The Tibetan Plateau (TP) in China’s southwestern region is the highest plateau with the most complex terrain in the world. It covers one-fourth of the total territory of China. The average elevation of the TP reaches up to the mid troposphere. For this reason, it is also called “the Roof of the World”. By its thermal-dynamical forcing, the TP significantly affects the atmospheric circulation, the surface-atmosphere momentum exchange, and the hydrologic cycle in its surrounding areas as well as in eastern China. Clouds and precipitation in the TP impose great impacts on the atmospheric moisture transport and surface heating. Under certain favorable synoptic conditions, various weather systems above the TP can move out of the plateau, leading to disastrous weather such as rainstorms in the downstream region. Cloud and microphysics processes above the TP are often different from those in low altitude areas. Due to the strong surface heating, factors that suppress the development of convection often disappear quickly after the noon in the TP. As a result, convective cloud is easy to develop. Convective processes are more frequently triggered in the TP than in the downstream plain areas. However, convective available potential energy (CAPE) is often small because of the low moisture content in the atmosphere above the plateau. The top of the cumulus clouds and that of the strong echoes from the radar are both low due to the small CAPE and low atmospheric moisture content above the TP. The horizontal scale of the convective system is often limited, too.

Due to the relatively rare field experiments and observations of clouds and precipitation in the TP, our underst and ing of the microphysical processes in clouds and precipitation over the plateau is still very limited. Large uncertainties exist in various numerical modeling studies of cloud physics over the TP, and some critical parameters that are used to describe cloud physics and precipitation process are probably not appropriate for the TP. As a result, most of the present numerical models cannot reasonably simulate the cloud microphysics processes over the TP, resulting in large biases in cloud radiative forcing and precipitation simulation. The above factors significantly deteriorate the capability of numerical models in their simulation and forecast of the cloud physics and precipitation. In addition, weather stations are scarce in the heartl and of the TP while from central to western Tibet are large uninhabited regions. Satellite measurements have become a necessity to obtain high-density observations over the TP. However, satellite measurements must be continuously calibrated by using surface observations. The lack of ground observations in the TP remains a major bottleneck affecting the calibration and quality control of satellite data. Therefore, it is imperative to conduct comprehensive observations of clouds and precipitation over the TP.

During the first TP Atmospheric Science Experiment conducted in 1979, scientists have already recognized the importance of precipitation observation. Two conventional X-b and 711 type radars were set up in Naqu, central TP, and Lasa (southern plateau) to measure precipitation. Based on the observations obtained during the experiment, Qin (1983) analyzed the statistical characteristics of cumulus clouds in Naqu and revealed the relationship between vertical distribution of moist static energy and convective development. In 1998, China Meteorological Administration and Chinese Academy of Sciences jointly launched the second TP Atmospheric Science Experiment. Scientists from China and Japan collaborated on measurements and studies of the energy and water cycles over Naqu region in the TP (GAME-TIBET). The X-b and Doppler radar from Japan, rain gauges, and radiosondes were deployed to collect comprehensive information about precipitation process in this region. Based on these measurements, characteristics of radar echoes from convective precipitation were analyzed; changes in the convective process before and after the monsoon onset and the diurnal variation of convection were explored. Precipitation structures by ground-based radar and TRMM (Tropical Rainfall Measuring Mission) precipitation radar were compared (Liu et al., 1999,2002; Shimizu et al., 2001; Uyeda et al., 2001; Liu,2003; Fu et al., 2006; Zhuang et al., 2013).

Satellite measurement is always an important approach for meteorological studies over the vast TP. Based on satellite measurements, TRMM precipitation radar observations, and cloud radar and lidar measurements from Cloudsat, many previous studies have analyzed the typical cloud structures related to deep convective precipitation in the TP. These studies revealed the statistical characteristics of convective precipitation in the plateau. Differences in characteristic cumulus clouds between the TP and other regions of East Asia have also been discussed. It is found that the deep convection developed over the TP usually remains weak with a feature of small horizontal scale due to the relatively dry environment and low CAPE (Fu et al., 2007; Li et al., 2009; Dai et al., 2011; Wang et al., 2011; Li et al., 2012; Cai et al., 2012). The microphysical structures of stratus clouds formed over Qinghai and the eastern TP region have also been analyzed based on in-situ aircraft and radar observations (Li and De Ligeer,2001; Zhao et al., 2002; Liu et al., 2008)

China Meteorological Administration has set up a new generation weather radar system in the TP for operational measurement of various precipitation processes. Unfortunately, the observations for precipitation are far less than satisfactory due to the topographic beam blocking. It is well known that weather radars (S-b and , C-b and , and X-b and weather radars) are mainly employed to obtain the three-dimensional structure of precipitation echoes, whereas their capability for cloud observation is very limited. Millimeterwavelength cloud radar and lidar are two primary tools for cloud observations.

Based on the above discussion, weather radars have been applied for precipitation observation over the TP (Zhuang and Liu, 2012) while satellite remote sensing has been used for cloud observation. However, little observation has been done in comprehensive and continuous measurements of microphysical parameters in the cloud physics using various active remote sensing technologies over the TP. Due to the lack of cloud observations, little research has been done on microphysical processes within cumulus clouds over the TP, and appropriate determination of the important parameters in the cloud physics still remains a question that has not been answered yet. So far, our knowledge about cloud physics over the TP is very limited. Meanwhile, new cloud observation technology such as millimeter-wavelength cloud radar has been applied in field experiments conducted in Guangdong and Yunnan provinces. The data analysis method has been developed by Liu et al. (2014) specifically to process cloud radar observations. The study by Liu et al. (2014) provides a solid basis to retrieve important cloud physics parameters for cumulus clouds over the TP, which is important for the scientific community to better underst and the temporal-spatial variation of clouds and precipitation over the TP, and further investigate the cloud physics and precipitation process. For this purpose, the reanalysis dataset based on multib and radar observations has been produced and applied for the development of cloud physics schemes and retrieval of parameters.

In the third TP Atmospheric Science Experiment in 2014, intensive observations were conducted from 1 July to 31 August 2014. Various vertically pointing radars, lidars, and dual-polarization radars were combined with passive remote sensing techniques to measure atmospheric water vapor, clouds, and precipitation during this intensive observation period. The millimeter-wavelength radar, C-b and frequency modulation and continuous wave (FMCW) radar, and Cb and dual-linear polarization radar, which represent the most advanced atmospheric observation technique and have been developed independently in China, were utilized to obtain first-h and field measurements. These measurements are valuable for cloud and precipitation studies in the TP.

In this paper, the instruments used in cloud and precipitation observation in this experiment are introduced. The observational data obtained in this experiment are described. Based on the cloud radar measurements, we analyze the statistical characteristics of clouds (cloud top and base, cloud depth, vertical distribution of clouds, etc.) in the summer over Naqu region. The radar echo structures of several types of typical clouds over the plateau are revealed and interesting results are provided.

The field experiment on clouds and precipitation is an effective approach that can help address the following scientific questions: (1) proposing a method to conduct comprehensive cloud and precipitation measurements using various radars; (2) providing quality control algorithms for observations; and (3) developing a method to retrieve cloud dynamical and microphysical parameters. The purpose of this study is to further underst and the microphysical processes involved in clouds and precipitation and their spatial-temporal variations. Multi-wavelength active remote sensing and passive remote sensing techniques are combined to obtain both micro- and macro-structures of atmospheric water vapor, clouds, and precipitation over the TP, which form the basis for the development of retrieval method for cloud and precipitation microphysics. Data obtained in this experiment will help to reveal the microphysical processes in clouds and precipitation over the TP, and provide evidence for correction and calibration of satellite remote sensing observations. These observations are also valuable in modeling studies of cloud and precipitation physics as well as in the development of parameterization schemes in numerical prediction models.

Most of the previous field experiments only used Doppler weather radar for the measurement of precipitation system. Neither the hydrometeor phase distribution in precipitation system nor the cloud process could be measured in these experiments. In the field experiment of the present study, various advanced multiple wavelength radar systems are utilized to obtain not only the macro-feature of precipitation echoes, but also the three-dimensional wind fields, phases of precipitation particles, and raindrop size distribution. The millimeter-wavelength radars are applied in this study for continuous measurements of the vertical structure of clouds, which can be retrieved to obtain the vertical profiles of microphysical and dynamical cloud parameters. These results are important for further studies of cloud physics over the TP.

The measurement instruments, location, and observational periods are given in Table 1. The exteriors of the instruments are shown in Fig. 1. Major technical specifications of the instruments are listed in Table 2.

Table 1 Measurement instruments, locations, and periods

Table options

Fig. 1 The main instruments used in the experiment. (a) Ka-band cloud radar, (b) C-band frequency modulation and persistent wave (FMCW) radar, (c) C-band mobile polarization radar, (d) vapor and cloud observation lidar, and (e) Ku-band micro-rain radar.

Figure options

Table 2 Major technical parameters of the cloud radar, continuous wave radar, and C-band dual-linear polarization radar

Table options

The Ka-b and solid-state transmitter-based millimeter wave vertically pointing cloud radar was used for continuous (uninterrupted) cloud measurements. The reflectivity, radial velocity, velocity spectrum width, and linear depolarization ratio were obtained. Meanwhile, power spectral density data were selected for further analysis. C-b and frequency modulated continuous wave (FMCW) vertically pointing radar system is the first frequency modulated continuous wave radar system used in China for cloud and precipitation observations. It utilizes continuous wave, all-phaseparameter Doppler radar to measure the reflectivity, radial velocity, and velocity spectrum width of cloud and precipitation at various levels with a vertical resolution of 15-30 m. Ka-b and micro-rain radar manufactured in German is used to measure the reflectivity of clouds and weak precipitation and power spectrum, which are used to retrieve the vertical profiles of raindrop size distribution for precipitation below 6 km (0.109-6-mm diameters) and total precipitation, etc. The atmospheric water vapor and clouds lidar used in this study transmitted three wavelengths (1064,532, and 355 nm) to measure and retrieve the mixing ratio of water vapor, cloud depolarization ratio, cloud base height, atmospheric extinction coefficient, and atmospheric backscatter coefficient profile. The maximum detection range of the system is 20 km, the range for water vapor mixing ratio is 0.2-5.0, and 0.2-15.0 km for cloud base and cloud depolarization ratio. These vertically pointing radars and lidars with three different wavelengths constitute the comprehensive observing system for measurements of clouds and precipitation with various intensities and at their different developing stages.

The portable C-b and dual-linear polarization radar operates with polarization base of simultaneous transmissions of horizontal and vertical radar wave with simultaneous reception using dual receivers to obtain the reflectivity, radial velocity, velocity spectrum width, differential reflectivity, differential phase, copolar correlation coefficient, etc. During the period of the field experiment, the radar scanned 9 elevation tilts once every 5 min. Together with the China new generation C-b and weather radar operated in Naqu Bureau of Meteorology, they constitute the dual-Doppler radar observation system.

In order to detect the characteristics of raindrop size distribution, and verify the accuracy of raindrop size distribution retrieved from radar observation, measurement of surface raindrop size distribution was conducted in this experiment using the HSCPS32 disdrometer. The HSC-PS32 disdrometer can detect liquid and solid particles. Diameters of liquid particles that can be detected are within 0.2-5.0 mm, while the diameters of solid particles are within 0.2- 25.0 mm. To coordinate with the retrieval of cloud water content from the vertically pointing radar measurements, microwave radiometer was used in this study. An MP-3000A 35-channel microwave radiometer made by U.S. Radiometrics Corp. was deployed. It can produce high-resolution temperature, relative humidity, and water vapor profiles from the surface to 10-km height. It also produces low-resolution liquid water profiles and relatively accurate total liquid water content.

The site for cloud and precipitation observation is located in Naqu of TP. This site is selected mainly because Naqu is the major area of the TP vortex genesis, where convective processes develop frequently, making it an ideal location for clouds and precipitation observation. The vertical measurements of cloud properties were conducted at Naqu Bureau of Meteorology (NQMET,31.48°N,92.01°E,4507 m AGL (above ground level)). The C-b and dual-linear polarization radar is installed at Naqu Climate-Environment Observation Station, Cold and Arid Region Environmental and Engineering Research Institute, Chinese Academy of Sciences (NQBJ,31.37°N,91.90°E,4509 m AGL). However, during the observational period, the C-b and continuous wave radar interfered with the China new generation weather radar installed at NQMET. For this reason, the C-b and continuous wave radar and a disdrometer were set up at Naqu Zhongxin Hotel (NQZX,31.29°N,92.03°E,4507 m AGL) for observation. The distance between NQZX and NQMET is about 2 km, and the intensive observation period is from 1 July to 31 August 2014.

When the intensive observation started on 1 July, most of the instruments worked properly, and continuous measurements of cloud evolution were obtained. In particular, continuous measurements have been conducted by the cloud radar and C-b and continuous wave radar since both radars use solid-state transmitters, which ensures the reliability and stability of the radar operation. However, the operation of some instruments became unstable due to the influence of the high elevation. For example, the microwave radiometer could not provide satisfactory measurements. Table 3 lists the instruments and the variables they can observe. The spatial-temporal resolution, the total amount of data, and the major weather processes involved during the observation period are also given in

Table 3 Raw measurements, spatial-temporal resolution, total amount of data, and weather processes involved during the observation period

Table options

Continuous evolution of the cloud vertical structures was obtained from cloud radar measurements during the period of 5 July to 4 August 2014. Based on the radar measurements during this period, we analyzed the diurnal variation of cloud base and top, cloud depth, number of cloud layers, and vertical distribution of clouds. Quality control has been performed first to remove echo interference and noise. The cell identification-based method was then applied to cloud classification, while the upper and lower boundaries of the cloud measured by radar were taken as cloud top and base respectively. The average cloud top and base, cloud depth, and number of cloud layers were then calculated. The results of the statistical analysis of these variables for the period 5 July to 4 August and their diurnal variations are analyzed. The diurnal variations of cloud top and base are shown in Fig. 2. The height shown in Fig. 2 and hereafter all refer to the height above the ground level. Figure 2 indicates that the clouds over the TP are generally classified into high clouds (with cloud top above 6 km), and mid- to low-level clouds with a top below 4 km. Few clouds with a cloud top of above 5 km are found. The top of high-level clouds above 6 km demonstrates a distinct diurnal variation. It reaches the highest level during 1600-2000 BT (Beijing time), and becomes the lowest during 0800-1200 BT. In contrast, no distinct diurnal variation can be found for the top of mid- to low-level clouds. Further analysis indicates that altostratus and altocumulus clouds account for parts of the high clouds after the deep convection dissipates, while the mid- to low-level clouds include developing cumulus and stratocumulus clouds.

Fig. 2 Diurnal variations of cloud top and base during 5 July-4 August 2014.

Figure options

In order to further analyze the cloud distribution at different altitudes, we analyzed the occurrence frequency of cloud at various altitudes. The occurrence frequency of cloud at a specific layer is defined by the ratio of the number of radar beams with significant cloud detection at the altitude to total number of radar beams. Note that the occurrence frequency of cloud defined here is closely related to the minimum reflectivity radar observed. Under the same condition, the occurrence frequency of cloud decreases with height because the minimum reflectivity increases with height. Figure 3 presents the diurnal variation of the occurrence frequency of cloud at 1-h interval and 1- km interval at different vertical levels. It is found that clouds over the TP largely distribute below 10 km, and the occurrence frequency of cloud above 10 km is less than 10%. The vertical distribution of clouds is clearly stratified. Few clouds are found at 5-km height, while large amounts of clouds concentrate at levels of 2-4 and 6-9 km. Clouds rapidly develop at levels above 5 km after the noon, especially during 1800-0400 BT. High frequency of cloud formation is found at levels between 6 and 9 km, where the number is larger than 50%. Above 6-9 km, parts of the clouds are deep cumulus clouds and parts are altostratus and altocumulus clouds that form after the dissipation of deep convection.

Fig. 3 Occurrence frequency of cloud at various levels during the period of 5 July to 4 August 2014.

Figure options

Cloud depth is defined as the difference between the heights of cloud top and base. The cloud depths are made detailed statistical analysis at 1-h and 1-km intervals at vertical direction. Clouds at multiple levels are considered. The diurnal variation of cloud depths distribution is shown in Fig. 4. It is found that the frequency of cloud deep deeper than 5-km clouds is less than 15%, while shallow clouds with a cloud depth less than 2 km has the maximum frequency. Note that there is no distinct diurnal variation in shallow clouds, whereas clouds deeper than 5 km show significant diurnal change. The depth of deep clouds increases rapidly after 1200 BT, reaches the maximum value at around 1400 BT, and maintains until before it starts to decrease in the morning. At 1000 BT, clouds deeper than 5 km are seldom to be observed.

Fig. 4 Diurnal variation of cloud depth from 5 July 2014 to 4 August 2014.

Figure options

The observed clouds could be single-layer clouds or multi-layer clouds. The occurrence frequency of cloud at a specific layer is defined as the ratio between the number of radar beams with clouds being observed at this layer and the total number of radar beams with cloud detection at any layers. The time interval for the calculation is 60 min. The occurrence frequency for total amount of clouds, single-layer clouds, doublelayer clouds, triple-layer clouds, and multi-layer (equal to or larger than four) clouds is shown in Fig. 5. It shows that the occurrence frequency for total amount of clouds is generally larger than 60%, and demonstrates a distinct diurnal variation. The occurrence frequency for total amount of clouds is the smallest at 1200 BT (about 10:30 am in local time), and increases rapidly with the intensified surface heating. It reaches the maximum value of 0.9 at 2300 BT. Single-layer clouds account for about 50% of the total amount of clouds, and also demonstrate a distinct diurnal variation. With the increase in the cloud layers, the percentage that they account for the total amount of clouds decreases. Wang et al. (2011) analyzed the Cloudsat measurements and found that the total amount of clouds in July is about 80% over the entire TP, among which 55% of clouds are single-layer clouds. This result is consistent with that of the present study.

Fig. 5 Diurnal variations of cloud occurrence frequency for total amounts of clouds, single layer clouds (curve A), double-layer clouds (curve B), triple-layer clouds (curve C), and multiple layer clouds (curve D; ≥ 4 layers).

Figure options

In the International Satellite Cloud Climatology Project (ISCCP), the values of cloud top pressure and optical thickness are used to classify different cloud types, i.e., cumulus, stratocumulus, stratus, altocumulus, altostratus, nimbostratus, cirrus, cirrostrarus, and deep convective clouds (Rossow and Schiffer, 1999). In order to better underst and the macroscopic features of clouds over the TP and their difference to that in other regions, different types of clouds in Naqu region are selected for further analysis with a focus on the vertical structure of reflectivity, vertical velocity, and cloud-particle phase.

Figure 6 shows the time-height cross-sections of reflectivity, radial velocity (positive upward), velocity spectrum width, and depolarization factor L_DR for the newly formed cumulus clouds. The vertical y-axis indicates height, the x-axis indicates time, and the origin indicates the radar antenna (4560 m AGL, the same hereafter). It shows that cumulus cloud height is about 3 km, and the cloud depth is 2 km. These cumulus clouds are the low-level clouds shown in Fig. 2. They passed the radar station in a very short time, indicating that the clouds have a small horizontal scale. Looking at the radial velocity, no precipitation formed in the two clouds with the maximum reflectivity of -30 dBZ, while the radar measured radial velocity are mostly upward with a speed larger than 5 m s⁻¹, indicating that the clouds are at their developing stage. Based on the statistical analysis of clouds in the TP,1400 BT corresponds to the time when cumulus clouds can develop rapidly. The cloud height and depth both grow quickly at this time. The radial velocity is negative in the other two clouds whose reflectivity reaches about -15 dBZ, suggesting that precipitate particles has formed in these clouds.

Fig. 6 Cloud radar observations of cumulus clouds during 1315-1428 BT 15 July 2014. (a) Reflectivity factor, (b) radial velocity, (c) velocity spectrum width, and (d) depolarization ratio.

Figure options

Figure 7 shows the case for altocumulus clouds that formed in early morning coexist with cumulus clouds and well-developed deep convection. The deep convective clouds reach up to 12 km with a maximum echo intensity of 15 dBZ (because of the attenuation of cloud, real reflectivity is larger than this value). Distinct upward draft is found in the upper part of the clouds, and a significant bright b and (characterized by abrupt increases in L_DR and sudden changes in reflectivity and radial velocity) occurs at 1.5 km. The base of the altocumulus clouds reaches up to 6 km while the top is above 10 km. Meanwhile, cumulus clouds are developing at the level of around 2.5 km above the ground with a depth less than 1 km. The cumulus cloud top is above the zero-temperature level, suggesting that ice cloud is the major component of the cumulus clouds. The above two clouds are altocumulus and low cumulus clouds shown in Fig. 2, respectively. Large L_DR is located at 3- and 8-km heights, respectively, corresponding to central parts of the deep convection and indicating that mixed phase clouds might exist at these levels.

Fig. 7 As in Fig. 6, but for echo features of altocumulus clouds, cumulus clouds, and deep convection measured by cloud radar during 0222-0449 BT 18 July 2014.

Figure options

Convection in the TP can reach very high levels but the convective intensity is relatively weak. There are many cases showing the coexistence of altocumulus clouds and deep convection. The altocumulus cloud top is often consistent with deep convective cloud top, possibly because the altocumulus clouds are generated at the dissipative stage of deep convective cloud. In addition, the cases we discuss in the following paragraph clearly show that the bright b and is quite distinct even during the process of significant convective development, which is quite different from what happens in the low altitude regions. One possible reason is that the zero-temperature level is closer to the ground in the TP than in low altitude regions, resulting in a relatively weak upward motion in this level.

Figure 8 illustrates the case when altostratus and stratus clouds coexist in the morning and convective development is weak. The base and top of the altostratus clouds are 6 and 11 km, respectively. The reflectivity is about -15 dBZ. The base of the stratus clouds is at around 2.5-km height, and the cloud distribution is horizontally homogeneous and very shallow. The reflectivity of the stratus clouds is about -35 dBZ, while upward motion is found in the clouds. This is a period when convection is the weakest, and the atmospheric stratification is stable. However, upward motion still exists in the stratus clouds.

Fig. 8 As in Fig. 6, but for the echo features of altostratus and stratocumulus observed by cloud radar during 0546-0700 BT 12 July 2014.

Figure options

Figure 9 is the time-height cross-sections of nimbostratus clouds in the morning. The cloud top is horizontally homogeneous and reaches up to 8 km. Reflectivity can be up to 10 dBZ and demonstrates spatial-temporal variation. Two strong echoes pass the cloud radar at 73 min. The bright b and is quite distinct, and the reflectivity increases by 15 dBZ after the melting of ice particles. Radial velocity changes by 6 m s⁻¹. The falling speed of liquid and solid precipitate particles can be up to 6 m s⁻¹ if the upward motion nearby the zero-temperature level is ignored. The figure of the radial velocity shows that positive radial velocity (3 m s⁻¹) exists at 3-km height, indicating a significant updraft at this level. The height of this updraft is consistent with that occurred in the cumulus in Figs. 6 and 7 and in the newly developed stratus clouds shown in Fig. 8.

Fig. 9 As in Fig. 6, but for the echo features of nimbostratus clouds observed by cloud radar during 0806-0919 BT 6 July 2014.

Figure options

To compare with nimbostratus clouds,Fig. 10 presents a deep convection case. The strong convective echo passed Naqu weather station at about 30 min, and the echo height reached up to 16.5 km. Looking at the radial velocity, it is found that upward motion largely occurred above 3 km, where the maximum radial velocity is 6 m s⁻¹. The upward velocity can be larger if considering the falling speed of precipitate particles. At 1805 BT, negative radial velocity occurred at the weak echo area at the levels between 2 and 4 km above the ground, where the radial velocity can be up to -8 m s⁻¹. The curved echo shape indicates that this is an area of inflow, where the air aloft subsides and continues to sink in the clouds. This feature suggests that updrafts and downdrafts occur simultaneously during deep convective process. Both large L_DR (> −24 dB) and small L_DR (< −28 dB) can be found in the convective updraft zone, implying that mixed phase and supercooled liquid water are present in deep convective clouds. Similar to the cumulus clouds shown in Fig. 7, even under such a strong convective condition, we can found the bright b and clearly. Apparently, vertical motion in this layer is not significant based on the consideration of radial velocity and bright band.

Fig. 10 As in Fig. 6, but for the deep convective case that occurred during 1725-1839 BT 5 July 2014.

Figure options

The third Tibetan Plateau Atmospheric Science Experiment was carried out from 1 July to 31 August 2014. Comprehensive measurements of water vapor, clouds, and precipitation were conducted at Naqu. The advanced radars in China, such as Kab and millimeter-wave cloud radar, Ku-b and microrain radar, C-b and continuous-wave radar and lidars, and microwave radiometer and disdrometer were deployed to observe high spatial-temporal resolution of vertical structures for clouds and precipitation. The C-b and dual-linear polarization radar was coordinated with the new generation weather radar to constitute a dual-Doppler radar system for the measurements of three-dimensional wind fields and the hydrometeor distributions within precipitations. Based on the radar measurements in this experiment, we analyzed the statistical features of clouds in the summer over Naqu region, and revealed the macro features of different types of clouds. The major conclusions are as follows.

(1) The cloud properties have been successfully measured by using various ground-based radars in the field experiment conducted in the summer of 2014. In particular, information about the vertical structure and evolution of clouds obtained in this experiment provides a strong basis for further studies in cloud physics and precipitation process.

(2) During the summertime over Naqu, clouds are largely distributed at levels above 6 km and below 4 km. Few clouds form at around 5 km. Statistical analysis showed that total amounts of clouds, the top of high clouds, and cloud depth, all demonstrate a distinct diurnal variation. Few clouds form at 1000 LST, whereas the strong surface heating after the noon effectively promotes the development of convection. There is distinct diurnal variation in the top of low and mid-level clouds. The highest frequency of cloud formation is found at 6-9-km levels during 1800-0400 BT.

(3) Newly formed cumulus and stratus clouds are often found at 3-km height, where there often exist significant updrafts. Various types of clouds and clouds at different levels of height often coexist during this period. Altostratus and altocumulus clouds are probably related to the dissipating process of the deep convection.

(4) Analysis of the observed deep convection cases indicates that updrafts and downdrafts often exist simultaneously in the convective system. Supercooled water and mixed phase might exist in such kinds of deep convective system.

The above measurements and preliminary analysis provide a basis for further study of clouds and precipitation in the TP. These observations are also valuable in modeling studies of cloud and precipitation physics as well as in the development of parameterization schemes in numerical prediction models.

Acknowledgments. We appreciate the contribution made by Meteorological Bureau of Tibet Autonomous Region, Naqu Bureau of Meteorology, the 23rd Research Institute of China Aerospace & Industry Corp., and Anhui Sun-create Electronics Limited Company. We also thank Professor Zhao Ping and Dr. Gao Wenhua for their suggestions and comments.