JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. D19, 4595, doi: /2003jd003489, 2003

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. D19, 4595, doi: /2003jd003489, 2003 Gravity wave generation in the lower stratosphere due to passage of the typhoon 9426 (Orchid) observed by the MU radar at Shigaraki (34.85 N, E) S. K. Dhaka, 1,2 M. Takahashi, 1 Y. Shibagaki, 3 M. D. Yamanaka, 4 and S. Fukao 5 Received 9 February 2003; revised 24 April 2003; accepted 16 May 2003; published 2 October [1] Intense gravity wave activities were investigated in the lower stratosphere during the typhoon Strong vertical winds were observed just a few hours before the arrival of the typhoon-center at the MU radar site. About 30 min to 1 hour after the typhooncenter had passed, a considerable reduction in vertical wind amplitude was detected. Dominant gravity waves showed time period in the range of 7 8 min, 15 min, and min in the upper troposphere and lower stratosphere. In the vicinity of the central region of the typhoon, a gravity wave was observed, which was monochromatic in nature with a vertical wavelength 3 km between 1.5 km and 23 km height. In the lower stratosphere, the horizontal wavelength for the prominent period was detected in the range of km (for 15 min wave period) and km (for min wave period). The vertical wavelength of these waves was examined from 2.5 km to 4.0 km. In the horizontal direction, the intrinsic group velocity was estimated between 9 ± 2 and 11 ± 2 m/s. Near the tropopause, the average direction of group velocity was assessed at about 20 ±3 from the horizontal. The generation of gravity wave like features, in the lower stratosphere, is believed induced by convection, as the low temperature of the clouds indicates a deep penetration over the radar region as seen in the satellite GMS images. INDEX TERMS: 3314 Meteorology and Atmospheric Dynamics: Convective processes; 3334 Meteorology and Atmospheric Dynamics: Middle atmosphere dynamics (0341, 0342); 3329 Meteorology and Atmospheric Dynamics: Mesoscale meteorology; 3384 Meteorology and Atmospheric Dynamics: Waves and tides; 6952 Radio Science: Radar atmospheric physics; KEYWORDS: convective updrafts, gravity waves, troposphere, lower stratosphere Citation: Dhaka, S. K., M. Takahashi, Y. Shibagaki, M. D. Yamanaka, and S. Fukao, Gravity wave generation in the lower stratosphere due to passage of the typhoon 9426 (Orchid) observed by the MU radar at Shigaraki (34.85 N, E), J. Geophys. Res., 108(D19), 4595, doi: /2003jd003489, Introduction [2] Convection is considered to be a potential source of energy to produce wind and temperature anomalies on a wide temporal and spatial scale. In the tropics, information on gravity wave characteristics, on a comparatively short horizontal scale produced by convection, is important to understand their contribution to the mean-flow acceleration in the lower stratosphere [Beres et al., 2002; Takahashi, 1999; Alexander and Holton, 1997; Takahashi and Boville, 1992]. The impact of these waves at the higher altitudes has also been noted, particularly in modulating the dynamics in the mesosphere, and in E region and F region in the 1 Center for Climate System Research, University of Tokyo, Tokyo, Japan. 2 Now at Department of Physics, Rajdhani College, University of Delhi, India. 3 Osaka Electro-Communication University, Osaka, Japan. 4 Graduate School of Science and Technology, Kobe University, Japan. 5 Radio Science Center for Space and Atmosphere (RASC), Kyoto University, Uji, Kyoto, Japan. Copyright 2003 by the American Geophysical Union /03/2003JD ionosphere [Šauli and Boška, 2001]. In previous studies, measurements based on rockets and balloons showed an enhancement of wind and temperature variance in the equatorial region in comparison to middle and high latitudes. This is most likely from the presence of convection in the region [Eckermann et al., 1995; Allen and Vincent, 1995; Ogino et al., 1995; Tsuda et al., 1994; Karoly et al., 1996; Vincent and Alexander, 2000]. A few flight experiments were also conducted to observe the effects of both shallow and deep convections on the generation of gravity waves in its vicinity [Kuettner et al., 1987; Pfister et al., 1993; Alexander and Pfister, 1995]. In most of the studies, convection has been shown to act as a source of gravity waves generating a large amount of momentum fluxes in its vicinity. Flight experiment results provide an estimate for the horizontal wavelength of between 10 to 100 km. Forcing on the horizontal scale has been found to be associated with the excitation of convection. However, little information on the vertical structure of waves was obtained from these flight experiments. [3] Dewan et al. [1998] were the first to observe thunderstorm generated gravity waves imaged from the space. They confirmed circularly symmetrical gravity wave feature ACL 4-1

2 ACL 4-2 DHAKA ET AL.: GRAVITY WAVE GENERATION in the stratosphere, which were linked to deep convection and thunderstorm event in the tropics. These waves possessed a short horizontal wavelength in the range of 25 km to 50 km and had wave periods of 9 min to 15 min. Within 1 2 hours gravity waves had propagated to a level of near 40 km after generating in the upper troposphere. The temporal and spatial variability of vertically propagating convectively generated gravity waves is poorly understood because all currently available observational techniques have serious limitations in either time or space. Radar observations have high time resolution, but are limited to a few locations on the globe. Indian MST radar observations, during convection at Gadanki (13.5 N, 79.2 E) revealed the presence of high frequency gravity waves (periods min) with a short horizontal wavelength (10 20 km) [Dhaka et al., 2002, 2003]. Results agreed well with satellite observations [Dewan et al., 1998]. Dhaka et al. [2002] also showed a time delay of about 30 min for gravity waves to appear in the lower stratosphere after their generation from a convective storm in the mid-troposphere. The time delay in gravity wave response in the stratosphere was attributed to the finite group velocity possessed by them. [4] In order to investigate the mechanism involved in generating the convection waves (gravity waves due to convection), many numerical experiments have been performed [Lane et al., 2001; Lane and Reeder, 2001; Piani et al., 2000; Piani and Durran, 2001; Alexander et al., 1995; Pfister et al., 1993; Fovell et al., 1992]. These all suggested different mechanisms may be important including diabatic-heating source, obstacle effects due to turrets in the presence of strong wind shear, and oscillating convective updrafts in the convective environment. More precisely, three theories exist at present, which support the view that gravity waves are generated by convection. These are: [5] 1. Mechanical oscillator Effect. It is proposed that oscillating updrafts and downdrafts of air parcels impinge on the interface between the unstable and stable regions. This causes oscillating displacements of the isentropes at the base of the stable layer, which in turn excite vertically propagating gravity waves. The frequency of waves generated in this process is controlled by the frequency of an oscillator [Clark et al., 1986; Fovell et al., 1992; Alexander et al., 1995; Lane et al., 2001]. [6] 2. Deep heating. In this mechanism, thermal forcing by latent heat release within the convective storm is the main gravity wave forcing [Salby and Garcia, 1987; Bergman and Salby, 1994; Garcia, 2000]. However, control on the vertical wavelength of convectively generated gravity waves is not clear and no unified view emerged yet to resolve this problem. Holton et al. [2002] have argued that for forcing of a given frequency, waves of vertical wavelength of twice the depth of the heating will be adequately excited only if the horizontal forcing projects significantly onto horizontal scales compatible with the vertical-tohorizontal ratio given by the dispersion relationship. The preferred vertical wavelength depends on a non-dimensional parameter relating to the frequency and the horizontal and vertical scales of the forcing. Detailed mesoscale model simulations by Song et al. [2003] have suggested a dependence on the basic-state wind relative to the rearward propagating convective cells. Recently, an analytical studies was published on thermally induced internal gravity waves by Chun [1995], Chun and Baik [1998, 2002], and Baik et al. [1999a, 1999b]. [7] 3. Obstacle effect. In this mechanism, the gravity wave forcing occurs when the pressure field produced by a rising convective element acts as an obstruction to the environmental horizontal flow, producing upstream propagating waves in a manner similar to a flow over a mountain. The pressure field associated with a convective downdraft can of course generate a similar forcing. In either case, the mechanism is equally effective when there are persistent updrafts and downdrafts. This mechanism is also known as moving or transient mountain effect or quasi-stationary forcing [Pfister et al., 1993]. [8] The mechanism for generating gravity waves related to convection is unclear in deciding which of these mechanism dominate. Because of the poor understanding about the characteristics and mechanisms, the gravity wave drag (GWD) could not be fully parameterized for the general circulation models (GCMs). However, several attempts have been made to parameterize GWD induced by sub-grid scale cumulus convection in large-scale models under different theoretical conditions for determining cloud top stresses. Kim et al. [2003] have discussed these parameterization schemes. Their proper implementation in GCMs [Kershaw, 1995; Chun et al., 1996; Chun and Baik, 1998, 2002] has improved our understanding on atmospheric dynamics. [9] The range of spatial (scale of convection system) and temporal scales (oscillation period of convection) in the atmosphere is strongly coupled. Temporal resolution may be clarified and investigated by radar experiments in a more explicit way. In order to obtain a higher resolution in space and time that appears to be the main difficulty associated with understanding the climate system, radar observations across the globe proving to be more useful. However, radar observations provided high time resolution information on atmospheric motions only at a particular location. [10] In the present article, a case study is described to report the characteristics of high frequency gravity waves in the lower stratosphere associated with a typhoon underneath. The main purpose of this article is to examine short period gravity waves in the vicinity of convection as described by Dhaka et al. [2001, 2002, 2003] in radar observations at a tropical station, and in numerical simulation by Lane and Reeder [2001], Lane et al. [2001] and Piani and Durran [2001]. The typhoon 9426 (Orchid) which originated in the tropical region lasted long enough and finally transformed into an extratropical one. Typhoon movement is associated with convective environments, giving rise to unstable and turbulent atmospheric background. It passed over the MU radar at Shigaraki providing an opportunity to study features of the gravity wave in its vicinity. The effect of deep convection is shown to be the dominant factor in producing such high frequency gravity waves in the lower stratosphere. [11] It is informative to mention here that Sato [1993] studied in detail the gravity wave characteristics during the typhoon Kelly (October 1987) which also passed near the

3 MU radar site. However, she focused more on the characteristics of inertia gravity waves, which were resolved successfully, showing time periods (observed) of about 20 hours and 6 hours with horizontal wavelengths of 300 km and 600 km, respectively. Such wave disturbances lasted for more than 10 hours in the lower stratosphere. In addition, she also investigated small-scale wind disturbances, which were stronger in comparison to normal atmospheric conditions. The present results are compared to the data of Sato [1993] and the similarities and differences between them are emphasized. DHAKA ET AL.: GRAVITY WAVE GENERATION ACL Radar Data [12] In this study, the Middle and Upper atmosphere (MU) radar of the Radio Science Center for Space and Atmosphere, Kyoto University at Shigaraki ( E, N) was operated during the typhoon. The MU radar is a monostatic VHF- band Doppler radar with an active phased array system composed of 475 Yagi antennas [Fukao et al., 1985a, 1985b]. The MU radar is capable of deriving three components of the wind field at a very fine height and time resolution. Vertical sampling is obtained at a resolution of 150 m whereas the time resolution is 1 minute. The MU radar has the ability to detect atmospheric and precipitation echoes simultaneously [Fukao et al., 1985c; Wakasugi et al., 1986, 1987] and so in deriving the vertical velocities the effect of the precipitation echoes was taken into account [Shibagaki et al., 1997]. In this study, the radar was operated in two observation modes covering height ranges km and km at 2 min intervals respectively. The two observation mode data are combined to cover the altitude range of 1.5 km to 24.1 km. 3. Results 3.1. Background Atmospheric Condition Trajectory and Pressure Field [13] Figure 1 shows 12 hourly trajectory and surface central pressures of the typhoon 9426, as reported by Japan Meteorological Agency (JMA). The typhoon evolved from a tropical storm near Guam Island on 19 September It developed as it approached the Japan Islands. Near Minamidaitoujima ( E, N), the central pressure and maximum wind speed were 930 hpa and 47.5 m/s, respectively. Then it moved north northeastward and landed on the south coast of Kii Peninsula at 1930 LST 29. At the time of landing the central pressure and maximum wind speed were 950 and 40 m/s, respectively. Over the Kansai area, while moving toward Japan Sea, the central pressure increased (975 hpa) and winds were decreased (30 m/s). Finally, it transformed to an extra-tropical cyclone near Hokkaido at 1200 LST 30. Shibagaki et al. [2003] have reported the general structure of this typhoon. They focused on the meso-b and meso-g scale characteristics of the associated precipitating clouds. In front of the typhoon, outer and inner rainbands (meso-b scale) were identified. They argued that the tilted outflow regions accompanied the outer rainbands to the middle troposphere, and these originated from the maximum of cyclonic wind. It was suggested that these outflows contributed to the formation Figure 1. Trajectory and surface central pressures of typhoon 9426 as reported by Japan Meteorological Agency (JMA). Locations of the typhoon centers are indicated at an interval of 12 hours. and development of the outer rainbands and the upper-level band shaped cloud Wind and Temperature Field [14] Figure 2a shows the time-height cross section of the wind field from 1200 hrs local standard time (LST), 29 September to 0600 hrs LST, 30 September An arrow on the time axis represents the arrival time of the typhoon center over the MU radar. The diagram is constructed to show the horizontal wind behavior in vector form using the zonal and meridional wind components derived from the MU radar observations. In front of the typhoon, about 10 hours in advance, northward winds were dominant in the middle and upper troposphere, whereas in the lower troposphere, the wind prevailed in northwest direction. After the typhoon passed over the MU radar site, a dramatic change was observed in the direction of the horizontal mean wind, turning from northward to eastward, a change inherently associated with the rear of the typhoon. Above 14-km height, winds were of smaller magnitude during the entire time of observation. Mean winds were related essentially to varying magnitude of the radial and tangential winds close to the typhoon-center and its effective spatial region. [15] Altitude profiles of temperature, potential temperature, relative humidity and Brunt-Väisälä frequency (converted to periods) are shown in Figure 2b at 2033 hrs LST, 29 September The data were obtained using radiosonde flight. The radiosonde observation was taken about 2 hours prior to the arrival of the typhoon-center over the MU radar. The temperature profile (panel i) shows a

4 ACL 4-4 DHAKA ET AL.: GRAVITY WAVE GENERATION Figure 2. (a) The time-height section of horizontal mean wind observed by the MU radar during the typhoon. Time is shown on the x axis in hours (local standard time). An arrow on the timescale represents passing time of the typhoon center over the MU radar site. (b) Altitude profiles of the temperature (K) (panel i), potential temperature (panel ii), relative humidity (panel iii), and the Brunt-Väisälä frequency (converted to periods, panel iv). The radiosonde flight was at 20:33 hrs LST on 29 September 1994 about two hours before the arrival of typhoon-center over the radar site.

5 DHAKA ET AL.: GRAVITY WAVE GENERATION ACL 4-5 Figure 3. Horizontal distribution of taller clouds based on T BB using Japan Geostationary Meteorological Satellite (GMS) IR data at 2100 hrs LST 29 September. Filled circle and cross signs represent the location of the MU radar site and typhoon center respectively. minimum temperature (198 K) at about 17.4 km with a fine structure in the vertical direction. The potential temperature (panel ii) profile shows a wavy structure in the lower troposphere with no well-defined stratification. However, between 7 km and 13 km it seemed to be neutrally stratified. The vertical profile of the relative humidity is shown in panel iii. Background atmospheric air mass appeared to be saturated from ground level to 7-km altitude as the relative humidity was almost 100% except close to 4 km, where it was nearly 80%. Turbulent behavior was observed from 11 km to 16 km, which maximized in the presence of the typhoon-center over the MU radar. This was revealed using Doppler width data (a turbulent parameter in the radar observation) of the vertical beam (not shown). The vertical profile of the Brunt-Väisälä period is shown at panel iv in Figure 2b. The magnitude of the Brunt-Väisälä period was larger in comparison to normal atmospheric conditions both in the troposphere (12 13 min) and lower stratosphere (6 min). This is a typical atmospheric behavior revealed close to a typhoon center. Spikes were noted in the Brunt-Väisälä period at km height region, but these do not resemble seen with normal atmospheric behavior. These spikes primarily correspond to a turbulent region. However, the turbulence declined substantially after the typhoon had passed over the radar region. The typhoon central-region was warm compared to its front and rear region. Vertical gradient in temperature in the core shows a funnel type of structure, which is investigated by Shibagaki et al. [2003] using rawinsonde data of JMA stations GMS Images [16] During the typhoon, presence of high convective clouds was confirmed using T BB (K) (the area-mean equivalent blackbody temperature) data obtained from a geostationary meteorological satellite (GMS) in Japan. Figure 3 shows band shape clouds along with inner clouds and other active cloud region in the vicinity of the eye of the typhoon. The temperature near the eye of the typhoon was relatively low and of the order of 200 K, which indicates the clouds had penetrated deeply in the upper troposphere. The rainbands were associated with different cloud systems, which passed over the radar region (not shown). The temperature profile shown in Figure 2b indicates a low temperature of the order of 198 K near the tropopause (about 17 km), which confirms the presence of clouds at higher levels in the upper troposphere Observed Vertical Winds at the Front and Rear of the Typhoon [17] As this typhoon passed over the MU radar site, it offered a unique opportunity for studying the vertical wind

6 ACL 4-6 DHAKA ET AL.: GRAVITY WAVE GENERATION structure associated with its central region as well as its front and rear region. Over about 8 hours, the structural variability in vertical winds from 1.5 km to 23 km was examined. Figure 4 represents the vertical wind behavior from 1830 hrs LST to 2200 hrs LST, 29 Sep at the front of typhoon and from 2230 hrs LST to 0200 hrs LST at the rear of typhoon. Altitude profiles were constructed sequentially by averaging at a time-interval of 30 min. Profiles were plotted with a constant offset of 0.5 m/s to show progression with time. A change was noticed in the structure of vertical winds in the troposphere. With the passage of time (in passing from the front to rear of the typhoon) vertical winds around the mid-troposphere underwent a gradual transition, replacing the downdrafts by the updrafts. In the lower troposphere, the updrafts gradually replaced by the downdraft. Strong updrafts around 6-km altitude in front of the typhoon seem to have a relation with the topography effect of the Kii Mountains in the presence of strong cyclonic winds. Above 6 km, the downdrafts seemed to be associated with falling ice particles produced by the upper level updrafts. [18] The magnitude of vertical winds in the mid-troposphere was larger (within the range of ±0.8 m/s) in comparison to the lower stratospheric winds. The vertical wind profile at about 22:30 hrs represents a typical structure associated with the center of typhoon. The wave structure near the center of the typhoon was unique, which seems to be coupled with a monochromatic gravity wave in the height range of 1 23 km. Amplitude of the wave is decreased in the vertical direction, which indicates that it was an evanescent wave generated near the center of the typhoon Gravity Wave Pattern in The Lower Stratospheric Region [19] To investigate gravity wave structure on a fine temporal scale, the lower stratospheric region was of prime interest as a consequence of the perturbed tropospheric region due to the typhoon. Vertical beam data at an interval of 2 min were used to construct a time series at a height resolution of 300 m. The gravity wave pattern from 15 km (this level is adjacent to the tropopause) to 23 km is illustrated in Figure 5. A time series for the consecutive heights shifted vertically with a magnitude of 0.5 m/s to produce distinctive patterns. [20] Differences in the intensity of wave pattern between the front and rear of typhoon were clearly apparent over a deep layer at km. The wave amplitude at the front of typhoon was 0.3 m/s, which was reduced by a factor of about two at the rear. A signature of decreasing wave amplitude was evident after about 30 min of the typhoon passing over the radar site. After about 2230 hrs (time the typhoon-center passed over the radar site), between 15 km to 17 km altitude, a few bursts in the vertical winds were noticed for a further two hours. These bursts were identified correspond to the inner rain shield that had been observed near the typhoon eye. Subsequent to the rain shield passing away over the radar, the gravity wave pattern became smoother in the vertical direction from 15 km to 23 km. The presence of taller clouds (shown in Figure 3) was associated with strong activities in gravity waves. Absence of these clouds marked the signature of reduced activities Figure 4. Vertical wind profiles are constructed at an interval of 30 min covering about 8 hours. The profiles from 1830 hrs LST to 2200 hrs LST are at the front of the typhoon. At 2230 hrs LST, and from 2230 hrs to 0230 hrs LST the profiles represent the wind structure at the center and rear of the typhoon, respectively. A systematic change in vertical winds in the form of updrafts and downdrafts are evident at midtropospheric region. above 17 km that suggests a close relationship between deep convection and strong wave activities Power Spectrum Vertical Winds [21] In order to investigate the prominent periods in vertical winds, power spectrum using maximum entropy method (MEM) was applied to the time series. We used log scale on time axis, and linear scale on vertical axis to represent the product of power and frequency. This method is preferred as it has the advantage of resolving the short period adequately and minimizing any spurious peaks, especially those appearing in the low frequency region [Dhaka et al., 1993; Zangvil, 1977]. Time series at different levels were analyzed from 1800 hrs to 0230 hrs LST. [22] Figures 6 and 7 show the dominant periods present in the gravity wave patterns averaged between 15 to 18 km and 20 to 23 km height respectively. In order to examine the consistency in the prominent periods near tropopause, we represented the lower stratospheric region as two thick layers, the first layer from 15 to 18 km (Figure 6) and the second one from 20 to 23 km (Figure 7). A horizontal wind field around km height was noted with a small magnitude. Above 20 km, horizontal winds could not be computed due to the weak radar signal for the oblique beams. The spectral peaks of the dominant periods showed some consistency in both the thick layers kept in view the background mean wind. Figure 6 illustrates the dominance of the period below one-hour focussing around 7 8 minute, 15 minute, and minute. Spectral analyses at km altitude (Figure 7) revealed a strong power around 15-minute and a weaker power around 30 minute wave period. Above 20 km height, there is a gradual transition of dominant periods from min to 15 min i.e., maximum power was shared by a wave of 15 minute. Spectral peaks

7 DHAKA ET AL.: GRAVITY WAVE GENERATION ACL 4-7 Figure 5. Time-height section of the gravity waves pattern in the lower stratosphere from 15 km to 23 km. The time series is plotted at a height interval of 300 m. A vertical line is drawn at the time of typhoon passing at MU radar site. The left and right time-height sections represent the front and rear of the typhoon, respectively. Strong wave activity is evident at the front of the typhoon. were also verified using Fast Fourier Transform (FFT). It is clear from these MEM spectra and Figure 5 (a 2-D graph) that the strong activities of gravity waves with a period of less than one hour were pronounced in the convective environment. It is informative to mention here that a weak signal of a few hours was also present in the lower stratosphere, which is seen by applying a low-pass filter (cut off at about 90 min). In the tropospheric region, at a few layers of thickness 2 km spectral peak corresponding to a low frequency of periods about hours was seen Rainfall Data [23] Figure 8 shows the MEM spectrum for the rainfall data collected at the radar site during the typhoon. Rain data were used from 1800 hrs LST 29 September to 0300 hrs LST 30 September with a 5-minute data interval, therefore the nyquist frequency (1/2t) limits the calculation to periods 10 min. Rainfall data show a quasiperiodic behavior manifesting similar periods as noticed for the vertical winds with dominant periods at around min and min. The periods below 10 min are rejected and therefore are not shown in Figure 8. However, a strong peak was seen in the spectrum near 10 min, which is similar to the peak in vertical wind data. Spectral analysis of the rainfall data suggest that a quasiperiodic pattern is involved in the convection process passing over the MU radar, which may activate and/or organize new cells in the convection at intervals of near 15 minute and minute. The generation of the gravity waves in the convective environment is strongly believed to have an association with the convection processes. Possible mechanisms are discussed in section Wave Characteristics Vertical Wavelength [24] The vertical wavelength associated with gravity waves was examined during the typhoon, which is important to understand forcing in the vertical direction. Figure 9 shows the MEM spectrum for a 30-minute time average of vertical winds near the typhoon center. It is clear that a vertical wavelength of the order of 3-km is dominant. The center- region of typhoon seems to be unique. A monochromatic wave was present near the center of the typhoon as shown earlier in Figure 4. Therefore a single peak in the spectrum was seen for vertical wavelength 3 km between 1 to 23 km heights. This is unlike the other timings when an almost vertical wavelength larger by a factor of two dominated the troposphere. In most of the cases (in sets of 30 min time averages) the vertical wavelength investigated was 6 8 km in the troposphere (from Figure 4) and km in the lower stratosphere Horizontal Wavelength [25] Using dispersion relations for linear wave theory, an attempt was made to illustrate the wave parameters during the typhoon. The following relationship applies to horizontal wavelengths of less than a few hundred km and short periods. w 02 ¼ N2 k 2 m 2 þ k 2

8 ACL 4-8 DHAKA ET AL.: GRAVITY WAVE GENERATION Figure 6. Power spectra using Maximum Entropy Method (MEM) show the dominant periods in the vertical winds at 7 8 min, near 15 min and min in the lower stratospheric region from 15 km to 18 km. [26] Where symbols w 0, N, m, k represent intrinsic frequency, Brunt-Väisälä frequency, vertical wave number, and horizontal wave number, respectively. In the lower stratosphere the mean wind is of low magnitude, so intrinsic frequency is almost equal to the observed frequency. Observed Brunt-Väisälä frequency was approximately 2p/6 minute 1 in the lower stratosphere, which is explained by the radiosonde observation being taken when the central region of the typhoon was close to the radar site and the upper tropospheric region was highly turbulent. The horizontal wavelength was computed corresponding to dominant periods (i.e., minute and min) having a vertical wavelength km. Figure 10 shows the variation of the horizontal wavelength for different wave periods in the observed range of vertical wavelengths. As the wave period increases (followed by the increase in vertical wavelength from 2.5 km to 4.0 km) the horizontal wavelength also increased as shown for different periods in Figure 10. The computed horizontal wavelength for 10 Figure 8. Same as Figure 6 but for the rainfall data at the MU radar site. Spectrum shows the prominent periods at about min, 30 min, and 60 min. Around 10 min and below periods are not plotted as a limit of nyquist frequency. 15 minute wave period was about km. Horizontal wavelengths were computed at 18 km height and above as the observed dominant waves of time period minute were present at that height region. [27] The horizontal wavelength for waves of minute period was about km (on average, the vertical wavelengths were 3.25 ± 0.75 km). The observed dominant periods minute were restricted to an altitude region of km. In this height region, we compared the computed horizontal wavelengths with the observations. The wind contours at the front and rear of the typhoon are illustrated in Figure 11. The actual distance from the center of the typhoon to the front and rear positions is inscribed on the x axis. Grey scale is marked to show the magnitude of vertical winds, where the limit of the magnitude was about 0.8 m/s. The updrafts and downdrafts are distinctly viewed dominating in the lower and mid troposphere. In front of the typhoon, near 12-km high and above, gravity wave features on short horizontal scale with Figure 7. Same as Figure 6 but for the altitude range of km. Spectrum shows the dominant periods at about 15 min and 30 min. Figure 9. Shows the MEM spectrum of 30-min averaged vertical winds at the center of the typhoon. The dominant vertical wavelength is 3-km.

9 DHAKA ET AL.: GRAVITY WAVE GENERATION ACL 4-9 Figure 10. Shows the distribution of the horizontal wavelengths for a given range of dominant periods and vertical wavelengths. The horizontal wavelengths are computed in the height range of 15 km to 23 km. wavelength km are more evident between 100 km to 300 km at the front of the typhoon. In the lower stratosphere, on average, the horizontal wavelength was about km for the fine gravity wave pattern. This seems to be in good agreement with computed horizontal wavelengths (10 20 km). At an altitude of 16 km and below, computed horizontal wavelengths around km (for min period) match to the observed range of horizontal wavelength (25 50 km). About 14-km height, a horizontal wavelength about km is more apparent at the front of typhoon compared to its rear as clearly seen in Figure 11. The mean winds are very weak at around 14 km and above; gravity waves periods are almost intrinsic in nature and do not seem to be biased by the Doppler shift as expected in the case of strong winds in upper troposphere. Emerging pattern of gravity waves in Figure 11, and computed results both reflect the presence of short horizontal wavelengths in a convective environment. The results in Table 1 from different height regions are summarized. First column of the table represents the different height regions at which observed wave periods, horizontal wavelengths and vertical wavelengths are shown. In the observed wave period column bold numerals correspond to primary peak in the power spectrum. However, secondary peak is also shown in the same column. It is noted that preferably a single period dominates at a particular height region. Therefore we assume that the observed horizontal wavelength corresponds to the dominant period Estimates of the Intrinsic Group Velocity in the Horizontal Direction Corresponding to Different Wave Periods and Horizontal Wavelengths [28] In order to determine the intrinsic group velocity possessed by the dominant wave periods, we used the following simple dispersion relation [Dewan et al., 1998]. n o C gx x ¼ ðl x =tþ 1 ðt B =tþ 2 ; where symbol C gx represents the intrinsic group velocity and the symbols l x, t, and t B show horizontal wavelength, wave period, and Brunt-Väisälä period, respectively. The distribution of intrinsic group velocities at 10 km, 25 km and 40 km horizontal wavelengths is illustrated in Figure 12. Mean values for intrinsic group velocity corresponding to wave periods of 17 ± 2 min, 40 ± 10 min and 60 ± 10 min are 9 ± 2 m/s, 10.5 ± 3, and 11 ± 3 m/s, respectively. For a fixed horizontal wavelength, the intrinsic group velocity decreases (increases) for a wave of large (small) time period. On the other hand, a wave group of given periods will be rapid (slow) for long (short) horizontal wavelengths. If group velocity makes an angle f from the horizontal line then the magnitude of the intrinsic group velocity in that particular direction will be given by C g =C gx /cos (f). As Figure 11. Vertical wind contours plotted at the front and rear of the typhoon. Distance is marked from the center of typhoon in kms. A gray scale shows the magnitude of the vertical updrafts and downdrafts. Positive and negative values represent updraft (away from the radar in the vertical direction) and downdraft (toward the radar) respectively. The horizontal wavelengths for a fine and large gravity wave patterns may be inferred from the Figure.

10 ACL 4-10 DHAKA ET AL.: GRAVITY WAVE GENERATION Table 1. Characteristics of Gravity Waves in Different Height Region Height Region, km Observed Wave Period, min Horizontal Wavelength, km, l x Vertical Wavelength, km, l z (i)15 ± 2 a ± (ii)13 ± 2, 40 ± ± ± (iii)13 ± 2, 60 ± ± ± 0.8 a Bold numerals denote the dominant periods in the respective height region. shown in Figure 5 after passing out of the typhoon from the MU radar site, the amplitude of gravity waves decreased in the lower stratosphere, which became apparent about 30 min later. It is hypothesized that a variation in the wave amplitude is linked to the variability in energy source within the convective system. The temperature distribution in GMS images shows that higher cloud turrets were present during the typhoon over the radar region, which suggests deep convection in troposphere. The low temperature for the higher clouds was below 200 K (at 2100 hrs LST), and the temperature of the tropopause using radiosonde data around the same time (at 20:33 hrs LST) was about 198 K at 17.4 km (Figure 2b). This directly confirms penetration into the lower stratosphere, and is confirmed by enhancement in the Doppler width data indicating the ascent of turbulence from the lower troposphere to the upper troposphere possibly due to mixing induced by the clouds. [29] Bearing in mind the propagation time (30 60 min) needed to transmit the effect of wave group velocity to lower stratosphere, we determined the possible limits of the angle f using the relationship C g =C gx /cos (f). In the lower stratosphere, we examined wave activities in the height range of km. The gravity waves with a period of about 15 min possessed an intrinsic group velocity of about 9 ± 2 m/s (32.4 ± 7 km/h) in the horizontal direction. Given the group velocity effect in a 6 8 km deep layer within min, a simple calculation Figure 12. Shows the horizontal intrinsic group velocity (m/s) corresponding to the observed wave periods (at about 15 min, min, and min) and horizontal wavelengths (at 10 km, 25 km and 40 km). Figure 13. Direction representation of the group velocity shown near the tropopause level for a wave period at about 15 min. The angles 26 ±4 and 14 ±3 correspond to a wave group responding in 30 min and 1 hour, respectively. The average direction is shown by an arrow which makes angle 20 ± 4 from the horizontal. was performed to estimate the angle f from the horizontal. We used the analogy that similar timings will be required in experiencing the effect of group velocity from the tropopause level to 23 km and vice versa. It is estimated that the angle formed from the horizontal ranged between 26 ±4 to 14 ±3. Figure 13 shows a schematic of the group velocity arranged on the basis of transmission of energy in a depth of 6 8 km within time interval min. The angle computed using the relationship C gz /C gx = k/m = l z /l x (where l z and l x represents vertical and horizontal wavelengths, respectively) was of the same order as cited above. The average angle estimate is 20 ± 3 from the horizontal. It is obvious from Figure 13 that if the group velocity effect were faster in the stratosphere then the angle formed with the horizontal would be larger and vice versa. In this case, as the angle formed from the horizontal is small, the magnitude of the group velocity along these directions is slightly higher (less than 10%) compared to horizontal directions. [30] In a recent study of gravity waves over a thunderstorm (latitude 28.8 N, longitude 86.9 E) [Dewan et al., 1998] some characteristic features were revealed in a convective environment. The responding time between 23 km to 40 km was reported as about 2 hours. The angles for the two dominant wave periods (9 min and 15 min) were calculated at about 33 ±7 and 20 ±7 with a horizontal wavelength from 25 km to 50 km. The characteristic features of these two cases are in good agreement. However, the group velocities computed by Dewan et al. [1998] were about 3 times larger than those obtained here. It is important to mention here that they carried out observations up to 40-km altitude, whereas we restricted our observations to below 23 km as there were no data in the vertical direction to further confirm the variations in the horizontal wavelength. The main difference noted in the typhoon case is of a comparatively short horizontal wavelength (15 km) corresponding to a wave period of 15 min. However, the responding time for the

11 DHAKA ET AL.: GRAVITY WAVE GENERATION ACL 4-11 group velocity in the vertical directions seems similar in these two studies. 4. Summary, Discussion and Concluding Remarks [31] In this article, an effort has been made to study the characteristic features of high frequency gravity waves in the lower stratosphere during a typhoon. These waves were observed with a prominent period below 1 hour, peaking in the range of 7 8 min, 15 min, and min. Above 14-km wave frequencies were not affected much by the mean wind, as its magnitude was only of few m/s in the northeast direction. Therefore the observed frequencies are almost within the range of intrinsic frequencies. [32] The vertical wavelengths determined in the troposphere and lower stratosphere during the typhoon was of different magnitude. In the troposphere, the dominant vertical wavelength ranged from 6 km to 8 km. However, in the lower stratosphere it was reduced to km. The central region of the typhoon was shown to have a monochromatic wave behavior with a vertical wavelength 3km from the troposphere to the lower stratosphere (the central region seemed to have no convection as the rainfall suddenly decreased after the typhoon center passed over the radar site). The horizontal wavelengths determined for the fine and large gravity wave pattern were in the range of km and km, respectively. Sato [1993] pointed out the dominance of bit large-scale (several hundreds km horizontal wavelength) vertical wind disturbances during a typhoon passage as observed by the MU radar. She also reported the intense gravity wave activities covering a large frequency spectrum up to 24-km height. Some similarities were noticed between these two studies. For example, the vertical winds were reduced in the lower stratosphere soon after the typhoon passed over the radar site [Sato, 1993, Figure 7]. The time response in reducing the vertical winds was almost 1 hour similar to noted in the present study. Also, one important agreement between these two studies was the difference between convective activities at the front (highly convective) and rear (almost no convection) of the typhoon. However, the emphasis of this paper is different, and more focused on investigating the characteristic features of high frequency gravity waves, which was not attempted in the earlier study. [33] In the light of recent published 3-dimensional numerical simulation results for the convection and generation of gravity waves in its vicinity [Piani et al., 2000; Piani and Durran, 2001], the present observations are important in understanding some of the theoretical issues. The horizontal wavelength reported by Piani and Durran [2001] is in the range of km, which coupled the distance between the bursts of strongest convective updrafts in a convective system [see Piani and Durran, 2001, Figure 4]. Horizontal wavelengths in the lower stratosphere in this study were of similar magnitude. Further, we examined the power spectrum of rainfall data during the typhoon, and an inference is drawn that some sort of convection process took place/passed across the radar site at an interval of min and min from 1800 hrs LST to 0300 hrs LST. That convection process may have had a role in inducing gravity waves on the same timescale, although it may be a complex process related via cloud systems, which needs to be explored further. For example, we need to resolve whether it is the same cloud system that deepens at a very high altitude or whether in the lower tropospheric region there are shallow clouds responsible for the rainfall. However, there is evidence [Dhaka et al., 2001, 2002; Chu and Song, 1998] that strong updrafts and downdrafts are present and coexist with a large turbulence in the troposphere suggest a deep convection reaching to the level of tropopause height. The present case seems to be in favor of deep convection. Piani and Durran [2001] proposed that gravity waves generated in the lower stratosphere due to storms of different intensities in the troposphere all have similar amplitudes. The only difference noted was of distinct levels of neutral buoyancy in the upper troposphere. One convincing explanation seems to be that the propagation mode depends on the w and k values and the external mode decreases in the stratosphere which varies with different storm. We found that in the present study, on average, the characteristic features of the gravity waves may be explained in the lower stratosphere on the basis of linear wave theory, which supports the numerical simulation results. We also compared these results with the tropical deep convection event [Dhaka et al., 2002, 2003] and found that wave period and horizontal wavelengths are almost in same range in the lower stratosphere. However, there are differences in the nature of updraft formation in the troposphere and the level of neutral buoyancy. The updraft in the tropical region coupled with the strong convection is large (6 8 m/s) and the level of neutral buoyancy is high at about km. On the other hand, in the typhoon case (observations are at mid latitude) the updraft forms with magnitude 1 m/s (Figure 4) and the level of neutral buoyancy is near km. [34] In general, the properties of gravity waves generated due to convection in a 2-dimensional numerical model of Alexander et al., [1995] match well with the present observational results. They showed that long period waves (>1 hour) were not common, on the other hand high frequency waves with a period of 7.8 min and 12.8 min were found dominant. The horizontal wavelength computed in the simulation results was in the range of km. However, the vertical wavelength was a little larger (>5 km) compared to that found in the present study (<5 km), which is attributed to a deep heating in the troposphere. A mechanical oscillator pumping mechanism was proposed to be the most suited one to explain the characteristic features of gravity waves in their simulation. Fovell et al. [1992] showed similar general properties in a 2-dimensional numerical simulation of convectively generated stratospheric gravity waves. They found the vertically oriented high frequency gravity waves with a time period 28, 21, and 16 min, horizontal wavelength about 20 km, and the vertical wavelength km. One striking result was the similarity between the time period of rainfall rate oscillation and the gravity wave perturbation. A likely conclusion drawn from their study is that a primary mode of excitation of the gravity waves is the mechanical forcing from the oscillatory updrafts. [35] Lane et al. [2001] and Lane and Reeder [2001] in a recent 3-dimensional numerical model, resolved both the mesoscale cloud clusters and generation of gravity waves.

12 ACL 4-12 DHAKA ET AL.: GRAVITY WAVE GENERATION They reported that when convective clouds evolve, they excite gravity waves. The gravity waves are generated with the largest amplitude when the cloud top reaches the upper troposphere. Wave amplitude in the stratosphere is at about 0.3 m/s. Gravity waves were found to be monochromatic in nature with a horizontal wavelength 17 km and a vertical wavelength 5 km. The modeled gravity waves are approximately circular in nature, implying that convective clouds do not generate gravity waves, which prefer a particular direction. The width of the convective updraft, the degree of turbulent mixing, and the distance through which air parcels overshoot their level of natural buoyancy were the factors that control horizontal wavelength. Within limitations, the forcing mechanism was indicated to be from the oscillation of the convective updrafts around their level of neutral buoyancy. The region of excitation was detected about km. [36] In the present study, periodic updrafts were formed at about km height in front of the typhoon about 4 hours before its arrival. The oscillatory behavior of the updrafts and downdrafts (Figure 4) coupled with a 6 8 km vertical wavelength was distinctly visible in the tropospheric region, a finding which seems to support the Lane et al. [2001] mechanism of oscillatory updraft. As convective systems are complex, it is difficult to observe their evolution and internal structure. A close examination of the appearance of upper level clouds (duration 6 8 hours, Figures not shown) and an instantaneous correspondence in enhancing vertical winds between 10 to 14 km height reveals a proximate relationship between the convection cells and origination of gravity waves. Observed horizontal wavelengths of the order of km and km are supposed to be important in the convection model, as convection organizes in this range [Houze, 1993]. [37] It is believed that close to the source, the amplitude of convectively triggered gravity waves is roughly proportional to the vertical velocities in the convective cells. Above the updrafts (formed in the upper troposphere) in the overlying region (lower stratosphere) enhancements in vertical wind were noted after some interval of time. Such continuously oscillating convective systems are capable of producing high frequency gravity waves. The results in this article relate to a typhoon associated with convection over the midlatitude. However, we found that most of the characteristic features agreed with the numerical simulation of gravity wave generation by tropical convection, which indicate a consistency in results over tropics and extratropics. The vertical oriented oscillating updrafts seems to be the dominating mechanism from the study. However, it is not the only mechanism responsible for generating the gravity waves during convection. There remains the possibility of generating gravity waves near the cloud tops in the presence of strong mean wind (though mean wind was of a low magnitude in this case). Similarly, another possibility may be that there is a source of diabatic-heating near the center of the typhoon - deep eye wall convection in the typhoon which in turn may generate gravity waves. Therefore it is difficult to determine statistically the fraction of waves generated by different mechanisms. How strong gravity waves would generate in the stratosphere is also related to the typhoon intensity that in turn is affected by many factors including sea surface temperature (SST), vertical wind shear, and distance from the land etc [Baik and Paek, 2001]. None of these discussed here. In order to extend this study further in particular to focus on the dominant mechanisms involved in producing high frequency gravity waves, some of the observational results at a tropical station will be addressed in future work. [38] Acknowledgments. The MU radar belongs to and is operated by the Radio Science Center for Space and Atmosphere, Kyoto University. We acknowledge T. Osawa for providing GMS IR data, and Japan Meteorological Agency (JMA) for providing the trajectory and relevant information about the typhoon One of the authors (SKD) is thankful to CCSR, University of Tokyo for providing the working facilities. Authors are thankful to Robert Seviour for his kindly reading the manuscript to improve the English grammar. References Alexander, M. J., and J. R. Holton, A model study of zonal forcing in the equatorial stratosphere by convectively induced gravity waves, J. Atmos. Sci., 54, , Alexander, M. J., and L. Pfister, Gravity wave momentum flux in the lower stratosphere over convection, Geophys. Res. Lett., 22, , Alexander, M. J., J. R. Holton, and D. R. Durran, The gravity wave response above deep convection in a squall line simulation, J. Atmos. Sci., 52, , Allen, S. J., and R. A. Vincent, Gravity wave activity in the lower atmosphere: Seasonal and latitudnal variations, J. Geophys. Res., 100, , Baik, J.-J., and J. S. Paek, Relationship between vertical wind shear and Typhoon intensity change and development of three-predictor intensity prediction model, J. Meteorol. Soc. Jpn., 79, , Baik, J.-J., H.-S. Hwang, and H.-Y. Chun, Transient critical level effect for internal gravity waves in a stably stratified flow with thermal forcing, Phys. Fluids, 11, , 1999a. Baik, J.-J., H.-S. Hwang, and H.-Y. Chun, Transient, linear dynamics of a stably stratified shear flow with thermal forcing and a critical level, J. Atmos. Sci., 56, , 1999b. Beres, J. H., M. J. Alexander, and J. R. Holton, Effects of tropospheric wind shear on the spectrum of convectively generated gravity waves, J. Atmos. Sci., 59, , Bergman, J. W., and M. L. Salby, Equatorial wave activity derived from fluctuations in observed convection, J. Atmos. Sci., 51, , Chu, Y.-H., and J.-S. Song, Observations of precipitation associated with a cold front using a VHF wind profiler and ground -based optical rain guage, J. Geophys. Res., 103, , Chun, H.-Y., Enhanced response of a stably stratified two-layer atmosphere to low-level heating, J. Meteorol. Soc. Jpn., 73, , Chun, H.-Y., and J.-J. Baik, Momentum flux by thermally induced internal gravity waves and its approximation for large-scale models, J. Atmos. Sci., 55, , Chun, H.-Y., and J.-J. Baik, An updated parameterization of convectively forced gravity wave drag for use in large-scale models, J. Atmos. Sci., 59, , Chun, H.-Y., J.-H. Jung, J.-W. Kim, and J.-H. Oh, Effects of mountaininduced gravity wave drag on atmospheric general circulation, J. Korean Meteorol. Soc., 32, , Clark, T. L., T. Hauf, and J. P. Kuettner, Convectively forced internal gravity waves: Results from two-dimensional numerical experiments, Q. J. R. Meteorol. Soc., 112, , Dewan, E. M., et al., MSX satellite observations of thunderstorm-generated gravity waves in mid-wave infrared images of the upper stratosphere, Geophys. Res. Lett., 25, , Dhaka, S. K., A. Kumar, and O. P. Nagpal, Some studies of tropical/equatorial waves over Indian tropical middle atmosphere: Results of the tropical wave campaign, Meteorol. Atmos. Phys., 51, 25 39, Dhaka, S. K., P. K. Devrajan, Y. Shibagaki, R. K. Choudhary, and S. Fukao, Indian MST radar observations of gravity wave activities associated with tropical convection, J. Atmos. Sol. Terr. Phys., 63, , Dhaka, S. K., R. K. Choudhary, S. Malik, Y. Shibagaki, M. D. Yamanaka, and S. Fukao, Observable signature of a convectively generated wave field over the tropics using MST radar at Gadanki (13.5 N, 79.2 E), Geophys. Res. Lett., 29, , Dhaka, S. K., M. Takahashi, Y. Kawatani, S. Malik, Y. Shibagaki, and S. Fukao, Observations of deep convective updrafts in tropical convection and their role in the generation of gravity waves, J. Meteorol. Soc. Jpn, in press, 2003.