 9 compares the predicted 24-h accumulated
rain valid at 0000 UTC 15 July 2001 obtained from the
experiments with the two schemes at the 48- and 3-km
grid intervals, and the tabulated statistical skill scores
are in Table 3. It is seen that all the experiments capture
the observed heavy rainfall extending from southwest
to northeast across the central part of the Korean Peninsula (cf. Fig. 3), with detailed features in the highresolution grids. An intense precipitation core with the
PLIN scheme results in an increase of the domain total
precipitation (Table 3). More precipitation is simulated
by the PLIN than by the WSM6 scheme and this results
in the deterioration of the bias score for precipitation.
Another distinct impact is that compared to the results
TABLE 3. The pattern correlation coefficients (PCs) and bias
score of 24-h accumulated precipitation over South Korea ending
at 0000 UTC 15 Jul 2001, obtained from the 48-km and 3-km
resolution experiments with the 3-km resolution values in parentheses. The bias score is a ratio of the domain-average precipitation from the model run against the corresponding observation.
Precipitation
Over the Korean Peninsula Over the whole domain
WSM6
PLIN
WSM6_vg
PLIN_vg
WSM6_nr
Correlation
Bias score
0.71 (0.66)
0.70 (0.55)
0.71 (0.64)
0.68 (0.67)
0.66 (0.44)
1.34 (1.42)
1.67 (1.64)
1.37 (1.42)
1.65 (1.66)
1.42 (1.56)
Max (mm) Avg (mm)
229 (257)
259 (290)
202 (238)
259 (286)
216 (203)
14.18 (21.44)
16.46 (23.40)
14.23 (21.79)
16.06 (23.34)
14.63 (21.70)
JANUARY 2009
HONG ET AL.
71
FIG. 10. Vertical distribution of water quantities, obtained from the 3-km experiments with the (a) WSM6 and (b) PLIN schemes,
and (c) their differences, averaged over the heavy rainfall region (33.38–41.08N, 121.58–130.58E) during the 24-h forecast period.
from the PLIN experiment, the WSM6 scheme shifts
the major precipitation band southward toward what
was observed, leading to the improved pattern correlation in the case of the WSM6 run (Figs. 9c,f).
To investigate the fundamental differences in the two
schemes, the vertical profiles of averaged condensates,
obtained from the 3-km experiments, are compared in
Fig. 10, together with differences (PLIN 2 WSM6).
The increase of cloud ice at warmer temperatures is
pronounced when the WSM6 scheme is used. In the
WSM6 experiment, coexisting ice and snow are seen at
warmer temperatures below 400 mb, whereas there is
negligible ice at these levels in the PLIN experiment.
These differences generally reflect the characteristics of
ice-phase microphysics proposed by H04. These char-
acteristics are also seen in the 2D results, but the
distribution of liquid phase hydrometeors differs (cf.
Fig. 6).
The differences in simulated precipitation and hydrometeors between the WSM6 and PLIN experiments
generally follow the scenario revealed in the idealized
experiment (Fig. 11), namely that the precipitation
amount is smaller and the sum of hydrometeors aloft is
larger in the WSM6 run than in the PLIN run. However, the impact of the sedimentation velocity for graupel is quite different from the results that are obtained
in the 2D run (cf. Figs. 7, 11). It is clear that the evolution of domain-averaged precipitation from the
WSM6_vg run (PLIN_vg) is very close to that of the
WSM6 (PLIN) experiment (cf. Figs. 11a, 7a), whereas
FIG. 11. Time series of the (a) precipitation rate and (b) hydrometeor water path, resulting from the WSM6
(solid), PLIN (dotted), WSM6_vg (dashed), and PLIN_vg (dotted–dashed) experiments at 3-km resolution, averaged over the heavy rainfall region (33.38–41.08N, 121.58–130.58E).
72
JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY
VOLUME 48
FIG. 12. As in Fig. 8, but for the heavy rainfall experiments at the 3-km resolution, averaged over the heavy rainfall
region (33.38–41.08N, 121.58–130.58E).
the evolution of volume-averaged water substances
in the WSM6_vg (PLIN_vg) experiment follows the
overall impact as shown in the 2D idealized experiment (cf. Figs. 11b, 7b). Table 3 confirms that the
maximum amount of precipitation and its domainaveraged amount are dominated by the ice-phase
microphysics. The horizontal distribution of precipitation also confirmed that the northward (southward)
shifting of the precipitation band, as seen in Fig. 9,
also appears in the PLIN_vg (WSM6_vg) run (not
shown). A possible reason for the different sensitivity is
given below.
The vertical profiles of the differences in hydrometeors generally follow the characteristics seen in the 2D
run, in which the faster sedimentation of graupel plays
a role in reducing the hydrometeors above the freezing
level (cf. Figs. 8a, 12a), and the WSM6 microphysics
increases the amount of cloud ice and snow (cf. Figs. 8b,
JANUARY 2009
HONG ET AL.
12b). Differences in the vertical distribution of graupel
due to the different microphysics are also similar to
those seen in the 2D run, but there is a relatively large
reduction of graupel in the upper troposphere when the
WSM6 physics is employed. A major difference is
found in the distribution of liquid phase hydrometeors.
The amount of surface rainfall from the WSM6_vg run
is very similar to that from the WSM6 run, which is
different from the results in the 2D case (Figs. 8a, 12a).
Also, the increase of cloud water in the WSM6_vg run
seen in the 2D run is not distinct in the 3D run. The
corresponding differences in temperature and specific humidity are not directly explainable in this 3D
run framework, but it is distinct that the changes due
to the microphysics are larger than those due to the
sedimentation velocity (cf. Figs. 12c,d). The relative
warmness and dryness of the PLIN scheme in the rain
evaporation layer is consistent between the 2D and 3D
run, possibly as a result of different rain evaporation
rates.
A reason for the different effect between the 2D and
3D runs can be deduced from the different thermodynamic environments. We understand that in nature,
evaporation depends on the difference in vapor pressure between the surface of the raindrop and the air,
not on the relative humidity, but the model relative
humidity is a variable for evaporation with the assumption that modeled raindrops are assumed to be at the
same temperature as the air. For example, in the 2D
case, layers around the freezing level are nearly saturated because of strong updrafts, so that the melting
process is more efficient than the evaporation of the
graupel, and vice versa, due to lower relative humidity
in the 3D case.
Another reason can be deduced from the interaction
between the ice clouds and radiation, which is not considered in the 2D case. The reduction of ice particles
through faster sedimentation of graupel in the
WSM6_vg run increases shortwave radiation reaching
the surface, which results in warming the lower troposphere (Fig. 12c). The reduction of cloud ice in the
PLIN physics also brings about the increase of solar
radiation at the surface (Fig. 12d). As a result, the decrease of the stability within the entire troposphere in
the PLIN scheme provides a favorable environment for
convective activity. Both effects enhance the buoyancy
for triggering convection, leading to enhanced rainfall
at the surface, but with a larger impact by the ice-phase
microphysics than by the fall velocity (Table 3). This
may be because an ice cloud has a stronger cloud–
radiation feedback than other ice particles since areal
coverage for ice is relatively large. The importance of
73
FIG. 13. The 24-h accumulated rainfall (mm) ending at 0000
UTC 15 Jul 2001 from the (a) WSM6_nora experiment and (b) the
difference (WSM6 2 WSM6_nora).
cirrus clouds to the radiation feedback and related precipitation was pointed out by H04.
To further confirm the role of the revised ice microphysics in the WSM6 scheme, another sensitivity experiment that excludes the cloud–radiation feedback is
conducted. In Fig. 13, it is seen that the WSM6 scheme
without the cloud–radiation feedback shifts the major
rainband northward, which is the same way as was
simulated by the PLIN scheme. By comparing the three
results from the WSM6, PLIN, and WSM6 without ra-
74
JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY
VOLUME 48
FIG. 14. Pressure–latitude cross sections of the time-averaged temperature (K, shaded), relative humidity (%,
positive with solid and negative with dotted contours), sum of hydrometeors (g kg21, positive with thick blue line
and negative with dotted contours) obtained from (a) WSM6 2 WSM6_nora and (b) WSM6 2 PLIN experiments;
and sum of ice-phase hydrometeors (g kg21, with thick green lines from the WSM6 experiment) averaged over the
longitude region between 125.58 and 126.58E.
diation feedback experiments, it can be deduced that
the southward displacement of the simulated precipitation in the WSM6 scheme, as compared to that from the
PLIN scheme, is due to the enhanced ice cloud amounts
and their radiation feedback. These are further explained below.
It is not straightforward to explain the reason for the
changes in precipitation due to the differences in microphysics in a three-dimensional model framework.
Thus, we try to investigate a plausible mechanism for
the cloud–radiation interaction, as shown by Fig. 14. In
Fig. 14a, it is seen that cloud–radiation interaction
warms the air in the upper troposphere. More cloud ice
exists between 600 and 200 hPa when the WSM6 is used
(Fig. 14b). As shown by H04, increased ice cloud above
leads to the reduced longwave cooling in the upper
troposphere, which shows as a relative warming effect
below 200 hPa. Cooling above 200 hPa is also enhanced
because of the increased longwave cloud-top effect.
Additionally, the decrease of downward solar energy
induces a cooling near the surface. This stabilization
effect appears broadly from south to north across the
precipitation band. Thus, the air to the north has
less chance for forming clouds since temperature is
colder and relative humidity is drier with latitude. The
air to the south is still buoyant, although the surface is
cooler. As a result, the WSM6 scheme tends to stabilize
the atmosphere, as compared with the PLIN scheme,
which enhances (suppresses) vertical motion to the
south (north). Further, it was confirmed that an enhanced vertical motion to the south plays a role in in-
creasing the convergence (divergence) of the meridional wind component in the lower (upper) troposphere, leading to the shift of the rainband southward.
This effect is smaller in the comparison of the WSM6
and PLIN schemes (Fig. 14b), but still visible. Because
of a reduced amount of ice clouds in the PLIN scheme,
the cloud–radiation feedback would be weakened in the
PLIN scheme, and consequently, the WSM6 scheme
displaces the rainband south through an enhanced
feedback between clouds and radiation processes.
5. Concluding remarks
Although this study provides the relative importance
of ice-phase microphysics and fall velocity for ice particles in the bulk-type parameterization approach of
clouds and precipitation, and sheds some light on the
cloud and radiation interaction in forming precipitating
convection, a more robust evaluation of the hydrometeor
profile is needed. Since the case chosen in this study is
associated with heavy rainfall within a stationary monsoon front, the impact of the vertical structure of hydrometeors on the simulated precipitation may not be
as distinct as the large-scale changes due to the cloud–
radiation feedback. Another future case study for locally
driven convection would help us to understand the role
of the vertical distribution of hydrometeors in forming
the precipitation and related mesoscale evolution.
We also recognize that the WSM6 has a deficiency
in reproducing a strong leading edge simulated reflectivity embedded within a squall line, as shown
JANUARY 2009
75
HONG ET AL.
by Thompson et al. (2006). Further revisions to the
WSM6 scheme are undertaken to improve the evolution of the storm by reducing the amount of graupel
(Dudhia et al. 2009).
The evolution of the simulated precipitation with the
inclusion of graupel (WSM6) is similar to that from the
simple (WSM3) and mixed-phase (WSM5) microphysics
in a low-resolution grid; however, in a high-resolution
grid, the amount of rainfall increases and the local maximum becomes stronger as the number of hydrometeors
classes increases (Hong and Lim 2006). This study, comparing WSM6 with PLIN, also implies that the impact of
the complexity in the microphysics due to the number of
prognostic water substance variables on simulated convective activity is smaller than the effects of the manner in
which each microphysical process is formulated in the
same category of prognostic water substance variables.
Finally, it is important to note that the bulk schemes
being compared were the WSM6 and PLIN schemes
within WRF, which are relatively similar bulk schemes,
indicating that the findings of this research can be
specific to these schemes, and that the relationships
observed may differ for other bulk schemes or other
models. Despite such a restriction, our findings of the
relative roles in ice-phase microphysics and its sedimentation velocity are certainly useful as a measure of
differences between typical bulk schemes. This indicates a need for future efforts toward the development
of a more realistic representation of microphysical
processes.
Acknowledgments. This study was supported by the
Korean Foundation for International Cooperation of
Science & Technology (KICOS) through a grant provided by the Korean Ministry of Science & Technology
(MOST) in 2007 and by the Climate Environment System
Research Center sponsored by the SRC program of the
Korea Science and Engineering Foundation (KOSEF).
APPENDIX
List of Symbols in Fig. 1
Symbol
Description
Pcond
Pgaci
Pgacr
Pgacs
Pgacw
Pgaut
Pgdep
Pgeml
Pgevp
Pgfrz
Pgmlt
Piacr
Pidep
Pidw
Pigen
Pihmf
Pihtf
Pimlt
Praci
Pracs
Pracw
Praut
Prevp
Psaci
Psacr
Psacw
Psaut
Psdep
Pseml
Psevp
Psfi
Psfw
Psmlt
Production rate for condensation–evaporation of cloud water
Production rate for accretion of cloud ice by graupel
Production rate for accretion of rain by graupel
Production rate for accretion of snow by graupel
Production rate for accretion of cloud water by graupel
Production rate for autoconversion of snow to form graupel
Production rate for deposition–sublimation rate of graupel
Production rate induced by enhanced melting rate of graupel
Production rate for evaporation of melting graupel
Production rate for freezing of rainwater to graupel
Production rate for melting of cloud ice to form cloud water
Production rate for accretion of rain by cloud ice
Production rate for deposition–sublimation rate of ice
Production of cloud ice by Bergeron process
Production rate for generation (nucleation) of ice from vapor
Production rate for homogeneous freezing of cloud water to form cloud ice
Production rate for heterogeneous freezing of cloud water to form cloud ice
Production rate for instantaneous melting of cloud ice
Production rate for accretion of cloud ice by rain
Production rate for accretion of snow by rain
Production rate for accretion of cloud water by rain
Production rate for autoconversion of cloud water to form rain
Production rate for evaporation–condensation rate of rain
Production rate for accretion of cloud ice by snow
Production rate for accretion of rain by snow
Production rate for accretion of cloud water by snow
Production rate for autoconversion of cloud ice to form snow
Production rate for deposition–sublimation rate of snow
Production rate induced by enhanced melting of snow
Production rate for evaporation of melting snow
Transfer rate of cloud ice to snow through growth of Bergeron process embryos
Bergeron process (deposition and riming) transfer of cloud water to form snow
Production rate for melting of snow to form cloud water
Value SI units
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
kg kg21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
s21
76
JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY
REFERENCES
Bryan, G. H., J. C. Wyngaard, and J. M. Fritsch, 2003: Resolution
requirements for the simulation of deep moist convection.
Mon. Wea. Rev., 131, 2394–2416.
Chen, F., and J. Dudhia, 2001: Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5
modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129, 569–585.
Chen, S.-H., and W.-Y. Sun, 2002: A one-dimensional time dependent cloud model. J. Meteor. Soc. Japan, 80, 99–118.
Dudhia, J., 1989: Numerical study of convection observed during
the Winter Monsoon Experiment using a mesoscale twodimensional model. J. Atmos. Sci., 46, 3077–3107.
——, S.-Y. Hong, and K.-S. Lim, 2009: A new method for representing mixed-phase particle fall speeds in bulk microphysics
parameterizations. J. Meteor. Soc. Japan, in press.
Ferrier, B. S., W.-K. Tao, and J. Simpson, 1995: A double-moment
multiple-phase four-class bulk ice scheme. Part II: Simulations of convective storms in different large-scale environments and comparisons with other bulk parameterizations. J.
Atmos. Sci., 52, 1001–1033.
Fletcher, N. H., 1962: The Physics of Rain Clouds. Cambridge
University Press, 390 pp.
Gilmore, M. S., J. M. Straka, and E. N. Rasmussen, 2004: Precipitation uncertainty due to variations in precipitation particle
parameters within a simple microphysics scheme. Mon. Wea.
Rev., 132, 2610–2627.
Hong, S.-Y., 2004: Comparison of heavy rainfall mechanisms in
Korea and the central United States. J. Meteor. Soc. Japan,
82, 1469–1479.
——, and J.-O. Lim, 2006: The WRF single-moment 6-class microphysics scheme (WSM6). J. Korean Meteor. Soc., 42, 129–151.
——, J. Dudhia, and S.-H. Chen, 2004: A revised approach to
ice-microphysical processes for the bulk parameterization of
cloud and precipitation. Mon. Wea. Rev., 132, 103–120.
——, Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes.
Mon. Wea. Rev., 134, 2318–2341.
Houze, R. A., Jr., P. V. Hobbs, P. H. Herzegh, and D. B. Parsons,
1979: Size distributions of precipitation particles in frontal
clouds. J. Atmos. Sci., 36, 156–162.
Kain, J., and M. Fritsch, 1993: Convective parameterization for
mesoscale models: The Kain–Fritsch scheme. The Representation of Cumulus Convection in Numerical Models. Meteor.
Monogr., No. 24, Amer. Meteor. Soc., 165–170.
Kessler, E., 1969: On the Distribution and Continuity of Water
Substance in Atmospheric Circulations, Meteor. Monogr., No.
32, Amer. Meteor. Soc., 84 pp.
Klemp, J. B., 2006: Convective-resolving forecasting with the
WRF model. Proc. Fourth Joint Korea-U.S. Workshop on
Mesoscale Observation, Data Assimilation and Modeling for
Severe Weather, Seoul, Korea. Korea Science and Engineering Foundation/U.S. National Science Foundation Office of
International Science and Engineering, 58–63.
Kuo, Y.-H., 2006: Assimilation of ground-based GPS data for
short-range precipitation forecast. Proc. Fourth Joint KoreaU.S. Workshop on Mesoscale Observation, Data Assimilation
and Modeling for Severe Weather, Seoul, Korea, Korea Science and Engineering Foundation/U.S. National Science
VOLUME 48
Foundation Office of International Science and Engineering,
38–41.
Lang, S., W.-K. Tao, R. Cifelli, W. Olson, J. Halverson, S. Rutledge, and J. Simpson, 2007: Improving simulations of convective system from TRMM LBA: Easterly and westerly regimes. J. Atmos. Sci., 64, 1141–1164.
Lin, Y.-L., R. D. Farley, and H. D. Orville, 1983: Bulk parameterization of the snow field in a cloud model. J. Climate Appl.
Meteor., 22, 1065–1092.
McCumber, M., W.-K. Tao, J. Simpson, R. Penc, and S.-T. Soong,
1991: Comparison of ice-phase microphysical parameterization schemes using numerical simulations of tropical convection. J. Appl. Meteor., 30, 985–1004.
Meyers, M. P., R. L. Walko, J. Y. Harrington, and W. R. Cotton,
1997: New RAMS cloud microphysics parameterization. Part
II: The two-moment scheme. Atmos. Res., 45, 3–39.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A.
Clough, 1997: Radiative transfer for inhomogeneous atmosphere: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102 (D14), 16 663–16 682.
Noda, A., and H. Niino, 2003: Critical grid size for simulating
convective storms: A case study of the Del City supercell
storm. Geophys. Res. Lett., 30, 1844, doi:10.1029/
2003GL017498.
Reisner, J., R. M. Rasmussen, and R. T. Bruintjes, 1998: Explicit
forecasting of supercooled liquid water in winter storms using
the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124,
1071–1107.
Rutledge, S. A., and P. V. Hobbs, 1984: The mesoscale and microscale structure and organization of clouds and precipitation in midlatitude cyclones. XII: A diagnostic modeling
study of precipitation development in narrow cloud-frontal
rainbands. J. Atmos. Sci., 41, 2949–2972.
Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M.
Barker, W. Wang, and J. G. Powers, 2007: A description of
the Advanced Research WRF version 2. NCAR Tech. Note
NCAR/TN-4681STR, 88 pp.
Tao, W.-K., and Coauthors, 2003: Microphysics, radiation and
surface processes in a non-hydrostatic model. Meteor. Atmos.
Phys., 82, 97–137.
Thompson, G., R. M. Rasmussen, and K. Manning, 2004: Explicit
forecasts of winter precipitation using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis. Mon. Wea. Rev., 132, 519–542.
——, P. R. Field, W. D. Hall, and R. M. Rasmussen, 2006: A new
bulk microphysical parameterization for WRF and MM6.
Proc. Seventh Weather Research and Forecasting Model
Workshop, Boulder, CO, NCAR Mesoscale and Microscale
Meteorology Division, 1–11.
van den Heever, S., and W. R. Cotton, 2004: The impact of hail
size on simulated supercell storms. J. Atmos. Sci., 61, 1596–
1609.
Walko, R. L., W. R. Cotton, M. P. Meyers, and J. Y. Harrington,
1995: New RAMS cloud microphysics parameterization.
Part I: The single-moment scheme. Atmos. Res., 38, 29–62.
Wang, J.-J., X. Li, and L. D. Carey, 2007: Evolution, structure,
cloud microphysical, and surface rainfall processes of monsoon convection during the South China Sea Monsoon Experiment. J. Atmos. Sci., 64, 360–380.