Variations of vertical air velocity (W) in the midlevel shallow-layer clouds are described by a case study observed at West Sumatra, Indonesia (0.2°S, 100.32°E), in the nighttime between 8 and 9 May 2004. By receiving echoes from refractive index irregularities, W and spectral width (σW), used as a proxy of W turbulence, were observed both in clear and cloud regions using frequency power spectrum obtained by a 47-MHz wind profiler with 150-m vertical and 166-s time resolutions. Using altitude profiles of received signal intensity of a 532-nm Mie lidar (Plidar), altitudes with significantly larger Plidar than below (or above) were considered as cloud regions. Most of the shallow-layer clouds were observed between 6.0 and 8.5 km. In the top part of clouds (∼0–500 m below the estimated cloud tops), downward W up to ∼0.2–0.3 m s−1 and σW up to ∼0.5–0.6 m s−1 were observed. In the middle part of clouds (∼500–1000 m below the estimated cloud tops), W showed large variations. Both the standard deviation of W during the observation period and σW were large (∼0.5–0.7 m s−1). These results demonstrate that a combination of VHF wind profiler and lidar is useful to observe wind variations in and around midlevel shallow-layer clouds with high time and vertical resolutions. Altitude profiles of temperature observed by radiosondes showed that the air was absolutely stable near the top part of clouds and conditionally stable below. Possible relationship between W and temperature is discussed.
 Shallow-layer clouds are confined to shallow layers of air in which the rate of cooling resulting in cloud formation is rather slight [Houze, 1993]. These clouds are generally ∼1 km or less in vertical extent, and include low-level ones (fog, stratus, and stratocumulus), midlevel ones (altostratus and altocumulus), and high-level ones (cirrus, cirrostratus, and cirrocumulus). Though shallow-layer clouds can occur quite locally, they can also be very widespread, covering mesoscale or even synoptic-scale regions. This prosperity of shallow-layer clouds to cover great areas has a major impact on the climate of the earth through the absorptive and scattering effects of these cloud layers in the earth's radiation balance [e.g., Liou, 1986].
 Though mean vertical air motion in shallow-layer clouds is smaller compared with ones in convective clouds, previous numerical computations have shown that vertical air velocity (hereafter W) plays a crucial role in the formation and maintenance of shallow-layer clouds [e.g., Lin et al., 1998; Starr and Cox, 1985]. Further, W, radiation, and turbulent mixing all interact to give shallow-layer clouds their particular character, because the condensation in the shallow-layer clouds cannot be accounted for upward air motion alone [Houze, 1993]. To know relationship between W and dynamical and radiative processes associated with shallow-layer clouds, in situ W observations using aircraft [e.g., Gultepe et al., 1995] and remote ground-based observations using millimeter-wave radars which receive echoes from cloud particles [e.g., Shupe et al., 2008] have been carried out. However, continuous W observations both in and around shallow-layer clouds are scarce because means to measure W are limited.
 By receiving echoes from refractive index irregularities, VHF wind profilers typically operated near 50-MHz frequency (6-m wavelength) have the capability to continuously observe vertical profiles of vertical and horizontal air velocities both in clear and cloud regions [e.g., Fukao, 2007; Gage, 1990; Röttger, 1980]. VHF wind profilers are also able to retrieve raindrop size distributions in the lower troposphere due to their capability of receiving echoes from raindrops and atmospheric turbulence separately [e.g., Sato et al., 1990; Wakasugi et al., 1986]. However, VHF wind profilers are not sufficiently sensitive to detect small-sized cloud particles in shallow-layer clouds. Ground-based lidars and millimeter-wave radars, which are able to detect small-sized cloud particles, are useful for observing shallow-layer clouds [Sassen and Mace, 2002]. Therefore lidars and/or millimeter-wave cloud profiling radars have been additionally used for observing turbulence and air velocities in shallow-layer clouds. Using a VHF wind profiler and lidar in India, Kumar et al.  have shown a preliminary report on turbulence and winds in upper tropospheric tropical cirrus. Using a VHF wind profiler and scanning cloud profiling radar, Wada et al.  have shown the development of midlatitude frontal cirrus in the presence of shear instability and background upward wind. Recently, it has been shown that a combination of VHF wind profiler and cloud profiling Doppler radar is useful for retrieving particle fall velocity in cirriform clouds [Yamamoto et al., 2008].
 In this study, observational results of W and turbulence in and around midlevel shallow-layer clouds are presented using a VHF wind profiler and a 532-nm Mie lidar. A case at a tropical station (0.2°S, 100.32°E) during 8–9 May 2004 is presented. In section 2, data used in this study are described. In section 3, observational results are presented. Sections 4 and 5 offer discussion and conclusions, respectively.
2.1. The 47-MHz Wind Profiler
 A VHF wind profiler (hereafter wind profiler) used in this study is a monostatic pulse Doppler radar which operates at 47.0 MHz. The wind profiler is referred to as the Equatorial Atmosphere Radar, and has been operated at Kototabang, West Sumatra, Indonesia (hereafter KT; 0.2°S, 100.32°E, 865 m above the mean sea level). Table 1 lists the specifications of the wind profiler. The wind profiler uses a circular antenna array, approximately 110 m in diameter, which consists of 560 three-element Yagi antennas. To produce total peak output power of 100 kW, each antenna is driven by a solid-state transmitter-receiver module with 180-W peak output power. The active phased-array antenna system enables the wind profiler to steer radar beams on pulse-to-pulse basis. For the detailed description of the wind profiler system, see Fukao et al. . During the period focused on (from 8 to 9 May 2004), the wind profiler was operated with two observation modes; a 1-beam mode to observe W and a 5-beam mode to observe horizontal wind. Two observation modes were alternatively carried out. By pointing only to the vertical direction, signal-to-noise ratio of the spectral data obtained with the vertically pointed radar beam of the 1-beam mode improves ∼7 dB from one obtained by the 5-beam mode [Nishi et al., 2007]. Because of the signal-to-noise ratio improvement of the frequency power spectra, W was computed from the first-order moment (Doppler shift) observed by the vertically pointed radar beam of the 1-beam mode. Zonal (meridional) wind was computed by Doppler shifts observed by pairs of oblique beams of the 5-beam mode symmetrically offset from the vertical in the zonal (meridional) direction. Pure observation time of the 5-beam and 1-beam modes are 81.92 s and 78.6432 s, respectively. Because of additional time necessary for storing observational data into hard disk drives, vertical profiles of W and horizontal wind were produced every 166 s. During the period we focused on, the wind profiler observed weak horizontal wind less than 8 m s−1 at 6–9 km. Therefore possible misestimation of W in the presence of large horizontal wind and large vertical wind shear [e.g., Yamamoto et al., 2003] was negligible.
Table 1. Specifications of the Wind Profiler System
Monostatic pulse Doppler radar
Quasi-circular antenna array of 560 three-element Yagi antennas
Antenna beam width
3.4° (one-way half-power full width)
100 kW (maximum duty: 5%)
Number of transmitter/receiver modules
560 units (180 W peak power/each)
 Spectral width, defined as the square root of the second-order moment of the frequency power spectra, is used as a proxy of turbulence within the sampling time and volume. Because the two-way half-power full width of a radar beam and vertical resolution were 2.4° and 150 m, respectively, the wind profiler observed ∼300-m × 150-m volume at 8-km altitude. The observed spectral width contains artificial contributions due to processes such as wind shears across the sample volume and the effects of finite beam width. Using equation (16) of Nastrom , the effects of artificial contributions were removed. Equation (16) of Nastrom  requires information on altitude profiles of horizontal wind, antenna beam width, and range resolution to compute the artificial components. Altitude profiles of horizontal wind observed by the wind profiler were used. Antenna beam width of the wind profiler and the range resolution of the 1-beam mode were 3.4° (one-way half-power full width; see Table 1) and 150 m, respectively. Using equation (13) of Yamamoto et al. , the accuracy of observed W and spectral width were estimated. In equation (13) of Yamamoto et al. , accuracy of W and spectral width were determined by the radar wavelength, observation time, and number of incoherent integrations. In the 1-beam mode, they were 6.38 m, 78.6432 s, and 3 times, respectively. The value of coefficient k, which is listed in Table 1 of Yamamoto et al.  and used in equation (13) of Yamamoto et al. , was 0.38 (0.24) for W (spectral width) accuracy estimation. Using the above mentioned values, the accuracy of observed W and spectral width were estimated to be 0.044 m s−1 and 0.028 m s−1 for a spectral width of 1.0 m s−1.
2.2. The 532-nm Lidar
 A lidar was operated at KT with a laser wavelength of 532 nm, pulse energy of 20 mJ, and pulse repetition frequency of 10 Hz. A Schmidt-Cassegrain telescope with a diameter of 20 cm, a photoncounting photomultiplier tube, and a multichannel scaler were used to detect backscattered signals from atmospheric molecules, aerosols, and cloud particles. Altitude profiles of received signal intensity (hereafter Plidar) were recorded with 150-m and 10-min intervals. Owing to the receiver saturation by sunlight, the lidar was not operated from 0800 local standard time (hereafter LST) to 1600 LST. Except the period of 0800–1600 LST, the lidar was continuously operated. Note that LST is 7 h earlier than the universal time. Because microphysical properties were difficult to be computed without making assumptions, only information on cloud boundaries was used for data analysis.
 Using the characteristic that Mie scattering by cloud particles is much greater than molecular Rayleigh scattering, cloud bottom and cloud top altitudes were estimated using altitude profiles of Plidar. Value of Plidar was defined by number of photon count during the observation time (10 min). Background noise level was estimated and removed by averaging received signals in high altitudes where molecular scatterings were not detected. Cloud bottom altitudes were estimated by detecting an altitude where Mie scattering by cloud particles occurred. The first altitude zi (zi denotes ith sampling altitude) from below, where the vertical increase of Plidar defined by Plidar(zi+1) − Plidar(zi) was larger than the threshold value, was judged as the cloud bottom altitude. The threshold value was selected to satisfy that (1) the threshold value is small enough to detect small increase of Plidar due to Mie scattering by cloud particles, and that (2) the threshold value is large enough to avoid misestimation of cloud bottoms due to Plidar fluctuations caused by background noise. Threshold values were changed to select the best one that satisfied both the conditions (1) and (2). As a result, the threshold value of 400 was used to judge cloud bottom altitudes. Note that changing the threshold value (e.g., 350 and 450) did not significantly change the results shown in this study.
 The method used for estimating cloud top altitudes is described. When clouds had two or more layers, cloud top altitudes were computed for every cloud layer. For a cloud layer whose altitude was not the highest one, a ratio of vertical decrease of Plidar (hereafter Pratio) above the cloud bottom was computed by
Because the magnitude of Plidar showed large variations due to attenuation of transmitted laser light by cloud particles, Pratio, in which effects of Plidar magnitude were small, was used to estimate cloud top altitudes. Around the cloud tops, the source of lidar signal changed from cloud Mie scattering to molecular Rayleigh scattering. Therefore Pratio showed a large negative value in the top part of clouds and was small above cloud tops. The cloud top altitude was determined as the first altitude zi from below where ∣Pratio(zi)∣ became smaller than the threshold value, under the condition that Plidar vertically decreased between the range zi−1 and zi (Plidar(zi) − Plidar(zi−1) < 0). Threshold values to determine cloud top altitudes were changed to select the best one that detected rapid changes from cloud Mie scattering to molecular Rayleigh scattering. As a result, the threshold value of 1/3 was used to judge cloud bottom altitudes. Note that changing the threshold value (e.g., 1/4 and 1/2) did not significantly change the results shown in this study.
 For the highest cloud layer, the following conditions were used:
where σlidar is a standard deviation of background noise intensity. The cloud top altitude was determined as the first altitude zi from below which satisfied the conditions shown by equations (2)–(4). The conditions were used because equation (1) was not able to be used frequently due to Plidar attenuation by cloud particles. The conditions of equations (2)–(4) were also used when clouds had only one layer. In exceptional cases that cloud boundaries were not able to be determined using the fixed threshold values, cloud boundaries were determined manually by taking into account continuity of cloud boundaries in time and altitude.
 Cloud-boundary altitudes estimated by the lidar were compared with relative humidity (hereafter RH) observed by radiosondes to show that Plidar is useful to estimate cloud boundaries. Radiosondes using Vaisala RS92-SGP sensor were launched at KT. RH is expected to increase with altitude around a cloud bottom, and decrease with altitude around a cloud top. Figure 1 shows altitude profiles of Plidar and RH. RH showed a rapid increase around the cloud bottom altitude (6.04 km). RH which was 38% at 6.05 km vertically increased to 91% at 6.55 km. Around the cloud top altitude estimated by the lidar (7.84 km), RH showed a significant decrease. RH was larger than 92% below 7.75 km, showed a rapid vertical decrease above 7.85 km, and decreased to ∼81% at 8.05 km. The radiosonde was located at ∼10 km away from the lidar around 6.0 km, and ∼9 km away around 7.9 km. Though the radiosonde was away from KT, the cloud boundaries estimated by the lidar were consistent with the vertical changes of RH observed by the radiosonde. Figure 2 shows another altitude profiles of Plidar and RH. RH showed a rapid increase around the cloud bottom altitude (6.49 km). RH which was 33% at 6.55 km vertically increased to 96% at 6.85 km. Though the cloud bottom altitude was estimated by Plidar profiles with the time resolution of 10 min, RH observed by radiosondes was in situ value without time averaging. Further, the radiosonde was located at ∼9.5 km away from the lidar around 6.5 km. Though both the difference in time averaging and the distance between the lidar and radiosonde can cause the discrepancy between cloud bottom altitude and the RH profile, the difference between the cloud bottom altitude and the altitude where RH vertically increased (60 m) was less than the vertical resolution of the lidar (150 m). Around the cloud top altitude estimated by the lidar (8.44 km), RH showed a significant decrease. RH which was larger than 95% below 8.35 km showed a rapid vertical decrease (82%) at 8.45 km. The radiosonde was located at ∼8 km away around 8.4 km. Therefore, as shown in the previous case (Figure 1), the cloud boundaries estimated by the lidar were consistent with the vertical changes of RH observed by the radiosonde.
3.1. Cloudiness Around KT
 Hourly equivalent blackbody brightness temperature (hereafter TBB) observed by the IR1 (10.20–11.20 μm) sensor of the Geostationary Operational Environmental Satellite 9 (GOES-9) is used to describe horizontal distribution of cloudiness around KT (see Figure 3). At 1600 LST 8 May, a signature of clouds was not observed at KT (TBB was 289 K; Figures 3a and 4) . TBB at KT showed a decrease (264 K) at 1800 LST 8 May, which indicates the appearance of clouds at KT. Cloud region (TBB of ∼254–256 K) was observed at KT during 2000–2200 LST 8 May (Figures 3c, 3d, and 4). TBB at KT temporarily increased to 280 K at 2300 LST. Afterward TBB decreased again at 0000 LST 9 May, and was less than 260 K until 0600 LST. Clouds having TBB less than 250 K were observed mainly in the northwest of KT (Figures 3e–3g).
TBB distributions are further examined to show that deep convective clouds were not observed. High clouds having TBB less than 240 K (temperature at ∼10.3 km) were almost absent, and most clouds had TBB greater than 245 K (∼9.7 km). Because clouds are not regarded as perfect black bodies, cloud top temperatures inferred from TBB are generally higher than the actual temperatures at cloud tops. Therefore cloud top altitudes inferred from TBB are generally lower than real ones [Sherwood et al., 2004]. However, TBB results indicate that clouds were optically thin and convectively inactive even if their tops were higher than ∼10.3 km. Absence of deep cumulus convection occurred on a synoptic scale. A synoptic-scale cloud cluster, which moved from the Indian Ocean to the western Pacific Ocean and caused deep cumulus convection over Sumatra in the beginning of May 2004, already moved to the east of Sumatra during the period focused on (8–9 May 2004) [Seto et al., 2006; Yamamoto et al., 2007].
 Variations of shallow-layer clouds at KT are presented using the time-altitude plot of Plidar presented in Figure 5a. From 1600 LST 8 to 0600 LST 9 May, most of the observed clouds were located at 6.0–8.5 km. 9.4-GHz weather radar was operated at ∼20 km southeast of KT to observe precipitation at and around KT. During the period we focused on, precipitation particles inside the clouds were not detected by 9.4-GHz weather radar (see Figure 6 of Seto et al. ). Therefore the observed clouds were shallow-layer ones, and did not have precipitation particles whose diameter is typically larger than 0.5 mm. Plidar at ∼7.5–8.0 km was smaller (dominantly less than 104) during 2230 LST 8–0000 LST 9 May. This decrease of Plidar seems to be consistent with the significant increase of TBB at KT at 2300 LST 8 and 0000 LST 9 May (Figure 4), though Plidar attenuation by clouds at ∼6.5–7.0 km could be a factor that partly contributed smaller Plidar observed at ∼7.5–8.0 km.
3.2. Vertical Air Motion in Shallow-Layer Clouds
3.2.1. Top Part of Clouds
W in the top part of clouds is presented. Figure 5b shows a time-altitude plot of W. Within ∼500 m below the estimated cloud top altitudes, downward W was dominantly observed from 1740 LST 8 May to 0420 LST 9 May. Figure 5c shows a time-altitude plot of spectral width observed by the vertically pointed radar beam (hereafter σW). σW was almost always larger than ∼0.3 m s−1 within ∼500 m below the estimated cloud top altitudes. σW in the upper part of clouds further showed that not only downward W but also upward W existed within the sampling time and volume, though W averaged over the sampling time and volume was downward (see also Figure 5b).
 To quantitatively describe W variations in the top part of clouds, time series of W and σW at 8.07 km from 1600 to 2300 LST 8 May 2004 are shown in Figure 6a. When estimated cloud top altitudes were located around 8.44 km (1930–2150 LST 8 May; Figure 5c), W at 8.07 km was dominantly downward and ranged from −0.83 to 0.26 m s−1. Averaged W and σW at 8.07 km were −0.21 and 0.51 m s−1, respectively. These results further confirm the fact that not only downward W but also upward W existed within the sampling time and volume, though W averaged over the sampling time and volume was downward. Further, these results indicate that VHF wind profilers have a capability of continuously observing W and its turbulence, which is impossible by other observation means. At the altitude 150 m higher (8.22 km), averaged W and σW became smaller. They were −0.09 and 0.43 m s−1, respectively. At the altitude 300 m higher (8.37 km), turbulent W was almost absent. Averaged W and σW at 8.37 km were 0.01 and 0.20 m s−1, respectively.
 Similar features of W were observed in the shallow-layer cloud observed later. Figure 7a shows time series of W and σW at 8.22 km from 2300 LST 8 to 0600 LST 9 May 2004. When estimated cloud top altitudes were higher than 8.44 km (during 0040–0250 LST 9 May; Figure 5c), averaged W and σW at 8.22 km were −0.27 and 0.57 m s−1, respectively. At altitude 150 m higher (8.37 km), averaged W and σW became smaller. They were −0.04 and 0.41 m s−1, respectively. At the altitude 300 m higher (8.52 km), turbulent W was almost absent. Averaged W and σW at 8.52 km were 0.02 and 0.22 m s−1, respectively.
3.2.2. Middle Part of Clouds
σW frequently exceeded 0.5 m s−1 in the middle part of clouds (∼500–1000 m below the cloud tops), which shows that W in the middle part of clouds was highly turbulent (Figure 5c). To quantitatively describe time variations of W in the middle part of clouds, time series of W and σW at 7.62 km from 1600 to 2300 LST 8 May 2004 are shown in Figure 6b. When estimated cloud top altitudes were located around 8.44 km (during 1930–2150 LST 8 May), W at 7.62 km showed large changes even every observation intervals (166 s). W at 7.62 km ranged from −2.22 to 0.97 m s−1, and standard deviation of W computed during 1930–2150 LST (0.68 m s−1) was much greater than that at 8.07 km (0.28 m s−1; Figure 6a). These results suggest that turbulence in clouds passed over the wind profiler had small-scale changes in time. Clouds advected ∼1179 m within the observation interval (166 s), because horizontal wind velocity observed around 7.62 km was ∼7.1 m s−1 (not shown). As seen in the upper part, σW, which is a standard deviation of W within sampling time and volume, was large. σW at 7.62 km averaged during 1930–2150 LST was 0.66 m s−1. Large W changes seen at 7.62 km were also observed around the altitude. At the altitude 150 m lower (7.47 km), standard deviation of W and averaged σW computed during the period were 0.51 and 0.54 m s−1, respectively. At the altitude 150 m higher (7.77 km), both of them were 0.70 m s−1. Previous radiative transfer models have shown that radiation plays an important role in producing turbulence caused by instability in the clouds [Heymsfield et al., 1991; Starr and Cox, 1985].
Figure 6c shows time series of W and σW in or near the estimated cloud bottoms (7.02 km). During 1930–2150 LST 8 May, averaged W and σW were much smaller compared with those in the middle (7.62 km) and upper (8.07 km) part of clouds. They were 0.03 and 0.24 m s−1, respectively.
W variations in the shallow-layer cloud observed in another case are presented. Figure 7b shows time series of W and σW at 7.92 km from 2300 LST 8 to 0600 LST 9 May 2004. During the period when estimated cloud top altitudes were almost always higher than 8.29 km and cloud bottoms were lower than 7.69 km (from 2350 LST 8 to 0220 LST 9 May; Figure 5c), W at 7.92 km ranged from −1.22 to 0.71 m s−1. Standard deviation of W computed over the period (0.51 m s−1) was much larger than that at 8.22 km computed over 0040–0250 LST 9 May (0.25 m s−1; Figure 7a). σW at 7.92 km averaged from 2350 LST 8 to 0220 LST 9 May was 0.61 m s−1. Clouds advected ∼913 m within the observation interval (166 s), because horizontal wind velocity observed around 7.92 km was ∼5.5 m s−1 (not shown). At the altitude 150 m higher (8.07 km), standard deviation of W and averaged σW computed over the period were still large (0.49 and 0.54 m s−1, respectively). At the altitude 150 m lower (7.77 km), though the standard deviation of W was relatively small (0.33 m s−1), averaged σW was still large (0.52 m s−1). It is noted that in the latter case, upward W was not frequent even in the middle part of clouds (Figure 7b). As previously shown, the center of cloud region having TBB of less than 250 K was observed mainly in the northwest of KT during 0000–0200 LST 9 May (Figures 3e and 3f). This result suggests that the cloud generation region, where upward motions were expected to be found, was located in the northwest of KT.
Figure 7c shows time series of W and σW in or near of cloud bottoms estimated (6.87 km). From 2350 LST 8 to 0220 LST 9 May, averaged W and σW were much smaller compared with those in the middle (7.92 km) or upper (8.22 km) part of clouds. They were 0.05 and 0.17 m s−1, respectively.
 To describe possible relationship between temperature and W, altitude profiles of temperature are presented. Figures 8c and 8d show altitude profiles of temperature and temperature lapse rate (hereafter Γ) observed by the radiosonde located at 5.5–9.0 km during 1856–1906 LST 8 May. Temperature was 256–267 K at 6.0–8.0 km. Static stability was conditionally unstable below 7.0 km (Γ values were between the moist adiabatic lapse rate and dry adiabatic lapse rate of 9.8 K km−1). Γ, which was large (8.5 K km−1) at 7.0 km, rapidly decreased with altitude and was smaller than the moist adiabatic lapse rate above 7.1 km, which shows that the static stability was absolutely stable in the top part of cloud. Significantly downward W was observed at 7.17–7.47 km, where the static stability was higher (Figure 8b). Figures 9c and 9d show altitude profiles of temperature and Γ observed by the radiosonde located around 5.5–9.0 km during 0058–0109 LST 9 May. Temperature was 251–264 K at 6.5–8.5 km. Except around 7 km, the static stability was conditionally unstable below 7.7 km. Γ, which was large (9.3 K km−1) at 7.7 km, rapidly decreased with altitude and smaller than the moist adiabatic lapse rate above 8.2 km, which shows that the static stability was absolutely stable in the top part of cloud. Downward W was observed at ∼7.92–8.22 km, where the static stability increased rapidly with altitude and hence the static stability was higher (Figure 9b). Therefore W and Γ features observed in Figures 9b and 9d were consistent with those in the previous case (see Figures 8b and 8d).
 From model computation, Heymsfield et al.  have suggested that radiative cooling causes negative buoyancy in cloud top parcels, and that the negative buoyancy in cloud top parcels produces convective instability and downward W. They further have shown that entrainment at cloud tops is responsible for significant drying of downward W and acts to warm parcels and reduces the convective instability produced by radiative cooling. The mechanism found by Heymsfield et al.  is a candidate that explains the downward W, large σW, and relatively small Γ in the top part of clouds observed in our case study (also see Figures 5b, 5c, 6a, and 7a). Radiative cooling rate is roughly estimated using the relation
where Θ is potential temperature, z is altitude, p0 is 1000 hPa, p is pressure in hPa, is diabatic heating rate, and cp is specific heat at constant pressure (1004 J · kg−1 · K−1). Other terms in thermodynamic equation are neglected to simplify the estimation. When typical observed values of W = −0.25 m s−1, = 6 K km−1, and p = 400 hPa are used, /cp is estimated to be −100 K day−1.
 It should be noted that it is not able to conclude that the stable layer found in the top of clouds was the result of cloud processes or the cause that determined the cloud top altitudes. If the stable layer already existed when shallow-layer clouds formed, it can be a factor that determined the cloud top altitudes by prohibiting ascents of the air. However, the observational results of temperature suggest that such temperature profile is important to understand dynamical and microphysical processes of shallow-layer clouds.
 Below the altitudes where Plidar reached to its peak (∼7.24 km), the air was dominantly conditionally unstable (Figures 8a and 8d). In the later case, below the altitudes where Plidar reached to its peak (∼7.84 km), the air was also conditionally unstable (Figures 9a and 9d). These results suggest that cloud particles fell down from their generation region to evaporate in the lower region, and evaporative cooling could play some role in producing larger Γ in the lower part of clouds.
 In this study, variations of vertical air velocity (W) in the midlevel shallow-layer clouds have been described by a case study observed at West Sumatra, Indonesia (0.2°S, 100.32°E) in the nighttime between 8 and 9 May 2004. Using the capability of VHF wind profiler to observe W both in clear and cloud regions, W and spectral width (σW), a standard deviation of W within the sampling time and volume, were obtained by the 47-MHz wind profiler with 150-m vertical and 166-s time resolutions. Using altitude profiles of received signal intensity (Plidar) of the 532-nm Mie lidar, altitudes with significantly larger Plidar than below (or above) were estimated as cloud regions. Most of the shallow-layer clouds were observed dominantly at 6.0–8.5 km. In the top part of clouds (∼0–500 m below the estimated cloud tops), dominantly downward W up to ∼0.2–0.3 m s−1 and σW up to ∼0.5–0.6 m s−1 were observed. In the middle part of clouds (∼500–1000 m below the estimated cloud tops), W showed large variations. Standard deviation of W and σW averaged during the observation period had values up to ∼0.5–0.7 m s−1.
 Altitude profiles of temperature observed by radiosondes have shown that the air was absolutely stable near the top part of clouds and conditionally stable below. Though it is not able to conclude that the stable layer found in the top of clouds was the result of downward W produced by radiative cooling or the cause that determined the cloud top altitudes, it seems that such temperature profile is important to understand dynamical and microphysical processes of shallow-layer clouds.
 This study describes initial results to demonstrate that a combination of VHF wind profiler and lidar is useful to observe wind variations in and around midlevel shallow-layer clouds with high time and vertical resolutions. A combination of wind profiler and lidar would be useful for online monitoring of turbulence for air traffic control, and improving the parameterization of turbulence in and near shallow-layer clouds for in-flight turbulence forecasts. However, in this case study, the time resolution of the wind profiler and lidar (166 s and 10 min, respectively) were not enough to observe the detailed relationship between W and cloud properties near cloud tops. Phase of cloud particles (ice or liquid) is an important factor that determines turbulence and lifetime of clouds. From model computation, Starr and Cox  have shown that nighttime shallow-layer clouds composed of supercooled liquid water have much stronger turbulence produced by radiative processes and longer lifetime than clouds composed of ice crystals, because supercooled liquid water has smaller fall velocity than ice crystals due to its small size. Using a Raman/Mie lidar having higher detectability and an additional receiver to observe linear depolarization ratio, we are now trying to observe the relationship between dynamical and microphysical processes of shallow-layer clouds.
 The 47-MHz wind profiler belongs to Research Institute for Sustainable Humanosphere (RISH), Kyoto University, and is operated by RISH and National Institute of Aeronautics and Space (LAPAN), Indonesia. The operation of the 47-MHz wind profiler and radiosonde launching were supported by Grant-in-Aid for Scientific Research on Priority Area-764 funded by the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan. The work was supported by Hydrometeorological Array for Isv-Monsoon Automonitoring (HARIMAU) project of the Japan EOS Promotion Program (JEPP) and Grant-in-Aid for Young Scientists (B) (19740293) funded by MEXT. GOES-9 was operated by Japan Meteorological Agency, and TBB data used in this study were distributed by Kochi University.