Professional Documents
Culture Documents
A Numerical Study of Clear-Air Turbulence (CAT) Encounters Over South Korea On 2 April 2007
A Numerical Study of Clear-Air Turbulence (CAT) Encounters Over South Korea On 2 April 2007
ABSTRACT
DOI: 10.1175/2010JAMC2449.1
occur above or below the jet stream where the VWS is strong vertical shear in the upper-level outflow of a
sufficiently strong to overcome its environmental stability. mesoscale convective system (MCS). The subsequent
In addition, the entrance region of the jet stream and the higher-resolution simulations of Trier et al. (2010) sug-
confluence zone of two different flows favor CAT gener- gested that this observed turbulence may have been re-
ation through upper-level frontogenesis (Mancuso and lated to shallow banded convection that was organized by
Endlich 1966; Roach 1970). The strong vertical wind shear the vertical shear in the near-neutral to weakly con-
and low static stability near the frontal zone result in KHI. vectively unstable outflow.
Ellrod and Knapp (1992) indicated that a turbulence index Another mechanism of CAT is a breakdown of gravity
1 (TI1 5 DEF 3 VWS, where DEF is total deformation) waves generated by complex terrain and convection. Jiang
that is based on upper-level frontogenesis is a good di- and Doyle (2004), Doyle et al. (2005), and Lane et al. (2009)
agnosis for CAT near the developing upper-level front and claimed that large-amplitude mountain waves and their
the jet core in a highly curved trough. However, it is not subsequent breaking over complex terrain are responsible
sufficient to describe a highly localized individual CAT for small-scale turbulence that directly affects aircraft.
event solely by using the KHI based on frontogenesis (e.g., Mountain waves can be a source of CAT in several ways.
Knox 1997; Kaplan et al. 2005; Ellrod and Knox 2010). First, the amplitude of the vertically propagating mountain
To identify the existence of clouds during this pe- 3. Experimental design of WRF model
riod, composite images of the Multifunctional Transport
Satellite (MTSAT) focused on Korea and East Asia are The numerical model used in this study is the ARW-
used (Fig. 2). In Fig. 2a, there are no clouds over the west WRF model, version 3.0, developed by the National
coast and the eastern mountain areas over the Korean Center for Atmospheric Research (NCAR), which is a
Peninsula, while stratiform clouds that extend from South nonhydrostatic and fully compressible model. Details of
China cover Jeju Island. In Fig. 2b, many clouds are lo- this model can be found in Skamarock et al. (2008). This
cated along the surface frontal system elongated from model has been used for several previous studies on
South China to South Japan. Although some patches of turbulence (e.g., Feltz et al. 2009; Fovell et al. 2007; Trier
cloud extend to Jeju Island, there are no signals of pre- and Sharman 2009) and gravity waves (e.g., Plougonven
cipitation in radar echoes or at the surface observation et al. 2008; S.-Y. Kim et al. 2009). Locations of all model
stations over South Korea, including Jeju Island on 2 April domains used in the present simulation are defined by
2007 (not shown). For these reasons, turbulence that oc- the Lambert conformal map projection shown in Fig. 3.
curred in Korea during this period is regarded as CAT, as Six domains (1, 2, 3, 4, 5, and 6) are considered in the
defined by Ellrod et al. (2003). present simulation with horizontal grid spacings of 30,
Before investigating the possible mechanisms of the 10, 3.3, 1.1, 1.1, and 0.37 km, respectively. Note that the
CAT events over South Korea, it should be noted that horizontal grid spacing of domains 4 and 5 is the same.
there are several uncertainties in the PIREP data. First, The locations of domains 1, 2, and 3 and domains 4, 5,
the locations of turbulence in PIREPs sometimes dif- and 6 are depicted in Figs. 3a and 3b, respectively. Also
fer by more than hundreds of kilometers from the actual shown in Fig. 3b are the locations of the operational
locations of the turbulence observed by in situ measure- radiosonde stations over South Korea and the nine
ments (Cornman et al. 2004). Second, the turbulence MOG-level and three light (LGT)-level CAT events
intensity reported in PIREPs tends to be different de- that occurred on 2 April 2007. Two-way nesting in-
pending on the particular pilot and aircraft type or size. teractions are conducted for domains 1 and 2, 2 and 3, 3
The aircraft types of the PIREP data in the present study and 4, and 3 and 5, whereas one-way interaction is con-
are the Boeing 747, Airbus 332, and Airbus 333. Al- ducted for domains 5 and 6. Horizontal grid spacing and
though PIREPs inherently have some uncertainties in the size (5730 3 5130 km) of domain 1 are set exactly the
location and intensity of the turbulence (Schwartz 1996), same as those of the operational RDAPS analysis to al-
they are invaluable resources for validating and evaluat- low direct comparison of synoptic features. Domain 3
ing various CAT diagnostics derived using several oper- includes all CAT events considered in this study, while
ational forecasting models (e.g., McCann 1999; Lee et al. domains 4 and 5 encompass those in Jeju Island and the
2003; Kaplan et al. 2006; Sharman et al. 2006; Knox et al. eastern mountain areas, respectively. Since detailed
2008; Jang et al. 2009; J.-H. Kim et al. 2009). examination of the mountain waves and their breaking is
DECEMBER 2010 KIM AND CHUN 2385
shown). This is a typical springtime flow pattern in this and Chung 2006; Jang and Chun 2008). The simulated
area, except for the relatively stronger westerly winds results are fairly good compared with the RDAPS analysis
(about 20 m s21 faster) and highly curved trough at in both wind and sea level pressure fields. One noticeable
300 hPa, compared with corresponding fields for the discrepancy, nevertheless, between the analysis and the
6-yr (2003–08) springtime mean calculated by the RDAPS simulated is that the simulated center of the Siberian high
analysis data (not shown). Also, the pressure gradient shifts southward with somewhat different isobaric lines to
force across the Korean Peninsula is exceptionally strong the west of the Siberian high.
on 2 April 2007 because of an intensified Siberian high For further investigation of the synoptic structure and
at the surface. This synoptic feature is similar to that for comparison between the model results and obser-
of CAT events near the upper-level fronts and fronto- vations, vertical profiles of wind and temperature de-
genesis environments shown by previous studies (e.g., rived from the simulations in domains 1–3 and observed
Roach 1970; Keller 1990; Kaplan et al. 2006). The pre- from two radiosonde stations are shown in Fig. 5. Be-
vailing wind is almost perpendicular to the meridionally cause the simulated soundings in the three domains are
oriented mountains located in the eastern part of the similar, only the result in domain 1 will be discussed and
Korean Peninsula, and this provides an optimal envi- compared with the observations. Figure 5 shows the
ronmental condition to generate large-amplitude moun- skew T–logp diagrams of temperature, dewpoint tem-
tain waves (Jiang and Doyle 2004; Doyle et al. 2005; Kim perature, and wind derived using the simulated (gray
DECEMBER 2010 KIM AND CHUN 2387
5. Model results
a. West coast area
The potential temperatures in domain 2 at z 5 6.1 km
(average level of number 3, 4, and 6 CAT events) at
1800 UTC 1 (left) and 0700 UTC 2 April 2007 (right) are
shown in Fig. 6a, and horizontal wind speed at z 5 9 km
(the level of the maximum jet stream) with horizontal
wind vectors and pressure at z 5 6.1 km are shown in
Fig. 6b. An intensifying upper-level front and associated
jet stream are identified during this 13-h period. At
1800 UTC, a relatively strong meridional temperature
gradient in a frontal zone is located from Balhae Bay
to the northern part of the Korean Peninsula. As this
frontal zone moves southward, the meridional temper-
ature gradient increases at 0700 UTC (Fig. 6b), implying
an intensified upper-level front during this period. As
FIG. 5. Skew T–logp diagrams (solid, temperature; dashed, the strengthening frontal zone induces strong VWS
dewpoint) obtained from the observed (black) and the simulated
(gray) soundings at (a) Baekryungdo and (b) Gwangju observation
through the thermal–wind relationship, the maximum
sites in South Korea at 0000 UTC 2 Apr 2007. (right) Simulated and wind speed of the jet stream above the frontal zone also
(far right) observed wind speeds at each level with full (10 m s21) increases from 70.2 to 82.9 m s21 over the 13 h. In ad-
and half (5 m s21) barbs and a pennant (50 m s21). Isotherms in dition, the convergence zone of north and south west-
108C intervals are denoted by slantwise thin lines in both panels. erlies in the entrance region of the jet stream ahead of
the highly curved trough approaches the west coast of
Korea, where the number 3, 4, and 6 CAT events occurred
lines) and the observed (black lines) soundings at (a) (Fig. 6b). This flow structure is consistent with several
Baekryungdo and (b) Gwangju at 0000 UTC 2 April other CAT events near the upper-level front and the jet
2007. Relatively strong stabilities in the layer between stream (e.g., Keller 1990; Ellrod and Knapp 1992; Kaplan
z 5 4 and 8 km at Baekryungdo and between z 5 3 and et al. 2005).
5 km at Gwangju stations are observed. This intensified To understand how localized turbulence relates to
stability in the midtroposphere is related to the sub- the evolution of the intensifying upper-level front/jet
sidence of stratospheric air because of deep tropopause system, we show in Fig. 7 the horizontal wind speed
2388 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 49
(Fig. 7a) and Richardson number (Fig. 7b) with poten- about z 5 10–12 km on the anticyclonic side of the jet
tial temperature and 1.5 potential vorticity unit (PVU) core to about z 5 5 km on the cyclonic side following the
in y–z cross sections along the line shown in Fig. 6b (right), upper-level front. At 0700 UTC (Fig. 7b; right), the
using the same data shown in Fig. 6. At 1800 UTC (Fig. 7a, overall structure of the upper-level front–jet system is
left), the jet core is located at z 5 9 km and the maxi- similar to that at 1800 UTC. However, the slope of the
mum wind speed is 74 m s21. The upper-level front that isentropes near the upper-level front is steeper, so the jet
is represented by slantwise and dense contours of the maximum at z 5 9 km becomes larger (89 m s21). In
potential temperature in the midtroposphere near z 5 addition, the tropopause is deeply folded down to about
3–6 km and y 5 600–1200 km is located below the jet z 5 4 km with a steep 1.5 PVU line following the upper-
core and has relatively strong vertical and horizontal level front. Because of this deep tropopause folding,
gradients of wind speed and temperature, respectively. relatively strong stability layers are located in the mid-
In addition, the height of the dynamic tropopause troposphere because of the intrusion of relatively stable
(1.5 PVU) decreases from its background height of stratospheric air. This is consistent with the observation
DECEMBER 2010 KIM AND CHUN 2389
profiles of the temperature shown in Fig. 5. This struc- sections of the Richardson number with the potential
ture is also consistent with tropopause folding that oc- temperature and 1.5 PVU in the different domains are
curred along the upper-level frontal zones in previous presented in Fig. 8. Figures 8a–8c are the simulated results
studies (e.g., Keyser and Shapiro 1986; Lane et al. 2004; from domain 1 (Dx 5 30 km), domain 2 (Dx 5 10 km),
Koch et al. 2005) and is important not only for aviation and domain 3 (Dx 5 3.3 km) in the region occupied by
safety but also for the mixing and stratosphere–troposphere two vertically dashed lines shown in Fig. 7b (right), re-
exchange (STE) of chemical constituents (e.g., Shapiro spectively. Note that the x axis in each plot denotes the
1980; Pavelin et al. 2001). As the development of the relative distance from the southernmost point of each
upper-level front–jet system causes a strong horizon- domain. Overall features of the tropopause folding and
tal temperature gradient in the upper-level frontal zone, upper-level frontal system are similar; except that the
VWS is maximized through the thermal wind rela- lines of 1.5 PVU become steeper as the horizontal grid
tionship. Consequently, the local Richardson number resolution becomes higher. In addition, there is a bump
becomes smaller along this frontal zone, mainly due to at the dynamic tropopause near x 5 700 km and z 5 5–
strong VWS regardless of strong stability (see Fig. 7b). 6 km of Fig. 8c, caused by relatively large-amplitude
The locations and timing of the number 3, 4, and 6 CAT waves generated by steep mountain terrain. The relative
events observed over the west coast of Korea are re- vorticity above the steep mountain near z 5 5–6 km (not
produced in domain 2 of the simulation. shown) is less than 23 3 1024 s21, and consequently,
To investigate the resolution dependency of the sim- potential vorticity has a local minimum there. This is less
ulation results illustrated in Fig. 7b (right), y–z cross evident in low-resolution domains that have smooth and
2390 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 49
small mountains. The regions of the smallest Richardson core, the local Richardson number is less than 0, im-
number along the upper-level front are located below plying that there exists local vertical mixing of potential
the jet core and are mainly due to enhanced VWS along temperature in the left side of Fig. 10c (described below).
the intensifying upper-level front/jet system in all three Figure 9a shows the accumulated convective precipita-
domains. These regions of small Richardson number tion with zero absolute vorticity, horizontal wind vector,
are sustained for about 5 h from 0400 to 0900 UTC, and pressure at the 10.7-km level (averaged level of
during which the intensifying upper-level front/jet system the number 2, 8, and 9 CAT events over the Jeju Island
is dominant over the Korean Peninsula. The smallest area) derived using the simulated results of domain 1 at
Richardson number is 0.224, near the location of the 1900 UTC 1 (left) and 0800 UTC 2 (right). Note that Fig. 9a
number 4 CAT in domain 3. In all three domains, the is not the whole area of domain 1 but focuses on the
regions of the small Richardson number are not obvi- regions of anticyclonic flow and the Jeju Island area.
ously different, implying no significant resolution de- Note also that the absolute vorticity enclosed by the zero
pendency on the simulation results in this area. The lines is negative. During this period, the Jeju Island area
simulation with 30-km horizontal grid spacing can iden- is located on the anticyclonic shear side of the jet streak
tify regions of the CAT events sufficiently in this area at that is passing through the west coast area of South
this time. Korea as described in the previous section. Because of
this strong anticyclonic shear flow, areas of negative
absolute vorticity increase during this period (Fig. 9a).
b. Jeju Island area
Considering that the inertial instability associated with
In Fig. 7b (right), a region of relatively low static sta- the negative absolute vorticity has been regarded as
bility (relatively sparse contours of potential tempera- a possible source of local turbulence that directly affects
ture) and small Richardson number is located on the aircraft in the anticyclonic flow (Knox 1997; McCann
anticyclonic shear side of the jet core near z 5 9–12 km 2001; Trier and Sharman 2009), CAT events observed in
and y 5 300–600 km. The number 2, 8, and 9 CAT events, the Jeju Island area are likely related to the inertial in-
observed over the Jeju Island area, are located in this stability. One can apply Stone (1966)’s criterion for
region at this time in domain 2 (shown in the left of inertial instability—f [ f(1 2 Ri21) 1 z] , 0, where f is
Fig. 10c, described below). This region has relatively the Coriolis parameter, Ri is the Richardson number,
strong horizontal wind shear but weak vertical wind shear and z is the vertical component of relative vorticity.
[vertically oriented contours of isotach; see Fig. 7b (right)]. When we apply Stone’s criterion, the areas of inertial
In some local region on the anticyclonic side of the jet instability are generally similar to those shown in Fig. 9,
DECEMBER 2010 KIM AND CHUN 2391
although they occupy somewhat broader regions. This is the accumulated convective precipitation in Fig. 9a (right),
likely because Stone’s criterion includes both symmetric a well-developed convective system appears along the
instability and pure inertial instability. surface frontal zone extending from the mid and southern
As the upper-level front–jet system is intensified dur- parts of China to southern Japan. The overall features of
ing this period, the maximized wind speed of the jet stream the simulated precipitation are in good agreement with
causes the anticyclonic shear region (negative absolute the observed convective bands shown in satellite imagery
vorticity) to be larger near the Jeju Island area (Fig. 9a, (Fig. 2b). As dominant westerly flow passes through the
right). On top of that, ageostrophic flow is also domi- developed convective system, wind vectors are signifi-
nant, which can be seen from the largely different angle cantly perturbed and distorted on the lee side of the
between the wind vectors and isobaric lines. This strong convective system (Fig. 9a, right). As locally enhanced
ageostrophic flow may be caused by well-developed con- cross-streamwise confluent flow (ageostrophic flow)
vective systems in the southern part of China. As seen by forms the local maximum of the vertical component of
2392 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 49
relative vorticity (vertical vorticity hereafter) on this 2, 8, and 9 CAT events. To understand the linkage be-
anticyclonic shear side of the jet streak region, local tween the developed convective system and inertial in-
concentration of vertical vorticity causes perturbations stability, an additional dry simulation is conducted. The
of absolute vorticity. As a result, the zero contours of experimental setting of the dry simulation is the same as
absolute vorticity are also perturbed (Fig. 9a, right). As- for the control simulation, with two-way nesting inte-
suming that strong horizontal gradients of vertical vor- gration through domains 1–5, except that convective
ticity can be translated to a microscale horizontal vortex parameterization is not used in domains 1 and 2 and
tube (Clark et al. 2000; Kaplan et al. 2005), this could microphysical processes are turned off in all domains.
generate small-scale turbulence that directly affects Figure 10a shows the subgrid-scale TKE with horizontal
aircraft. As can be seen in Fig. 9a (right), the region of wind vectors and boundaries of zero absolute vorticity in
perturbations of zero absolute vorticity along the maxi- the control simulation with moisture process (left) and
mized streamwise-oriented gradient of vertical vorticity dry simulation (right) at 10.7 km at 0800 UTC 2 April
is consistent with the location of the number 2, 8, and 9 2007. Note that the light contours in the left figure are
CAT events over the Jeju Island area. the pressure difference between the control and dry
Figure 9b shows the NCSU2 index with zero absolute simulations. Figures 10b and 10c are the same as Figs. 7a
FIG. 10. Comparison between the (left) control and (right) dry simulations at 0800 UTC 2 Apr, 2007: (a) horizontal wind barbs, pressure
(contours on right), pressure difference between the control and dry (contours on left), and zero absolute vorticity (s21; thick contour)
superimposed on the subgrid-scale TKE (m2 s22; shading) in domain 1, (b) potential temperature (contours) and 1.5 PVU (potential
vorticity unit; 1026 K kg21 m2 s21; thick line) superimposed on horizontal wind speed (m s21; shading), and (c) the Richardson number
(shading) along the dashed line in Fig. 6b (right). Contour intervals in (a) are 1 and 4 hPa, respectively. Contour intervals for (b) and (c)
are 4 K. Locations of CAT events over the Jeju Island area are depicted in (a) and (c) as asterisks with numbers corresponding to those in
Table 1.
2394 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 49
other, and small-scale features can be seen as horizontal dependent on the structure of topography as well as up-
grid resolution increases. The maximum subgrid-scale stream wind and stability. For this reason, an additional
TKE (1.212 m2 s2) occurs at the location of the number domain, domain 6 (Dx 5 0.37 km), is included in domain
2 CAT in domain 3, whereas it is less close to the number 5 (Dx 5 1.1 km) of the eastern mountain area, as shown in
2 CAT in domain 1. In all domains, the regions of local Fig. 3b, with one-way nesting interaction. Initial condi-
vertical mixing induced by ageostrophically disturbed tions and physical parameterizations in domains 5 and 6
flow on the anticyclonic shear side of the jet core are not are the same as those in the coarser domains, but the
obviously different. This suggests that numerical simu- boundary conditions in domain 6 are from the results of
lation even with 30-km horizontal grid spacing can domain 5 updated every 30 min. Large-scale flow during
identify the regions of CAT events in this area at this this period is in good agreement with observations, al-
time. In addition, the modeled turbulence in this area is ready shown in Figs. 4 and 5.
maintained for about 5 h from 0400 to 0900 UTC in do- Figure 12 shows vertical velocity (shading) with hor-
mains 1, 2, 3, and 4 (not shown). This is because the en- izontal wind vectors in domain 3 at four different levels
vironmental conditions conducive to CAT continue in (z 5 15, 11, 7, and 3 km) at 0300 UTC 2. In this figure,
this area as long as the synoptic or mesoscale forcings the localized wave structure is captured over the eastern
support their generation. These results are somewhat side of South Korea. These waves are generated by the
consistent with those of Trier and Sharman (2009), where strong zonal flow across the meridionally oriented
the simulated lower Richardson number and the subgrid- mountains over the eastern part of the Korean Penin-
scale TKE with 3-km horizontal and 400-m vertical grid sula. In particular, their amplitudes near the locations of
spacings in the anticyclonic outflow of the MCS are sus- the number 1, 5, and 7 CAT events over the eastern
tained for several hours. mountain area are larger than other regions at z 5
11 km, with the maximum amplitude of vertical velocity
of 13.0 and 24.4 m s21 near the location of the number
c. Eastern mountain area
7 CAT. This clearly shows that smaller-scale waves ap-
This section investigates the generation mechanism pear more in the lower altitudes, while only components
of the CAT events that occurred in the eastern moun- with relatively long wavelength reach up to the upper
tain area of South Korea on 2 April 2007. To precisely troposphere.
simulate mountain waves generated by complex terrain, To understand the filtering process of the short-
topography and large-scale flow must be represented re- wavelength waves at low altitude, the vertical profile of
alistically. This is because characteristics of the mountain m2 (where m is the vertical wavenumber) for a station-
waves such as amplitude and wavelength are highly ary mountain wave is calculated using the following
DECEMBER 2010 KIM AND CHUN 2395
FIG. 12. Horizontal wind vectors superimposed on vertical velocity (m s21; shading) obtained from the simulated results of domain 3 at
selected heights z 5 (top left) 15, (top right) 11, (bottom left) 7, and (bottom right) 3 km at 0300 UTC 2 Apr 2007. Locations of CAT
events over the eastern mountain area are depicted as asterisks with numbers corresponding to those in Table 1. Location of domain 6 is
indicated as an inner box in the upper-left plot.
equation: m2 5 (N2/U2) 2 (1/U2)(d2U/dz2) 2 k2, where k 0300 UTC 2 April 2007. The vertical wavenumber of the
is the horizontal wavenumber, and U and N are the basic- stationary gravity waves for a given horizontal wave-
state wind and Brunt–Väisälä frequency, respectively. length changes with altitude as the background stability
Figure 13 shows m2 profiles of waves with horizontal and wind change. Once a vertically propagating gravity
wavelengths of 12.5 km (Fig. 13a) and 4.2 km (Fig. 13b). wave meets a level of m2 5 0, it can be reflected backward
The U and N profiles used to calculate m2 are the domain- from the reflection level (level of m2 5 0) (Gossard and
averaged (domain 6) zonal wind and temperature at Hooke 1975). In Fig. 13, a significant reduction in m2
2396 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 49
appears between z 5 4 and 8 km because of the exis- The 0 wind is derived from the breaking of the gravity
tence of strong variance in background stability in the waves, and it becomes a wave-induced critical level
troposphere, which is mainly due to the deep tropo- (WICL) of the stationary mountain wave (Clark and
pause folding as mentioned in section 5a. There is no Peltier 1984). Once the WICL is formed, consecutive
level of m2 5 0 for the gravity wave with a relatively stationary mountain waves can be either reflected from
longer horizontal wavelength (12.5 km; Fig. 13a), while or absorbed at the critical level, depending on the local
a deep evanescent layer exists (z 5 5.2–10.6 km) for the Richardson number. Since the local Richardson number
wave with short horizontal wavelength (4.2 km; Fig. in the present case is lower than 0.25 near the WICL
13b). For the given U and N used in Fig. 13, we found (Fig. 14a), mountain waves are likely to be reflected.
that waves with horizontal wavelengths shorter than There are two interesting features in Fig. 14d. First,
10.5 km can be reflected in the 4–8-km-height range. almost erect or weakly tilted vertical velocities are shown
Thus, smaller-scale waves appear more in the lower alti- below about z 5 12 km. Spectral analysis of the vertical
tudes, while only relatively long wavelength components velocity with respect to horizontal wavelength reveals
reach up to the upper troposphere, as shown in Fig. 12. that there exists a single spectral peak at 22.1 km for all
Figure 14 shows x–z cross sections of the Richardson heights below z 5 12 km (right side of Fig. 16b, described
number (Fig. 14a) and subgrid-scale TKE with potential below). To examine the possibility of ducting between the
temperature (Fig. 14b), zonal wind (Fig. 14c), and ver- surface and z 5 12 km for a wave with a horizontal wave-
tical velocity (Fig. 14d) in domain 6 at 0300 UTC 2 along length of 22.1 km, the m2 profile of the wave (Fig. 15a) is
the line of y 5 75 km in domain 6 (36.868N in Fig. 12) at calculated using the zonal-mean zonal wind and temper-
which the number 7 CAT is located. Zero and negative ature in the x–z cross section, as shown in Fig. 14d.
values of the wind component are depicted as bold and Figure 15a shows that although m2 decreases with height
dashed lines in Figs. 14c and 14d, respectively. It shows up to about z 5 7 km, it is larger than 0 and increases
that mountain waves simultaneously break down in z 5 above z 5 7 km, with relatively large magnitude at z 5
13.5–17 km. This breaking activates subgrid-scale TKE 12 km. We found that a 0-m2 level exists below z 5 12 km
in that region (Fig. 14b). However, the simulated tur- only for waves with horizontal wavelengths less than
bulence is located above the observed number 7 CAT. 7.7 km. This implies that the wave below z 5 12 km with
This discrepancy will be examined in Fig. 17, which is a 22.1-km horizontal wavelength is not a ducted wave.
described below. Between z 5 3 and 5 (5 and 9) km, Rather, it is a weakly tilted upward-propagating wave.
a relatively strong (weak) stability layer appears, and Figure 15b is the angle of the phase line to the local
this is likely due to the deep tropopause folding, as vertical for the nonhydrostatic
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi internal gravity wave
shown in Fig. 7. In Fig. 14c, the maximum zonal wind, given by cosa 5 k/ k 2 1 m2 , which is calculated using
which exceeds 70 m s21, locates near z 5 8 km. How- the zonal wavenumber corresponding to the 22.1-km
ever, the zonal wind decreases with height above about z horizontal wavelength and the vertical wavenumber ob-
5 10 km, and 0 wind regions appear near z 5 15 km. tained in Fig. 15a. Figure 15b demonstrates clearly that
DECEMBER 2010 KIM AND CHUN 2397
the phase angle is almost 0 below z 5 12 km, while it two regions are the wave-breaking regions shown in
increases above z 5 12 km as m2 increases. The increase Figs. 14a–14c. The waves with horizontal wavelengths
in m2 above z 5 12 km is mainly due to increases in the much smaller than those below z 5 12 km are the sec-
Brunt–Väisälä frequency in the stratosphere. The profiles ondary waves, and there have been several numerical
of m2 and a calculated using the observed sounding at modeling studies on this topic associated with mountain
Osan (not shown) are similar to those shown in Fig. 15. waves (e.g., Bacmeister and Schoeberl 1989; Satomura
This confirms that the waves with horizontal wavelength of and Sato 1999) and convective gravity waves (e.g., Holton
about 22 km below z 5 12 km are vertically propagating and Alexander 1999; Zhou et al. 2002; Snively and Pasko
waves rather than ducted waves. 2003; Lane and Sharman 2006; Chun and Kim 2008).
Second, two regions exist of small horizontal wave- Because of the horizontal resolution, the secondary
lengths in the layer between z 5 12 and 17 km, one near waves appear only in the two high-resolution sub-
x 5 70–100 km and the other near x 5 0–20 km. These domains (domains 5 and 6) with 1.1- and 0.37-km grid
2398 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 49
FIG. 15. Vertical profiles of (a) m2 (where m is the vertical wavenumber of the mountain
spacings, and waves with horizontal wavelengths less performed an additional simulation with a different
than about 5 km appear only in domain 6. Shown in subgrid-scale turbulence scheme, the Yonsei Univer-
Fig. 16 are the x–z cross sections of vertical velocities sity (YSU) PBL scheme (Noh et al. 2003; Hong et al.
in domains 5 and 6 (left) and their power spectral 2006), in domain 6 through one-way nesting to domain
densities with respect to horizontal wavelength (right). 5 that used the MYJ PBL scheme. The simulated tur-
The vertical velocity in domain 6 is the one shown in bulence near the number 7 CAT using the YSU PBL in
Fig. 14d. In domain 5 (Fig. 16a), the magnitude of the domain 6, however, is not evidently different from that
mountain waves is very small compared with that in using the MYJ PBL scheme (not shown). This implies
domain 6, not only for small-scale secondary waves that that simulated turbulence in domain 6 is not very sensi-
are not fully resolvable but also for the dominant wave tive to the choice of the subgrid-scale turbulence scheme.
with horizontal wavelength of about 22 km below z 5 Rather, simulated variables that are important to de-
12 km. The spectral analysis of the vertical velocity in termine the wave-breaking conditions such as wind and
domain 5 below z 5 12 km reveals that the dominant temperature fields can differ from the observations,
horizontal wavelength is 19.4 km (Fig. 16a right), which and consequently the simulated wave-breaking height
is similar to the dominant horizontal wavelength of is higher than the observed height. To examine this
22.1 km in domain 6 (Fig. 16b right). Although relatively possibility, we estimated the wave-breaking height
small-scale disturbances related to the wave breaking from the mountain-drag parameterization by Chun
appear near x 5 140–160 km and z 5 15–17 km in do- et al. (1996) based on Palmer et al. (1986), using the
main 5, they are much less evident than in domain 6. As basic-state wind and stability profiles obtained from the
the horizontal grid spacing decreases, more complex simulation and observations. The following is a brief
terrain features becomes resolvable and mountain height description from Chun et al. (1996) of how to estimate
generally increases. Consequently, simulated waves have the wave-breaking height: (i) Calculate mountain-wave
a wider spectrum including short wavelength components stress at the surface (denoted by subscript s) t s 5
and larger amplitudes. Since wave breaking is directly 2K0rsNsUsjh9j2, where K0 is the horizontal wavenumber
related to wave amplitude, less or no wave breaking oc- (2.85 3 1024 is used, corresponding to a 22-km horizontal
curs in the relatively low-resolution simulations. wavelength), rs is air density, Ns is the Brunt–Väisälä
Figure 14 clearly demonstrated that regions of wave frequency, Us is the basic-state wind, and jh9j2 is the
breaking exist in the 13.5–17-km-height range, and this variance of subgrid-scale mountain height, which is
can generate local turbulence. However, the simulated calculated in this study using the 30-s digital elevation
wave-breaking regions are located somewhat higher model (DEM) dataset at the nearest grid point of the
than the location of the observed number 7 CAT. To RDAPS analysis data from the number 7 CAT (h9 5
understand the factors that create the discrepancy, we 300 m is used). (ii) Calculate the height variation due
DECEMBER 2010 KIM AND CHUN 2399
to the gravity wave at any level using (dh)k 5 [jtk21j/ zonal-mean zonal wind and temperature in the cross
(K0rk21Nk21Uk21)]1/2. (iii) Find the wave-breaking section shown in Fig. 14. The observed basic-state wind
level based on the minimum Richardson number in- and Brunt–Väisälä frequency are derived by applying
cluding wave effects [Rim 5 Ri(1 2 «)/(1 1 «Ri1/2)2], a nine-point vertical average to the rawinsonde data.
where Ri is the local Richardson number, and « is The simulated basic-state wind and stability resembles
the inverse Froude number [« 5 (dh)kNk21/Uk21]. If the observations in a reasonable sense, except with
Rim , 1/ 4, wave breaking occurs and the saturation relatively weak vertical variation and different mag-
wave stress is calculated (Lindzen 1967). If Rim $ 1/ 4, there nitude. Because of this discrepancy, vertical profiles of
is no wave breaking and wave stress is the same as that the minimum Richardson number differ significantly
at the level below (Eliassen and Palm 1960). (iv) Repeat between the simulated and the observed. Based on the
the previous step at higher levels until the wave stress is simulated basic-state flow, the first wave breaking oc-
zero or the model or observation top is reached. curs at z 5 14 km, which is consistent with the altitude
Figure 17 shows vertical profiles of the basic-state wind of the simulated turbulence shown in Fig. 14. In con-
(Fig. 17a), Brunt–Väisälä frequency (Fig. 17b), minimum trast, the wave-breaking height estimated using the
Richardson number (Fig. 17c), and mountain-wave stress observed basic-state flow in the upper troposphere is
(Fig. 17d) derived using the simulated result in domain 12.8 km, which is close to the altitude of the observed
6 (solid) and an observed rawinsonde sounding at number 7 CAT (12.6 km). Detailed inspection of the
Osan (dashed) at 0000 UTC 2 April 2007. Osan is lo- simulated and observed wind and stability near 12–
cated upstream of the eastern mountain area and is 14 km reveals that a slight difference in the vertical
depicted in Fig. 3b. The model-produced basic-state variation and magnitude of the basic-state flow causes
wind and Brunt–Väisälä frequency used in the calcu- a difference in the wave-breaking height of more than
lation of m2 shown in Fig. 15a are obtained from the 1 km.
2400 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 49
6. Summary and conclusions the WRF model, including six nested domains with the
finest horizontal grid spacing of 0.37 km. Large-scale
On 2 April 2007, nine MOG-level CAT events were features simulated in the lowest-resolution domain (Dx 5
reported in PIREPs during 6.5 h over the Korean Pen- 30 km) resemble those in the RDAPS analysis data with
insula. This is the highest number of CAT occurrences the same resolution. In addition, vertical profiles for wind
for this region in one day for the last 10 yr. These CAT direction and speed, temperature, and dewpoint tem-
events were distributed in three different regions: the perature derived from the simulations closely match ra-
west coast, Jeju Island, and the eastern mountain areas of diosonde observations.
South Korea. Composite images from the MTSAT satel- Based on the simulation results, we found that gen-
lite reveal no clouds over the west coast and eastern eration mechanisms of the CAT events in the three
mountain areas, whereas the Jeju Island area is covered regions are somewhat different. In the west coast area,
with stratiform clouds extending from the southern part of an intensifying upper-level frontal zone is formed by
China. Based on the available observations such as radar a strong meridional temperature gradient through the
reflectivity and record of precipitation, all CAT events thermal–wind relationship, and upper-level fronto-
including those in Jeju Island occurred in clear or non- genesis causes strengthening of the jet stream over the
convective weather conditions. west coast area, which finally introduces maximized
The characteristics and possible generation mechanisms vertical wind shear below the jet core. Deep tropopause
of the CAT events in the three areas are investigated folding on the cyclonic shear side of the jet core leads to the
using available observations and simulation results of local Richardson number being smaller than 0.25, which
DECEMBER 2010 KIM AND CHUN 2401
can trigger the Kelvin–Helmholtz instability. The lo- can be sufficiently identified by current NWP models
cations of the KHI are consistent with those of the with a horizontal grid spacing of ;10–30 km, while those
observed CAT events over the west coast area. Reso- generated by mountain waves need much higher reso-
lution dependency of the simulated turbulence in this lution. This has an important implication for current and
area is not significant, and a model with 30-km hori- future CAT forecasting. In current CAT forecasting,
zontal grid spacing and 300-m vertical grid spacing can based on the combination of various CAT indices cal-
identify the regions of the CAT events during this period. culated using NWP model output, CAT events associ-
CAT events observed in the Jeju Island area are likely ated with mountain waves cannot be diagnosed well, and
associated with inertial instability. As the intensifying thus alternative approaches may be required. For future
upper-level front/jet system approaches the Korean CAT forecasting, NWP model–forecasted turbulence
Peninsula, the region of inertial instability becomes larger can be used directly instead of calculating diagnostic
and moves to the Jeju Island area in the anticyclonic shear CAT indices, given that turbulence generated by large-
side of the maximized jet core. As strong westerly flows scale forcing is explicitly represented in the numeri-
passing through a well-developed convective system be- cal simulation with a current NWP model resolution.
come disturbed, an ageostrophically induced mesoscale As the resolution of NWP models increases, direct
——, W. D. Hall, R. M. Kerr, D. Middleton, L. Radke, F. M. Ralph, Turbulence Forecasting Algorithm (KITFA). Atmosphere,
P. J. Nieman, and D. Levinson, 2000: Origins of aircraft 19, 255–268.
damaging clear-air turbulence during the 9 December 1992 Janjic, Z. I., 2002: Nonsingular implementation of the Mellor-Yamada
Colorado downslope windstorm. J. Atmos. Sci., 57, 1105–1131. level 2.5 scheme in the NCEP Meso Model. NCEP Office Note
Cornman, L. B., G. Meymaris, and M. Limber, 2004: An update 437, 61 pp.
on the FAA Aviation Weather Research Program’s in situ Jiang, Q., and J. D. Doyle, 2004: Gravity wave breaking over the
turbulence measurement and reporting system. Preprints, central Alps: Role of complex terrain. J. Atmos. Sci., 61, 2249–
11th Conf. on Aviation, Range, and Aerospace Meteorology, 2266.
Hyannis, MA, Amer. Meteor. Soc., P4.3. [Available online at Joseph, B., A. Mahalov, B. Nicolaenko, and K. L. Tse, 2004: Var-
http://ams.confex.com/ams/11aram22sls/techprogram/paper_ iability of turbulence and its outer scales in a model tropo-
81622.htm.] pause jet. J. Atmos. Sci., 61, 621–643.
Derber, J. C., D. F. Parrish, and S. J. Lord, 1991: The global Kain, J. S., 2004: The Kain–Fritsch convective parameterization:
operational analysis system at the National Meteorological An update. J. Appl. Meteor., 43, 170–181.
Center. Wea. Forecasting, 6, 538–547. Kaplan, M. L., A. W. Huffman, K. M. Lux, J. D. Cetola, J. J.
Doyle, J. D., M. A. Shapiro, Q. Jiang, and D. L. Bartels, 2005: Charney, A. J. Riordan, Y.-L. Lin, and K. T. Waight III,
Large-amplitude mountain wave breaking over Greenland. 2005: Characterizing the severe turbulence environments as-
J. Atmos. Sci., 62, 3106–3126. sociated with commercial aviation accidents. Part II: Hydro-
Dudhia, J., 1989: Numerical study of convection observed during static mesoscale numerical simulations of supergradient wind
Lane, T. P., and R. D. Sharman, 2006: Gravity wave breaking, Prusa, J. M., P. K. Smolarkiewicz, and R. R. Garcia, 1996: Propa-
secondary wave generation, and mixing above deep convec- gation and breaking at high altitudes of gravity waves excited
tion in a three-dimensional cloud model. Geophys. Res. Lett., by tropospheric forcing. J. Atmos. Sci., 53, 2186–2216.
33, L23813, doi:10.1029/2006GL027988. Roach, W. T., 1970: On the influence of synoptic development on
——, and ——, 2008: Some influences of background flow condi- the production of high level turbulence. Quart. J. Roy. Meteor.
tions on the generation of turbulence due to gravity wave Soc., 96, 413–429.
breaking above deep convection. J. Appl. Meteor. Climatol., Satomura, T., and K. Sato, 1999: Secondary generation of gravity
47, 2777–2796. waves associated with the breaking of mountain waves. J. Atmos.
——, ——, T. L. Clark, and H.-M. Hsu, 2003: An investigation Sci., 56, 3847–3858.
of turbulence generation mechanisms above deep convection. Schwartz, B., 1996: The quantitative use of PIREPs in developing avi-
J. Atmos. Sci., 60, 1297–1321. ation weather guidance products. Wea. Forecasting, 11, 372–384.
——, J. D. Doyle, R. Plougonven, M. A. Shapiro, and R. D. Sharman, Sekioka, M., 1970: Application of Kelvin–Helmholtz instability to
2004: Observations and numerical simulations of inertia–gravity clear-air turbulence. J. Appl. Meteor., 9, 896–899.
waves and shearing instabilities in the vicinity of a jet stream. Shapiro, M. A., 1980: Turbulent mixing within tropopause folds
J. Atmos. Sci., 61, 2692–2706.
as a mechanism for the exchange of chemical constituents
——, ——, R. D. Sharman, M. A. Shapiro, and C. D. Watson, 2009:
between the stratosphere and troposphere. J. Atmos. Sci., 37,
Statistics and dynamics of aircraft encounters of turbulence
994–1004.
over Greenland. Mon. Wea. Rev., 137, 2687–2702.