The diurnal variability of raindrop size distribution (DSD) in precipitating clouds over Kototabang, West Sumatra, Indonesia (0.20°S, 100.32°E), is studied using three types of Doppler radars, operated at VHF (47 MHz), UHF (1.3 GHz), and X band (9.4 GHz) frequencies. Two precipitation events from 5 to 6 May 2004 in the first observation campaign of the first Coupling Processes in the Equatorial Atmosphere (CPEA-I) project reveal a difference between clouds precipitating in the early afternoon and clouds precipitating in the nighttime. In the early afternoon, the precipitating clouds were dominated by shallow convective types with high rainfall rate at the surface. In the nighttime, precipitating clouds were dominated by stratiform types with small rainfall rate at the surface. A diurnal variation of horizontal wind was observed over this area. The westerly in the lower troposphere and the easterly in the middle troposphere began to be enhanced in the afternoon (1400–1700 LT). DSD parameters were retrieved from VHF band Doppler radar data. A modified gamma distribution was used to model DSD parameters. The shape parameter (μ) was larger during stratiform precipitation than during shallow convective precipitation events, as shown by previous studies. During stratiform rain events on 5 May 2004, the median volume diameter (D0) was dominantly greater than 1 mm, which is larger than D0 during shallow convective rain events. Results presented in this paper indicate that DSD has a diurnal cycle over the mountainous region of Sumatra.
 The Indonesian maritime continent is one of the most active convection regions on the globe [Ramage, 1968]. So far, many studies have shown interannual and seasonal variations of convective activity over the Indonesian maritime continent [e.g., Hamada et al., 2002; Aldrian and Susanto, 2003]. Convection over the Indonesian maritime continent is also significantly modulated by intraseasonal variations (ISV) [e.g., Seto et al., 2004] and with periods of less than several tens of days [e.g., Tsuda et al., 1994]. Other than convection with a period of longer than several days, the diurnal cycle of convection induced by land- sea breeze circulation is also prominent in the Indonesian maritime continent [e.g., Houze et al., 1981; Nitta and Sekine, 1994; Hadi et al., 2002; Mori et al., 2004], because the Indonesian maritime continent is composed of many islands and the surrounding sea.
 Sumatra is one of the largest islands in the Indonesian maritime continent and is located at the eastern edge of the Indian Ocean (around 100°E). Because of its unique topography, Sumatra plays an important role in convection around 100°E. There is a mountain range along this southwestern coastline with an average height of ∼2000 m (see Mori et al. [2004, Figure 4] for a detailed map of Sumatra). Active convection with a diurnal cycle exists along its southwestern coastline [Mori et al., 2004]. Several studies have revealed the formation mechanism for active convection over Sumatra by topographically induced local circulation. From observations in the mountainous area of Sumatra, it has been shown that deep convective events tend to occur from the early afternoon to the evening [Renggono et al., 2001; Wu et al., 2003]. By using a cloud-resolving model, Sasaki et al.  showed that thermally induced upslope winds which converge over the mountain summit in the daytime play an important role in the formation of mesoscale precipitation systems with a diurnal cycle over western Sumatra and the adjacent sea region. Further, Renggono et al.  showed the dominance of stratiform clouds in the nighttime.
 Microphysical or cloud-modeling studies require knowledge on the rain drop size distribution (DSD) to understand the mechanism for precipitation development and to parameterize these processes in cloud-scale and mesoscale numerical models. To clarify the characteristics of DSD over the Indonesian maritime continent is important because the most active cumulus convection occurs in the region.
 Over Sumatra, the variations of DSD in relation to ISV have been recently reported by Kozu et al.  by using data derived from disdrometer and VHF band Doppler radar which is named Equatorial Atmosphere Radar (EAR) data. However, variations of DSD in relation to the diurnal cycle of convection is not clarified over Sumatra.
 To clarify the vertical coupling processes of the atmosphere over the Indonesian maritime continent, the Coupling Processes in the Equatorial Atmosphere (CPEA) project carried out a first observation campaign (CPEA-I) in the troposphere from 10 March to 9 May 2004 [Fukao, 2006]. During CPEA-I, the observation parameters of EAR were set to several modes, but only on 5–9 May 2004 they were set to a special mode to intensively observe the vertical wind in the troposphere. In this study, a preliminary report on diurnal features of DSD over the mountainous region of Sumatra from 5 to 6 May 2004 is presented using the fine-resolution vertical wind results derived from EAR data.
2. Data Description
 Three kinds of radars, operated at VHF band (47 MHz), UHF band (1.3 GHz), and X band (9.4 GHz) frequencies, were used in this study.
2.1. VHF Band Wind Profiler
 A VHF band wind profiler, the Equatorial Atmosphere Radar (EAR), was used for the retrieval of DSD and the investigation of the time-height variation of winds over this area. EAR is installed at Kototabang, located in the mountainous region of Sumatra (see Figure 1). From 5 to 9 May 2004, EAR was operated in two observation modes. The observation parameters of EAR are shown in Table 1. The pulse repetition frequency was 2,500 Hz in both observation modes. One mode was a standard observation mode to monitor vertical and horizontal winds in the troposphere. The observation time for this mode was 82 s, and the vertical wind Doppler velocity resolution was 0.195 m s−1. The beam direction was changed on a pulse-to-pulse basis. Horizontal wind data were derived from the standard observation mode. The other mode was a special mode (hereinafter called vertical mode) to monitor only the vertical wind in the troposphere. In this mode radar beams were pointed only to the vertical direction through an observation time of 79 s. Three Doppler spectra collected with the same observation time were averaged offline prior to the retrieval of DSD. Since this mode monitored only the vertical direction, an SNR improvement of 7 dB over the standard observation mode was expected. This improvement of SNR was important for retrieving the DSD during weak stratiform precipitation events. Furthermore, Doppler velocity resolution was 0.041 m s−1. This improvement of Doppler velocity resolution contributed to the better estimation of DSD by increasing the number of data points used for DSD retrieval.
Table 1. Principal Online Observation Parameters of EAR
 A UHF band wind profiler is used for observing wind and hydrometeors in the lower troposphere, including the planetary boundary layer (PBL) [e.g., Gage et al., 1994; Carter et al., 1995]. A UHF band wind profiler, Boundary Layer Radar (BLR), is located at Kototabang Global Atmosphere Watch (GAW) station, about 300 m from the EAR site. The radar beams of this BLR were electronically steered to the same directions as those of the EAR standard mode. The original BLR data had a vertical resolution of 150 m and time resolution of about 1 min. Previous results of BLR observation in Indonesia have been reported by Hashiguchi et al. , Renggono et al.  and so on. Using data derived from BLR, Renggono et al.  have shown that BLR can be used to classify precipitating cloud types by slightly modifying the algorithm proposed by Williams et al. . In this study, BLR data obtained by the vertically pointing beam are analyzed to classify precipitating cloud types.
2.3. X Band Doppler Weather Radar
 X band Doppler weather radar (XDR) can detect precipitating clouds within an observation range of 68 km. XDR is operated at 9.445 GHz and with 40 kW peak transmitted power. From 10 April to 9 May 2004, XDR was operated in a volume scan mode with 17 zenith angles from 0.7° to 40.0°. The time and range resolution of XDR were 4 min and 250 m, respectively. To observe precipitating clouds over Kototabang, XDR was installed approximately 20 km southeast of the EAR site (0.36°S, 100.41°E, 1121 m above MSL; see Figure 1). Precipitating clouds over Kototabang can be observed with XDR up to 14 km altitude. Radar reflectivity data is used for calibrating the DSD estimation by EAR (see Section 3.2.2 for details).
3.1. Precipitation Cloud Classification
 The classification of precipitating clouds in this study is based on the algorithm adopted from Williams et al. . In the present study each sample of data is analyzed to determine which of four types of precipitating clouds, that is, stratiform, mixed stratiform-convective, deep convective, and shallow convective clouds (hereinafter referred to as STR, MIX, CNV, and SHL, respectively), came from. The algorithm to classify precipitating clouds is based on a judgment of the presence of a melting layer and the presence of turbulence or hydrometeors above the melting layer. If BLR data reveal the existence of a melting layer (i.e., large received signal at 0°C isotherm level), a precipitating cloud type is classified as STR or MIX. If BLR data do not show the existence of a melting layer, the precipitating cloud type is classified as CNV or SHL. MIX type is classified if a broadening of spectral width above the melting layer, which indicates turbulent motion above the melting layer, is observed. CNV is classified if hydrometeors (i.e., presence of echoes) appears above the melting level. For more details of the classification method, see Williams et al.  and Renggono et al. .
Williams et al.  determined the onset of precipitation events by surface rain gauge and classified precipitation events every 30 min using a UHF band wind profiler. However, by modifying the determination of the onset of precipitation events, Renggono et al.  have applied this method for analyzing the diurnal and annual variations of precipitating clouds at Kototabang. Instead of using surface rain gauge, they used radar reflectivity and Doppler velocity at 1 km height from the surface obtained with the vertically pointing beam of BLR to determine the onset of precipitation events. Murata et al.  used the same method for the determination of precipitation events over Kototabang.
3.2. Retrieval of DSD
3.2.1. Mathematical Description of Doppler Velocity Spectra
 To retrieve DSD, we used Doppler spectral data obtained with EAR. In this study, we used the vertical mode of EAR observation, which has single-beam observation in the vertical direction only (see Section 2.1 for details). We assume that DSDs have the form of a modified gamma distribution,
where N0, μ and Λ represent the concentration parameter, shape and slope factors of the distributions, respectively. The relationship between drop size D (mm) and drop fall velocity v (m s−1) is expressed as
 The measured Doppler spectrum in precipitation condition is the sum of the two spectra corresponding to echoes from precipitating particles (hereinafter called Sp(v)) and refractive index irregularities from turbulent air (hereinafter called St(v)). The observed Doppler spectrum S(v) can be expressed as
where S0(v) is normalized from St(v), Pn is the noise level on the spectra, and W(v) is the window function [Wakasugi et al., 1986]. The asterisks (*) denote the convolution operation.
St(v) is approximated by the following Gaussian function,
where P0 is echo power, is mean wind velocity along the radar beam direction and σ is spectral width.
 With no atmospheric turbulence or wind Sp(v) may be expressed in terms of drop size (D) as
where v(D) is measured positive upward and C is a radar system constant.
Sato et al.  proposed a method of DSD retrieval by assuming the Marshall-Palmer distribution for the rainfall particle size. In the present study, the fitting method is improved by applying a modified gamma distribution to estimate precipitation parameters.
Figure 2 shows an example of Doppler spectra observed with EAR during a shallow convective precipitation event (Figure 2a) and a stratiform precipitation event (Figure 2b). The peak near 0 m s−1 corresponds to echoes from refractive index irregularities of turbulent air, and the smaller peak on the left corresponds to precipitation echoes. The estimated precipitation parameters for convective precipitation are μ = −0.6 and Λ = 7.06 mm−1, and for stratiform precipitation are μ = 3.9 and Λ = 4.95 mm−1. It is clear here that μ (Λ) for shallow convective precipitation is smaller (larger) than μ (Λ) for stratiform precipitation, which means that the number of large-sized raindrops is greater in stratiform clouds than in shallow convective clouds. Tokay et al.  have shown a similar result.
3.2.2. Determination of N0
 To obtain N0, we compared the radar reflectivity factor estimated from spectra obtained with EAR with the radar reflectivity factor obtained with XDR. Radar reflectivity factor Z is given by
If the modified gamma distribution is assumed to be N(D), equation (6) is expressed as
by applying equation (1) [e.g., Williams, 2002], where Γ is the complete gamma function. N0 is computed from μ and Λ estimated from EAR data, and Z obtained from XDR.
4. Large-Scale Convective Feature Around Sumatra
 From 10 April to 9 May 2004, convection over Kototabang was significantly modulated by an ISV event. Several large-scale cloud clusters developed over the Indian Ocean and reached Sumatra. Convection over Kototabang was formed within a large-scale convective envelope during the active phase of ISV [Kozu et al., 2005]. Figure 3 shows a time-longitude section of equivalent black body temperature (TBB) at 0.2°S from 1 to 9 May 2004. From 4 to 6 May 2004, the last large-scale cloud cluster appeared over the radar site, and convection over Kototabang developed within a large-scale convective envelope. It is to be noted that a clear diurnal variation is seen in TBB around Kototabang. Cloud tops were higher (or colder) at night (18—−0300 LT), consistent with results shown in a previous study [e.g., Mori et al., 2004], and suggesting that a topographically induced local circulation plays an important role in the development of convection in the mountainous region of Sumatra. After 7 May convection over Kototabang was not observed as the large-scale convective envelope moved east of 100°E.
5. Convection Over Kototabang
5.1. Convection on 5 May 2004
 Convective features over Kototabang during 5–6 May 2004 were measured by radars. Figure 4 shows observations from radars and surface rain gauge from 0900 LT on 5 May 2004 to 0600 LT on 6 May 2004. Precipitation events were observed during 1300–2230 LT on 5 May and during 0030–0300 LT on 6 May (Figure 4d). Strong updraft greater than 0.4 m s−1 was observed at 2–5 km altitude during 1300–1700 LT and during 1830–2030 LT (Figure 4a). The appearance of this updraft was associated with a large Z (greater than 22 dBZ) at 2–5 km altitude (Figure 4b) and high rainfall rate (1.0–15.5 mm h−1) at the surface (Figure 4d). Shallow convective precipitating clouds were dominant in the same period (Figure 4b). It is consistent that Z was significantly weakened above the melting level (∼5.2 km altitude). Strong updraft of greater than 0.6 m s−1 first appeared at 7–9 km altitude around 1945 LT, when precipitating clouds started to change from shallow convective type to deep convective type (see Figure 4b). Afterward, updraft of greater than 0.2 m s−1 continuously existed above 8 km. These updrafts found in the middle troposphere caused a development of stratiform precipitating clouds observed by XDR and BLR after 2045 LT. After 14 LT westerly winds below 3 km and easterly above 6 km altitude strengthened with time (Figure 4c).
 From 2000 to 2300 LT cloud tops higher than 8 km altitude were observed at Kototabang (Figure 4b). Stratiform precipitating clouds were dominant during the same period (Figure 4b). Weak rainfall of less than 2.5 mm h−1 was observed at the surface (Figure 4d). Below 3 km altitude, downdraft was dominant (Figure 4a). During 2300 LT on 5 May to 0300 LT on 6 May a weak Z of 7–22 dBZ was observed at Kototabang. Echoes from precipitation particles obtained by XDR disappeared after 0300 LT on 6 May.
5.2. Convection on 6 May 2004
Figure 5 shows observations made with radars and surface rain gauge during 0600 LT on 6 May to 0300 LT on 7 May 2004. In the morning (0600–1030 LT) downdraft was dominant below 3 km altitude (Figure 5a). Echoes from precipitating particles were not observed at this time (Figure 5b). From 1030 to 1230 LT updraft greater than 0.2 m s−1 was observed below 3 km. Afterward high rainfall rate (1.0–15.5 mm h−1) was observed at the surface during 1400–1730 LT (Figure 5d).
 Updraft greater than 0.6 m s−1 was observed at 2–4 km altitude during 1400–1630 LT (Figure 5a). This strong updraft was associated with a large Z of greater than 17 dBZ below 6 km altitude (Figure 5b). Although BLR was not in operation before 1530 LT, it was clear that the precipitation event during 1400–1730 LT showed similar features in relatively high rainfall rate at the surface, low echo top observed by XDR, and updraft limited in the lower troposphere as observed on 5 May. The westerly below 3 km altitude and easterly above 6 km altitude strengthened after 17 LT (Figure 5c).
 Stratiform precipitation was observed from 2030 LT on 6 May to 0100 LT on 7 May. During this period downdraft was observed at 3–6 km altitude (Figure 5a). On the other hand, updraft was dominant above 6 km altitude. However, precipitation at the surface was almost absent during this stratiform precipitation event (Figure 5d). Echoes from precipitating particles vanished after 0100 LT on 7 May 2004 (Figure 5b).
5.3. Zonal Movement of Precipitating Clouds Around Kototabang
 The zonal movement of precipitating clouds around Kototabang during this period is shown in Figure 6. On 5 May 2004 precipitating clouds developed from 1200 LT about 10–30 km to the west of Kototabang (Figure 6a), around which a steep hill surrounding Maninjau lake is located (see Figure 1). Shallow convective clouds over Kototabang observed from 1300 LT were transported from here by low-level westerly (Figures 4b and 4c). However, cloud tops higher than 7 km altitude were observed only in the west of Kototabang during 1200–1800 LT.
 After 2000 LT precipitating clouds with tops higher than 8 km altitude propagated from the east of Kototabang (Figure 4b). These clouds developed 30–40 km to the east of Kototabang (Figures 6a and 6b). Easterlies were dominant above 6 km altitude (Figure 4c) and caused the westward movement of stratiform precipitating clouds. This kind of cloud system, which moves eastward in the afternoon and westward in the late evening, was reported by Shibagaki et al. .
 On 6 May 2004 a similar pattern in the diurnal variation of precipitating clouds was observed. From 1215 LT precipitating clouds developed about 10–30 km west of Kototabang (Figure 7a). These clouds moved eastward and caused rainfall events over Kototabang from 1400 LT. From 2030 LT precipitating clouds with stratiform features moved eastward and covered Kototabang (Figure 7b). These clouds developed 30–40 km east of Kototabang from 1800 LT. However, Z was generally weaker on 6 May than on 5 May 2004.
6. Raindrop Size Distribution
Section 5 showed that precipitation events during 5–6 May 2004 were dominated by shallow convective and stratiform types and that the precipitation events in the afternoon and in the nighttime had different characteristics in the vertical structure of Z and the rainfall rate at the surface. This section describes the DSD characteristics of each kind of precipitation event. DSD parameters (N0, μ, and Λ) are computed from EAR spectra at 3 km height above the surface (or 3.85 km altitude) during the precipitation events of 1200–2400 LT on 5 May 2004. 47 sets of DSD parameters from shallow convective rain and 27 sets from stratiform rain are selected on the basis of the cloud type classified by BLR.
 DSDs for shallow convective and stratiform precipitation clouds are shown in Figure 8. It was found that DSDs during shallow convective rainfall events are characterized by a large amount of small-sized rather than larger-sized drops (Figure 8a). The average values of μ and Λ were −0.3 and 10.9 mm−1, respectively. The median volume diameters, which can be expressed as D0 = (3.67 + μ) Λ−1 for a modified gamma distribution, are between 0.18–0.8 mm (see Figure 9). On the other hand, the DSD during stratiform rainfall events had a drop size distribution broader than that seen during shallow convective events (Figure 8b). The average values of μ and Λ were 2.5 and 5.9 mm−1, respectively. The N(D) peaks of 0.3–0.8 mm were dominant while D0 was 0.4–1.9 mm (see Figure 9). This is consistent with the previous study on the point that μ and D0 are larger during stratiform precipitation events than during shallow convective precipitation events [Tokay et al., 1999]. Tokay et al.  also showed that Λ is smaller during stratiform precipitation events than that during shallow convective precipitation events.
 The time variation of DSD on 5 May are shown in Figures 9a and 9b. Precipitation events during 1400–1700 LT were dominated by shallow convective type (Figure 4b). Total rainfall was 17.6 mm at the surface (Figure 4d). The distribution of raindrops was concentrated on the small size (smaller than 0.5 mm). The relatively broad distribution with larger D0 (about 0.7 mm) around 1420 LT was associated with high rainfall rate at the surface (15.5 mm h−1) and Z of greater than 22 dBZ up to 6.0 km altitude (see Figures 4b and 4d).
 During 1900–2200 LT two types of DSDs were detected (Figure 9b). The first type, which appeared during 1925–2010 LT, had the same characteristics as the precipitation in the afternoon (1400–1700 LT). During this period precipitating clouds were dominated by the shallow convective type (see also Figure 4b). Therefore the DSD consisted of a large amount of small drops with smaller D0 between 0.2–0.8 mm. The second type, which appears after 2040 LT, had different characteristics. During this period the precipitating clouds were dominated by the stratiform precipitation type (see also Figure 4b). The peak of N(D) was not as large as that during the shallow convective events, but the DSD had a broad distribution. The peak of N(D) was around 0.5 mm, and D0 was larger than 1.0 mm.
 Similar patterns were found in DSD during precipitation events on 6 May 2004 (Figure 10). Precipitation events in the afternoon (from 1425 to 1605 LT) show DSDs characterized by narrower distribution. D0 of around 0.2 mm was predominantly observed, though the number of profiles was small and D0 was variable at that time. It is consistent with the smaller rainfall rate at the surface and the weaker Z than those for the shallow convective case on 5 May (see Figures 5b and 5d). During 2140–2240 LT the DSD was concentrated in the smaller drop size (around 0.5 mm), but during 2315–2335 LT a larger D0 of about 1.0 mm was observed. However, D0 in this stratiform case is smaller than that seen in the stratiform case on 5 May (see Figure 9). The smaller D0 on 6 May is consistent with smaller Z on that day (Figures 4b and 5b).
7. Discussion and Conclusion
 In this study, the diurnal features of DSD over Kototabang on 5–6 May 2004 have been studied. Doppler spectra with fine resolution were obtained by pointing radar beams of EAR only to the vertical direction. A diurnal variation of convection was clearly seen at Kototabang on both 5 and 6 May. In the afternoon (1300–1730 LT) precipitating clouds were dominated by shallow convective clouds. They had their origin about 10–30 km west of Kototabang. Strong updraft and high rainfall rate at the surface were also observed in the presence of these shallow convective clouds. In the nighttime (2000–0300 LT) stratiform clouds, which developed about 30–40 km to the east of Kototabang, caused mainly stratiform precipitation events at Kototabang. Renggono et al.  statistically showed a diurnal variation of precipitating clouds over Kototabang by classifying precipitating clouds seen by BLR. They showed that convective precipitating clouds are dominant in the afternoon whereas stratiform precipitating clouds are dominant in the nighttime. The diurnal variability of convection over Kototabang during 5–6 May 2004 agrees well with their statistical results.
 DSDs for shallow convective and stratiform cases were retrieved from Doppler spectra obtained with EAR. A modified gamma distribution was assumed as the form of the DSD. Different characteristics were found during precipitation in the afternoon (shallow convective precipitation) and the nighttime (stratiform precipitation). In the afternoon precipitation events are dominantly characterized by smaller D0. This variability of D0 in relation to cloud type is consistent with previous studies [Tokay et al., 1999; Cifelli et al., 2000].
 A diurnal variation of the zonal wind was also observed during 5–6 May. A westerly below 3 km (easterly above 6 km) intensified after 1400–1700 LT. However, the relationship between the diurnal variations of the zonal wind and convection needs further investigation.
Kozu et al.  studied the variability of DSD in relation to ISV and found that the DSD is broader during the inactive phase of ISV than during the active phase. Seto et al.  studied convection over Kototabang in relation to ISV and found that the deep convective events caused by topographically induced local circulation are dominant in the inactive phase of ISV. They further found that both convective and stratiform type precipitating events are observed at Kototabang in the active phase of ISV [see Seto et al., 2004, Figure 14].
 In this study we have shown a diurnal feature of DSD during the active phase of ISV (see Figure 3). However, the diurnal variability of the DSD may change with ISV. Statistical features of DSD over Kototabang will be studied by using EAR data in subsequent studies.
 The authors thank Toshiaki Kozu of the Department of Electronic and Control Systems Engineering, Shimane University, and Yasushi Fujiyoshi of the Institute of Low Temperature Science, Hokkaido University, Japan, for providing XDR data. The authors would also like to thank Masayuki Ohi of Hokkaido University for his effort in preparing the radar system and the operators who maintain and operate EAR and BLR. They thank William L. Oliver for careful reading of the manuscript. Acknowledgments extend to the three anonymous reviewers for their constructive comments. The present study was partially supported by Grant-in-Aid for Scientific Research on Priority Area-764 of the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan. The first author was supported by the Japan Society for the Promotion of Science (JSPS) RONPAKU (Ph.D. dissertation) Program.