Classification and Diurnal Variations of Precipitation Echoes Observed by a C-band Vertically-Pointing Radar in Central Tibetan Plateau during TIPEX-III 2014-IOP

Ruoyun MA, Yali LUO,2*, and Hui WANG,3

1 State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, Beijing 100081

2 Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters,Nanjing University of Information Science &Technology, Nanjing 210044

3 Key Laboratory for Cloud Physics and Weather Modification of China Meteorological Administration,Chinese Academy of Meteorological Sciences, Beijing 100081

ABSTRACT This study investigates classification and diurnal variations of the precipitation echoes over the central Tibetan Plateau based on the observations collected from a C-band vertically-pointing frequency-modulated continuous-wave(C-FMCW) radar during the Third Tibetan Plateau Atmospheric Scientific Experiment (TIPEX-III) 2014-Intensive Observation Period (2014-IOP). The results show that 51.32% of the vertical profiles have valid echoes with reflectivity > -10 dBZ, and 35.06% of the valid echo profiles produce precipitation at the ground (precipitation profiles); stratiform precipitation with an evident bright-band signature, weak convective precipitation, and strong convective precipitation account for 52.03%, 42.98%, and 4.99% of the precipitation profiles, respectively. About 59.84% of the precipitation occurs in the afternoon to midnight, while 40.16% of the precipitation with weaker intensity is observed in the nocturnal hours and in the morning. Diurnal variation of occurrence frequency of precipitation shows a major peak during 2100-2200 LST (local solar time) with 59.02% being the stratiform precipitation; the secondary peak appears during 1300-1400 LST with 59.71% being the weak convective precipitation; the strong convective precipitation occurs mostly (81.83%) in the afternoon and evening with two peaks over 1200-1300 and 1700-1800 LST, respectively. Starting from approximately 1100 LST, precipitation echoes develop with enhanced vertical air motion, elevated echo top, and increasing radar reflectivity. Intense upward air motion occurs most frequently in 1700-1800 LST with a secondary peak in 1100-1400 LST, while the tops of precipitation echoes and intense upward air motion reach their highest levels during 1600-1800 LST. The atmospheric conditions in the early morning are disadvantageous for convective initiation and development. Around noon, the convective available potential energy (CAPE) increases markedly, convective inhibition (CIN) is generally small, and a super-dry-adiabatic layer is present near the surface (0-400 m). In the early evening, some larger values of CAPE, level of neutral buoyancy, and total precipitable water are present, suggesting more favorable thermodynamic and water vapor conditions.

Key words: TIPEX-III, vertically-pointing radar, precipitation echoes, diurnal variation

1. Introduction

Clouds and precipitation over the Tibetan Plateau (TP)significantly influence the atmospheric circulations surrounding the TP (Flohn, 1957) and affect formation and variability of the Asian monsoon (Liu et al., 2012; Wu et al., 2012), mainly through modulating the interactions among surface sensible heating, vertical distribution of latent heating, and large-scale dynamical circulations(Luo and Yanai, 1984; Hsu and Liu, 2003). It has long been noticed that the clouds and precipitation over the TP in boreal summer are primarily convective in nature.Flohn (1968) documented a high frequency of afternoon cumulonimbus activity that continues into the night in eastern TP during boreal summer based on satellite images. The studies on ground-based radar observations during the GEWEX Asian Monsoon Experiment in Tibet(GAME-Tibet; Koike et al., 1999) have shown that the summer clouds over central TP are mostly convective,with substantial diurnal variations (Qian et al., 1984;Uyeda et al., 2001; Bhatt and Nakamura, 2005).

Characteristics of convective clouds over the TP have been identified based on observations from the spaceborne sensors aboard theCloudSat(Stephens et al.,2002), CALIPSO (Winker et al., 2003), and Tropical Rainfall Measuring Mission (TRMM; Simpson et al.,1988). Using theCloudSat/CALIPSO observations, Luo et al. (2011) found that, compared to the southern slope of the TP and the southern Asian monsoon region, deep convection over the TP is weaker in convective intensity and embedded in smaller-size convection systems owing to a lower level of neutral buoyancy (LNB) and a much drier atmosphere. By analyzing the TRMM observations,Qie et al. (2014) found that deep convective systems over the TP are relatively weak in convective intensity and small in size, confirming the findings of Luo et al.(2011); Xu (2013) concluded that convective storms over the TP are smaller horizontally and weaker in convective intensity than over low elevations in the East Asian monsoon region but stronger than deep convection over the slopes and foothills of the eastern plateau. Xu (2013) further pointed out that deep convective clouds over the plateau region have higher cloud bases and shallower mixed-phase depths than over low elevations in the East Asian monsoon region, but display active mixed-phase microphysical processes (strong radar echoes in the mixed-phase region).

However, classification of precipitation over the TP remains unclear because of insufficient observations and the difficulty in accurately detecting the presence of bright band by both ground-based volumetric scanning radars and space-borne radars due to the close proximity of the 0°C level to the highlands’ surface. The spaceborne radars observe from the top of the atmosphere downward and their detection capability near the ground[about 1 km above ground (AG)] is reduced greatly due to the noise caused by ground clutter, resulting in the misidentification of ground echoes as bright band in the precipitation classification algorithm of TRMM (Awaka et al., 1998). On the other hand, the lowest elevation angle of the ground-based volumetric scanning radars is normally 0.5°, which leads to deficient observations of the precipitation echoes near the ground. Furthermore,the relatively coarse spatial resolutions also contribute to their limited ability to reveal the vertical structure of the precipitating clouds. For example, Uyeda et al. (2001)showed vertical profiles of reflectivity for stratiform and convective echoes, respectively, observed by a groundbased X-band radar at Naqu in central TP during the GAME-Tibet. Their profiles clearly indicate a puzzling lack of bright band (Austin and Bemis, 1950) [a signature of “stratiform” precipitation (Houze, 1993)] in the stratiform echoes despite a strong contrast between the convective and stratiform echoes (see Fig. 8 therein).Based on the Precipitation Radar (PR) observations aboard the TRMM satellite, Fu and Liu (2007) surprisingly found a high percentage of stratiform rain (80% in version 5 and 93% in version 6) and a low percentage of convective rain (6% in both versions 5 and 6) over central TP in summer.

During the field campaign of the Third Tibetan Plateau Atmospheric Science Experiment (TIPEX-III; Zhao et al., 2018), a C-band vertically pointing radar was placed at Naqu in central TP during the 2014-Intensive Observing Period (IOP) from 1 July to 31 August (see Fig. 1 for a map of the study area with the locations of the observing facilities of relevance to the present study).It is a C-band frequency modulated continuous wave (CFMCW) Doppler radar (Peters et al., 2002) that measures vertical profiles of reflectivity (Ze), radial velocity(Vr), and velocity spectrum width (SW) of echoes overhead. TheVrand SW measurements offer quantitative indication of vertical velocity and turbulence intensity in convective precipitation. The C-FMCW radar has a 5.53-GHz frequency (about 5-cm wavelength), a 30-m range resolution, a 3-s temporal resolution, and a nominal sensitivity of -10 dBZ at 15 km overhead. The PR on board TRMM, by contrast, has a 13.8-GHz frequency (2.2-cm wavelength), a field-of-view diameter of about 5.0 km(after the boost in August 2001) at the nadir, a 250-m range resolution, and a nominal sensitivity of approximately 17 dBZ (Simpson et al., 1996). As previously stated, the bright band is located close to the TP’s surface,making it difficult to correctly identify its existence from the measurements by both space-borne radar (due to the adverse effect of ground clutter) and ground-based volumetric scanning radar (due to the limitations of the lowest elevation of about 0.5° and the coarser vertical resolution). Therefore, the vertically pointing observing mode and high resolutions of the C-FMCW radar are of particular importance for observing the bright band in stratiform precipitation regions for the first time over the TP.

Fig. 1. Topography (shadings) of the study area, and locations of the C-FMCW radar (blue dot) and the Naqu radiosonde station (red triangle).

The present study uses the C-FMCW radar measurements during the TIPEX-III to gain further insight into classification of precipitating clouds and their diurnal variations over central TP. Section 2 introduces the methodology to identify precipitation profiles and separate them into three types: stratiform, strong convective, and weak convective. The results of precipitation classification and diurnal variations are presented in Sections 3 and 4, respectively. Statistics of environmental conditions are described in Section 5. The paper ends with summary and conclusions, including a conceptual model of the classified precipitation types, in Section 6.

2. Precipitation classification method

The C-FMCW radar operated nearly continuously from 1 July to 31 August 2014 with measurements available during 96.6% of the IOP by average (Fig. 2), although its operation was unstable from 1-4 July and occasionally in the later time of the IOP. It measured 1,338,622 valid vertical profiles ofZe,Vr, and SW, respectively, during the IOP. In the present study, these 3-s profiles are running-averaged in a 1-min time window and the averaged individual profiles are used to identify the echo types by using the procedure illustrated in Fig.3. A vertical profile is identified as a valid-echo profile if it contains radar echo of > -10 dBZ in at least a 0.5-km depth anywhere from 0.12 to 15 km AG; otherwise, it is a clear-sky profile. The value of -10 dBZ is used since it is the minimum value of valid radar reflectivity that the C-FMCW radar can detect (Ruan et al., 2015). If a validecho profile contains vertically continuous radar echo (>-10 dBZ) from 0.12 to 0.5 km AG, it is believed that precipitation is produced at the surface and the profile is therefore classified as a precipitation profile. The precipitation profiles are further classified into three categories according to the vertical distribution of radar reflectivity:stratiform, strong convective, and weak convective. A precipitation profile is classified as stratiform if it contains a bright band and does not contain a thick continuous layer (> 1-km depth) of large reflectivity (> 35 dBZ).The precipitation profiles that are not classified as stratiform and have maximum reflectivity exceeding 35 dBZ are classified as strong convective. All the remaining profiles are classified as weak convective.

The presence of a bright band in a precipitation profile is determined by objectively judging bright-band boundaries following Fabry and Zawadzki (1995), combined with visual examination. The height of the brightband top (htop) is defined as the height where the curvature in logZeis maximum. A similar criterion is used to locate the bright-band bottom (hbot). Ifhtopandhbotare not identified, it is believed that the vertical profile does not contain a bright band. Afterhtopandhbothave been determined by the computer, time-height distributions of radar reflectivity are visually examined to make sure thathtopandhbotare indeed at the top and bottom of the bright band and the layer of large reflectivity(> 35 dBZ), if any, is not thicker than 1 km. Under occasional circumstances whenhtopandhbotof a certain pro-file are located evidently higher or lower than those at adjacent times, the abnormalhtopandhbotvalues are replaced by the representative values at adjacent times.

Fig. 2. Time fraction (%) of daily available observations of the C-FMCW radar during 1 July-31 August 2014.

Fig. 3. Flow chart of the precipitation classification procedure.

3. Precipitation classification results

The vertical profiles observed by the C-FMCW radar during the entire IOP are classified according to the algorithm described above. Statistics from the results are shown in Table 1. A total of 686,947 valid-echo profiles are identified, accounting for 51.32% of all profiles, i.e.,the remaining 48.68% are clear-sky profiles. Among the valid-echo profiles, 240,842 profiles are identified as precipitation profiles, accounting for 17.99% and 35.06% of all profiles and valid-echo profiles, respectively. That is to say, slightly more than 1/3 of the valid-echo profiles produce precipitation at the ground. The number of stratiform precipitation profiles with an evident bright-band signature is 125,301, which occurs at a frequency (divided by the number of all profiles) of 9.36% during the IOP and accounts for 52.03% of the total precipitation profiles. This fraction of stratiform precipitation is smaller than that (67%) obtained over North Mexico during the North American Monsoon Experiment (NAME) by Williams et al. (2007), who also used well-defined bright bands in the reflectivity profiles obtained from a groundbased profiling radar as a signature of stratiform rain in their rain-classification methodology. It is further revealed in the present study that the bright band centers at 0.97 km AG with an average depth of 420 m, i.e., boundaries of the bright band (htopandhbot) on average are located at 1.18 and 0.76 km AG, respectively.

The occurrence frequency of strong convective precipitation profiles (total number of 12,019) is 0.90%, which is comparable to that of deep convective core over the TP(1.19%) obtained by Luo et al. (2011) based on theCloudSat/CALIPSO measurements, and is also comparable with those of deep convective cores over subtropical North America and East Asia (0.92%-1.06%), but smaller than those over the south slope of the TP (1.88%)and southern Asian monsoon region (1.92%). The strong convective precipitation profiles account for merely 4.99% of the total precipitation profiles, indicating that only a small portion of precipitation over central TP contains large reflectivity (> 35 dBZ). A total number of 103,522 profiles are identified as weak convective precipitation, which occur at a frequency of 7.73% and account for 42.98% of the total precipitation profiles.

Table 1. Number and occurrence frequency (%) of vertical profiles, fraction of precipitation profiles in the valid-echo profiles, and fractions of three different precipitation types (i.e., stratiform, strong convective, and weak convective) in the precipitation profiles, respectively

The results suggest that slightly more than half of the precipitation (52.03%) is stratiform with a distinct brightband signature, followed by the weak convective precipitation (42.98%), whilst the fraction of strong convective precipitation is much less (4.99%). It is worth noting that the presence of a bright band and the absence of a continuous layer with depth > 1 km of large reflectivity (>35 dBZ) are required in identifying the stratiform precipitation profiles in the present study, which is stricter than those adopted in the previous studies (e.g., according to the TRMM PR precipitation classification algorithm, a vertical profile is identified as stratiform as long as a bright band is found near the melting layer). Therefore, it is unlikely that the occurrence frequency of stratiform precipitation is overestimated in the present study. The relatively high fractions of the stratiform precipitation and the weak convective precipitation in the current study suggest that the strong convective precipitation over the TP tends to weaken in convective intensity shortly after its genesis and turn into the stratiform and weak convective precipitation types. This is further confirmed by examination of the time-height distributions of radar reflectivity and radial velocity observed by the CFMCW radar during the entire IOP. It is believed that the short duration of strong convective activities is caused mainly by the characteristic atmospheric conditions over the TP, namely, low water vapor content and small convective available potential energy (CAPE; see Section 5 and Luo et al., 2011).

To better illustrate the ability of the C-FMCW radar in revealing the internal structures of the precipitation systems and the rationality of our classification results, the time-height distributions of radar reflectivity and radial velocity of a representative precipitation event occurring during 1400-2300 LST 19 August 2014 is shown in Fig. 4.Weak convective precipitation echoes are observed during 1513-1626 LST, with transient strong convective precipitation echoes appearing from 1520 to 1535 LST.In the early evening, the C-FMCW radar observes the most intense convection overhead during the IOP. The extreme reflectivity (> 45 dBZ) extends up to 4.5 km AG, large reflectivity (35-40 dBZ) to 11 km AG, and precipitation echo top reaches about 13 km AG during 1650-1700 LST. In the following 10 minutes, the extreme reflectivity extends from near the ground to 5 km AG, but the top of large reflectivity decreases to 5-7 km AG despite the existence of intense updrafts (Fig. 4b)near the echo top until about 1800 LST. The corresponding radial velocities from 1640 to 1800 LST suggest strong upward and downward motion (maximal speed up to 16.09 m s-1) from 2 km AG to the echo top. Afterwards, the convective intensity weakens significantly and most echoes are classified as weak convective after about 1810 LST. Subsequently, a bright band starts to appear,but disappears at 1848 LST; it reappears at about 2010 LST and remains quite stable until 2230 LST, with weak convective precipitation being occasionally identified amid the stratiform precipitation.

Fig. 4. Time-height distributions of (a) reflectivity (Ze) and (b) radial velocity (Vr) observed by the C-FMCW radar during 1400-2300 LST 19 August 2014. The red, black, and light blue colors on the top of the figure illustrate the classified stratiform, strong convective, and weak convective precipitation types, respectively. The black solid and dashed lines, respectively, denote the heights of the 0 and -15°C isotherms estimated using the sounding observations on 19 August 2014.

4. Diurnal variation of the precipitation echoes

4.1 Echo top height

The echo top height (ETH) herein refers to the maximum height reached by the minimum effective reflectivity (-10 dBZ) observed by the C-FMCW radar. The median and average values of the ETH during the entire IOP are 6.21 and 6.06 km AG, respectively. Figure 5 shows the diurnal variation of ETH. The echo top rises gradually after 1000 LST and remains stable at approximately 8.00 km AG during 1400-2100 LST. The highest ETH appears during 1700-1800 LST, showing a concentrated distribution with a median value of 8.25 km AG and the 25th percentile value reaching the median ETH of the whole day (as indicated by the red dashed line in Fig. 5).The ETH descends almost continuously from 0000 to 1100 LST, and remains relatively low during 0600-1000 LST with a median of about 4.00 km AG, which is 2.21 km lower than the median ETH throughout the whole day. The median ETH hits the lowest level around 1000-1100 LST. These features are consistent with the results of Chang and Guo (2016), who examined the diurnal variations of ETH by analyzing observations from the C-FMCW radar and a cloud radar during the IOP.

Fig. 5. Diurnal variation of echo top height (ETH) observed by the CFMCW radar. The red dashed line denotes the median ETH during the entire IOP. In each box-and-whisker diagram, the two shorter transverse lines outside of the box are the 90th (upper) and 10th (lower)percentiles, two transverse lines enclosing the top and bottom of the box are the 75th and 25th percentiles, and the transverse line inside the box is the 50th percentile. The 50th percentile values at each hour are connected with the black solid line.

4.2 Strong echoes

The average occurrence frequency of strong echoes with radar reflectivity ≥ 30 dBZ (35 dBZ) during the IOP, defined as the number of profiles containing the strong echoes divided by the total number of profiles, is 6.58% (3.71%). The diurnal variation of the occurrence frequency of strong echoes (Fig. 6) shows a major peak during 2000-2200 LST, being more pronounced in the diurnal variation pattern of reflectivity ≥ 30 dBZ. The occurrence frequencies are also relatively high during 0100-0300, 0500-0600, 1200-1500, and 1700-1800 LST,while relatively low in the morning (0600-1200 LST),with 1000-1100 LST being the lowest.

Fig. 6. Diurnal variation of the occurrence frequencies of radar reflectivity ≥ 30 dBZ (blue) and ≥ 35 dBZ (red). The dashed lines in blue and red represent the corresponding average occurrence frequencies during the IOP.

Fig. 7. As in Fig. 5, but for the top heights of radar reflectivity of(a) 30 and (b) 35 dBZ.

Figure 7 demonstrates diurnal variations of the maximum heights reached by the 30- and 35-dBZ echoes(ETH_30dBZ, ETH_35dBZ), respectively. The median(mean) values of ETH_30dBZ and ETH_35dBZ during the IOP are 1.24 (1.67) and 1.13 (1.40) km AG, respectively. Tops of the strong echoes are generally lower dur-ing 0000-1000 LST. The top heights of strong echoes start to rise at about 1000 LST and remain relatively high until 2000 LST with two evident peaks: one around 1100-1300 LST with the highest median values and the other during 1600-1800 LST with decreased medians but distinctively higher 90th percentiles, suggesting that the strong echoes can extend to the highest altitudes in the late afternoon/early evening. It is worth pointing out that the strong echoes have low tops during 2000-0600 LST(Fig. 7) despite their relatively high occurrence frequency (Fig. 6), reflective of the strong echoes in the profiles of stratiform precipitation rather than vigorous,vertically-erect convective precipitation systems.

4.3 Intense upward air motion

The radial velocity Vrmeasured by the C-FMCW radar reflects the sum of the vertical air motion and the fall speed of cloud/precipitation particles. An upward vertical motion of air is bound to exist when negative Vrvalues appear. Therefore, negative Vrwith larger absolute values can predominantly represent the stronger upward air motion. Although this representation underestimates the actual values of upward air velocity, especially at the lower levels where the fall speeds of large solid precipitation particles and liquid rain drops are larger than those of solid cloud/precipitation particles at the upper levels, direct measurement of air vertical velocity is a world-wide challenge and the representation adopted herein allows us to analyze qualitatively the upward air motion at the higher levels. The Vrvalues of ≤ -3 and ≤ -5 m s-1are used here to represent intense upward air motion. Their average occurrence frequencies during the IOP are 0.61% and 0.23%, respectively.

The diurnal variation of occurrence frequency of the intense upward air motion (Fig. 8) presents three peaks during 0900-2400 LST, with the major peak over 1700-1800 LST, the secondary one over 1100-1400 LST, and the third one over 1900-2200 LST. The intense upward air motion is hardly observed during 0000-0900 LST.Figure 9 shows diurnal variations of top heights of the intense upward air motion (the results during 0000-0900 LST are not shown owing to the rare occurrence of intense upward air motion during this period). The tops of intense upward air motion are apparently higher during 1600-2300 LST than during 1000-1600 LST, with a major peak appearing over 1700-1800 LST and a secondary peak over 2100-2200 LST, respectively.

4.4 Vertical distributions of the valid echo and radial velocity

Fig. 8. As in Fig. 6, but for the occurrence frequencies of radial velocity ≤ -3 m s-1 (blue) and ≤ -5 m s-1 (red).

Fig. 9. As in Fig. 5, but for the top heights of radial velocity of(a) -3 and (b) -5 m s-1.

Figure 10 shows vertical distribution of the number of valid-echo in each hour of the day. The echoes occur mostly below 12 km AG, with a minimum number around 4-5 km AG and a maximum near 3 km AG during most times of the day. The echo number near 6-9 km AG starts to increase around noon, apparently in close association with the development of strong upward air motion in the precipitation systems (Figs. 8, 9), leading to the formation of a “double peak” structure in the vertical distribution, which is more evident during 1400-2100 LST. The echo number around 6-9 km AG is even larger than that near 3 km AG during 1700-1900 LST.The vertical distributions and diurnal variations of the echo number described herein are qualitatively consistent with those documented by Liu et al. (2015) and Chang and Guo (2016).

Fig. 10. Vertical distributions of the number of valid-echo (> -10 dBZ) in each hour of the day.

Fig. 11. Vertical profiles of radar reflectivity in each hour of the day. Profiles in dark blue, light blue, black, orange, and red represent the 10th,25th, 50th, 75th, and 90th percentiles of radar reflectivity, respectively.

Figure 11 shows vertical profiles of the 10th, 25th,50th, 75th, and 90th percentiles of radar reflectivity in each hour of the day. Profiles of all percentile values above 8 km AG are concentrated at the low end of reflectivity (-10 to 5 dBZ) during 2300-1000 LST. Starting from about 1100 LST, the echoes develop with larger reflectivity and elevated tops (up to about 12 km AG). Of note is the presence of large reflectivity values within 11-12 km AG during 1600-1800 LST, reflective of strong upward air motion supporting large precipita-tion particles at these high altitudes (Fig. 9). Further comparisons of the 50th and 90th percentiles in each 3-h period of the day are performed (Fig. 12). For both the 50th and 90th percentiles, the profiles from 1 km AG upward during 0000-0900 LST are located to the left of those during 0900-2400 LST, which again suggests that the precipitation echoes during the nocturnal to early morning hours are generally weaker. For the 50th percentile profiles, the reflectivity within 4-8 km AG increases rapidly with decreasing altitude due to the growth of ice-phase particles by aggregation when falling downward and the larger fall speed of larger particles. An obvious curve to the right near 4 km AG is observed at most hours of the day except for 1200-1500 LST when the reflectivity maximum is located within 2-3 km AG,whereas a local reflectivity minimum is present around 3 km AG. Further, the 90th percentile profiles show that,across approximately 1 km AG where the 0°C level is located, the reflectivity values first increase and then decrease with decreasing altitude, reflective of the “bright band” structure of the stratiform precipitation.

Figure 13 shows vertical profiles of the selected percentiles of radar radial velocity in each hour of the day. It is suggested that, except that the radial velocities of the 10th and 25th percentiles are negative (upward velocities) at high levels (above 8 km AG), the radial velocities of all percentiles in the whole layer are positive (downward velocities) throughout the day. Large downward radial velocity is present within 1-2 km AG due to the strong drag effect of large solid precipitation particles and liquid-phase rain drops. The radial velocity is observed increasing during 1000-2300 LST as a result of enhanced drag effect of precipitation particles due to increased occurrence frequency of precipitation (Fig. 14a).It is noteworthy that more intense upward air motion exists within 8-13 km AG during 1600-1800 LST, which might have lifted the precipitation particles to higher levels, in accordance with the maximum top height of strong echoes in the early evening (Fig. 7). The stronger upward air motion is closely related to the positive buoyancy resulting from latent heat release by the ice-phase microphysical processes, and the upward motion near the echo top is also associated with the enhanced turbulent movement inside the clouds arising from the longwave radiative cooling (Luo et al., 2008). These analysis results from vertical profiles of radar reflectivity and radial velocity are not only consistent with the results from the ETH, strong echoes, and intense upward air motion in the previous subsections, but also adding more details concerning the vertical structure of the precipitation echoes.

4.5 Classified precipitation profiles

Fig. 12. Vertical profiles of the (a) 50th and (b) 90th percentiles of radar reflectivity in each 3 h of the day. The colors denote different times of the day.

Fig. 13. As in Fig. 11, but for vertical profiles of radial velocity. The dashed grey line in each panel indicates radial velocity of 0 m s-1.

Fig. 14. As in Fig. 6, but for the occurrence frequencies of (a) total precipitation, and (b) stratiform (blue), strong convective (red), and weak convective precipitation (black).

As described in Section 2, the precipitation profiles refer to the vertical profiles that contain continuous radar echo from 0.12 to 0.5 km AG. Therefore, the precipitation profiles are profiles that produce precipitation at the ground. The average occurrence frequency of the precipitation profiles during the IOP is 17.99%. Figure 14a shows the diurnal variation of occurrence frequency of total precipitation profiles. The occurrence frequency during 2300-1200 LST is lower than the daily average occurrence frequency (the blue dashed line in Fig. 14a)with the minimum value (11.21%) appearing during 0800-0900 LST; the occurrence frequency during 1200-2300 LST is higher than the daily average occurrence frequency, with the main peak during 2100-2200 LST(28.19%) and the secondary peak during 1300-1400 LST(23.68%).

Based on the classification results of the precipitation profiles, diurnal variations of occurrence frequencies of the classified stratiform precipitation, strong convective precipitation, and weak convective precipitation have been analyzed (Fig.14b). The stratiform precipitation occurs most frequently between 2000 and 2400 LST with a peak during 2100-2200 LST (16.64%); the occurrence frequency is also relatively high during 0000-0800 LST,but distinctly low (about 4.03%) between 1000 and 1300 LST with a minimum during 1100-1200 LST. The strong convective precipitation mainly occurs (81.83%) between 1100 and 2300 LST, with the major peak over 1700-1800 LST (3.47%) and the secondary peak over 1200-1300 LST (3.30%). Diurnal variation of the occurrence frequency of weak convective precipitation exhibits an evident peak during 1200-1500 LST (about 13%).

The major peak of precipitation occurrence frequency during 2100-2200 LST is contributed significantly(59.02%) by the stratiform precipitation, and the secondary peak during 1300-1400 LST significantly by the weak convective precipitation (59.71%). Due to its rare occurrence, the strong convective precipitation contributes less to the total precipitation in terms of occurrence frequency. Comparison of the occurrence frequency of the classified precipitation types suggests that the strati-form precipitation has the highest occurrence frequency among the three precipitation types, except for during 1000-1600 LST when the weak convective precipitation occurs most frequently among the three precipitation types. During 1200-1300 LST, the occurrence frequency of weak convective precipitation reaches its peak; in the meantime, that of strong convective precipitation reaches its secondary peak (slightly lower than its major peak during 1700-1800 LST), leading to the largest frequency of (strong plus weak) convective precipitation of the day.This is attributed to the increased atmospheric instability due to solar heating and surface sensible heat flux in the morning hours (Section 5). Moreover, the occurrence frequency of weak convective precipitation around noon is much higher than that of strong convective precipitation(the former is approximately 5 times the latter), suggesting that despite the highest occurrence frequency of convective precipitation around noon, the majority is of weak intensity. In contrast, during 1700-1800 LST, the occurrence frequency of weak convective precipitation declines considerably while that of strong convective precipitation reaches its own peak, leading to the comparable occurrence frequencies between the two precipitation types. This indicates that nearly half of the total convective precipitation is of the strong convective type in the early evening, consistent with the favorable atmospheric conditions for stronger convective development at this time of the day (Section 5).

Statistically, the ETH of each precipitation type indicates that the median top heights of the stratiform, strong convective, and weak convective precipitation are 7.20,9.51, and 5.67 km AG, while the corresponding averages are 6.94, 9.04, and 5.94 km AG, respectively. This suggests that, in terms of the average or median ETHs, the strong convective precipitation is the highest, followed by the stratiform precipitation, and the weak convective precipitation is the lowest. Figure 15 shows the diurnal variation of ETH of each precipitation type. The ETHs of each precipitation type during the second half of the day(1200-2400 LST) are generally higher than their correspondences during the first half of the day (0000-1200 LST). The stratiform precipitation’s ETH peaks over 1400-1500 LST and remains near 8 km AG during 1500-2300 LST, while the minimum appears over 1000-1100 LST with a median value of 4.05 km AG. The amplitude of the diurnal variation of the ETH of strong convective precipitation is larger than those of the other two precipitation types (Fig. 15b). Despite the limited sample size of strong convective precipitation during the first half of the day, it is clear that the ETHs in this period are significantly lower (by about 4-5 km) than the median value during the whole day (as indicated by the red dashed line in Fig. 15). The minimum (4.14 km AG) appears during 0700-0800 LST with a median less than half of the median during the whole day (9.51 km AG). These indicate that the strong convective precipitation occurs less frequently during the first half of the day with weaker convective intensity. The ETHs of strong convective precipitation rise rapidly around 1100 LST, and the median reaches about 11-12 km AG during 1200-1300 LST.During 1600-2200 LST, the ETHs remain at a high level(11-12 km AG). The ETHs peak over 1700-1800 LST with a median of 11.91 km AG, which indicates that the strong convective precipitation reaches its peak intensity in the early evening when it also occurs most frequently(Fig. 14b). The diurnal variation of weak convective precipitation (Fig. 15c) is characterized by higher (lower)ETHs during the second (first) half of the day with medians being 7-9 (4-6) km AG.

Fig. 15. As in Fig. 5, but for the ETHs of the (a) stratiform, (b) strong convective, and (c) weak convective precipitation.

5. Diurnal variation of the atmospheric conditions

During the TIPEX-III 2014-IOP, 9 extra soundings at 1200 LST were launched in July at the Naqu meteorological station in addition to the twice-daily routine sounding observations at 0600 and 1800 LST, which provide valuable information on the atmospheric conditions around noon over the central TP. In total, there are 62, 9,and 59 observations at 0600, 1200, and 1800 LST, respectively, during the IOP. These soundings are used to analyze the diurnal variation of the atmospheric condi-tions in order to better understand the diurnal variations of the classified precipitation types.

First, composite vertical profiles of the temperature and dew point temperature at 0600, 1200, and 1800 LST,respectively, are depicted in the skew T-lgp diagram. The corresponding lifting paths of the 0-500 m AG mixing air parcel are also given in the diagram (Fig. 16). In view of the fact that the near-surface air in 15 soundings at 1800 LST is contaminated by the convectively-generated cold pool in the afternoon, the 44 intact observations and 15 contaminated observations at 1800 LST are composited separately. Six variables that are closely associated with the initiation and development of convection are calculated for each composite profile (Table 2),including CAPE, convective inhibition (CIN), lifting condensation level (LCL), level of free convection(LFC), LNB, and total precipitable water (TPW). At 0600 LST (Fig. 16a), the temperature of near-surface air(0-1.5 km AG) is low mainly due to the nocturnal longwave radiative cooling effect, and the relative humidity is high. The air parcel lifting curve is on the left side of the environmental air temperature curve in the whole layer(from ground to about 12 km AG) and CAPE is 0 J kg-1,suggesting quite disadvantageous conditions for convective initiation and development.

At 1200 LST (Fig. 16b), the near-surface air temperature increases remarkably owing to continuous solar heating after sunrise and subsequent warming through surface sensible heat flux (Xu et al., 2002). The increase of air temperature below 0.4 km AG is particularly substantial, leading to a temperature lapse rate greater than the dry adiabatic lapse rate, i.e., a super-dry-adiabatic layer is present. In a thin layer (about 250-m depth) above the super-dry-adiabatic layer, the lapse rate is close to the dry adiabatic lapse rate. The dew point depression below 1.5 km AG is relatively large, indicating dry near-surface air;however, the dew point depression decreases (i.e., relative humidity increases) evidently with increasing altitude, indicating the existence of humidity inversion as also noted by Xu and Chen (2006). An evident dry layer is observed at 4-5 km AG due to the rapid decrease in the dew point temperature, which has an adverse effect on the formation and maintenance of precipitation particles. This is in accord with the low occurrence frequency of precipitation echoes at 4-5 km AG (Fig. 10).The 1200 LST composite profile has the highest CAPE(370.64 J kg-1) among three times, its CIN is almost 0 J kg-1, LCL is 1.26 km AG, LFC is 1.36 km AG, and LNB is 9.90 km AG, which demonstrates that the atmospheric conditions around noon are conducive to the convective initiation and development.

Fig. 16. Composite profiles of air temperature (red), dew point temperature (blue), and air parcel lifting curve (green) based on the Naqu sounding observations during the TIPEX-III 2014-IOP. (a) 62 observations at 0600 LST, (b) 9 observations at 1200 LST, (c) 44 observations at 1800 LST that are not affected by rain evaporative cooling, and (d) 15 observations at 1800 LST that are affected by rain evaporative cooling.

Table 2. Six convection-related variables derived from the composite sounding profiles in Fig. 16

For the composite profiles of the unaffected soundings at 1800 LST (Fig. 16c), the air temperature over 0-500 m AG declines slightly (by about 0.73°C) compared to 1200 LST, and the humidity inversion phenomenon below 1.5 km AG remains. Of importance is that the CAPE decreases, CIN increases, LCL rises to 1.36 km AG, and LFC rises to 1.99 km AG, suggesting that the near-surface air needs to overcome larger CIN and reach higher LFC before ascending freely with the support of positive buoyancy. This is consistent with the finding that the convective precipitation occurs less frequently in the early evening than around noon (Fig. 14b).For the composite profiles of the contaminated soundings at 1800 LST, the near-surface air temperature declines noticeably, CAPE is merely 5.71 J kg-1, LNB is only 6.04 km AG, CIN is 53.24 J kg-1, LCL is only 0.77 km AG, and LFC is 2.35 km AG.

The CAPE, CIN, LCL, LFC, LNB, and TPW are also calculated based on each of the radiosonde observations at Naqu station during the IOP, and the statistical distribution of these variables at 0600, 1200, and 1800 LST,respectively, are demonstrated in Fig. 17. The TPW is low (< 20 mm) at three times over Naqu and exhibits a small magnitude of diurnal variation, while the other 5 parameters exhibit distinct diurnal variations. At 0600 LST, CAPE is generally very small with a median (maximum) of only 4.48 (448.38) J kg-1; LNB is significantly low with a median value of only 5.45 km AG; CIN is large with the 75th percentile value reaching 51.89 J kg-1, which could be attributed to the nocturnal longwave radiative cooling and the evaporative cooling of precipitation; LCL distributes in a narrow range and is significantly lower than those at 1200 and 1800 LST due to the generally lower temperature and higher relative humidity near the surface in the early morning; and the median value of LFC is also relatively low.

At 1200 LST, CAPE increases significantly with a median of 457.60 J kg-1(exceeding the maximum CAPE at 0600 LST) and a maximum of 759.92 J kg-1. LNB increases significantly too and distributes in a narrow range with a median of 10.02 km AG. CIN decreases dramatically with a median value of only 5.45 J kg-1and a 75th percentile value of 17.23 J kg-1. The LCL distribution is still relatively concentrated, but the values are significantly higher than those at 0600 LST. The LFC distribution is broader, with values also higher than those at 0600 LST.

The median value of CAPE at 1800 LST reduces to 109.96 J kg-1, being significantly lower than that at 1200 LST. However, the 75th percentile value is in proximity to the median value at 1200 LST, and the maximum value (973.84 J kg-1) is even larger than that at 1200 LST(759.92 J kg-1). The CIN distribution at 1800 LST extends to larger values compared with noon, due mainly to the impact of precipitation evaporative cooling in the afternoon. The distribution of LCL is wider with smaller values appearing compared with 1200 LST, indicating that the near-surface air in the early evening is sometimes more likely to reach its LCL. The median LNB decreases compared to 1200 LST, but the 75th, 90th percentiles and maximum are equal to or greater than their corresponding values at 1200 LST, which indicates that higher altitudes of the LNB are likely to appear in the early evening.

The above results are more clearly illustrated in the two-dimensional distribution of normalized occurrence frequency of the environmental air parameters (Fig. 18).In the early morning, the CAPE values are concentrated in the small values (mostly < 200 J kg-1), the LNB distributes between 3 and 7 km AG, and CIN ranges from 0 to 150 J kg-1. At noon, the normalized occurrence frequency in each bin is higher due to the smaller sample size (nine observations). The CAPE scatters between 0 and 800 J kg-1and the smaller (larger) CAPE values correspond to larger (smaller) LFC values. Most of the CIN values are < 10 J kg-1, and the LNBs are mainly distributed at 9-11 km AG. Distribution of the variables at 1800 LST is obviously wider than those at 0600 LST. Of importance is that larger CAPEs (800-1000 J kg-1) appear compared to 1200 LST, and these larger CAPEs correspond to lower LCLs, lower LFCs, higher LNB, and smaller CINs. Such configuration of parameters in the early evening is more conducive to the development of stronger convection. These diurnal variations of the atmospheric parameters are overall consistent, and partially explain those of the classified precipitation types.

Fig. 17. Box-and-whisker diagrams of (a) CAPE, (b) CIN, (c) LCL, (d) LFC, (e) LNB, and (f) TPW, calculated by using the Naqu sounding observations during the TIPEX-III 2014-IOP at 0600 LST (black), 1200 LST (red), and 1800 LST (blue), respectively. The two dots outside of the box denote the maximum and minimum values, respectively; the two shorter transverse lines outside of the box are the 90th and 10th percentiles;the two transverse lines enclosing the box are the 75th and 25th percentiles; and the transverse line inside the box is the 50th percentile.

6. Summary and conclusions

In the present study, the vertically-pointing C-FMCW radar observations during the 2014-IOP of the TIPEX-III are used to examine the classification and diurnal variations of the vertical profiles of precipitation echoes in the central Tibetan Plateau during boreal summer (July-August). The sounding observations at Naqu radiosonde station (including 9 extra launches at 1200 LST) are used to analyze the diurnal variation of atmospheric conditions. The major conclusions are as follows.

(1) Of all the vertical profiles observed by the C-FMCW radar during the IOP, the valid-echo and clear-sky profiles account for 51.32% and 48.68%, respectively.Precipitation profiles with non-zero surface precipitation account for 17.99% of the total profiles and 35.06% of the valid-echo profiles, respectively. Stratiform precipitation with a bright band signature, which is precisely observed for the first time over the TP, accounts for 9.36%and 52.03% of the total profiles and the precipitation profiles, respectively. The occurrence frequency of strong convective precipitation with maximal reflectivity exceeding 35 dBZ is 0.90% and accounts for only 4.99% of total precipitation profiles; while that of weak convective precipitation is 7.73% and accounts for 42.98% of the precipitation profiles. Strong convective precipitation tends to weaken rapidly in convective intensity and transform into the stratiform and weak convective precipitation types.

(2) Approximately 59.84% of the precipitation with a higher echo-top height and more intense upward air motion occurs in the afternoon-to-midnight hours, while the remaining 40.16% occurs in the nocturnal and morning hours. Structural features of the classified precipitation types differ distinctively from each other, and differences are also noticed in the same precipitation type between the two sub-periods of the day. Consequently, a conceptual model of the vertical structures of the classified stratiform, strong convective, and weak convective precipitation types during 0000-1200 and 1200-2400 LST, respectively, is summarized (Fig. 19). The precipitation types in the nocturnal-to-morning hours tend to have smaller horizontal spans and lower ETHs relative to their counterparts in the afternoon-to-midnight hours.The precipitation, especially the stratiform precipitation,in central TP presents unique vertical structures in comparison with the precipitation over low elevations, e.g.,those shown in the conceptual models of squall lines(Houze et al., 1989).

Fig. 18. Normalized occurrence frequency of CAPE and (a-c) LCL, (d-f) LFC, (g-i) CIN, (j-l) LNB at (a, d, g, j) 0600 LST, (b, e, h, k) 1200 LST, and (c, f, i, l) 1800 LST, based on the Naqu sounding observations during the TIPEX-III 2014-IOP. Normalization refers to the division of sample number in every bin by the total sample number at the corresponding observation time.

(3) Precipitation echoes occur least frequently at about 4-5 km AG and most frequently near 3 and 6-9 km AG,respectively. Starting from 1100 LST, precipitation echoes develop with increasing radar reflectivity, enhanced vertical air motion, and elevated echo top. Intense upward air motion (speed > 5 m s-1) occurs most frequently at 1700-1800 LST with a secondary peak during 1100-1400 LST, while the tops of precipitation echoes and intense upward air motion reach their highest levels during 1600-1800 LST.

(4) At 0600 LST, the vast majority of CAPE values are < 200 J kg-1, the LNB is mostly 3-7 km AG, and the CIN ranges from 0 to 150 J kg-1, indicating generally unfavorable conditions for the convective initiation and development in the early morning. Around noon, the CAPE increases significantly, while CIN is mostly < 20 J kg-1,and a super-adiabatic layer is present near the surface(0-400 m AG), suggesting that convection can be easily initiated. At 1800 LST, the average thermodynamic conditions are not as advantageous as those around noon, but some larger values of CAPE, LNB, and TPW appear,suggesting that more favorable thermal and water vapor conditions can appear in the early evening.

Fig. 19. Schematics of the classified (a) strong convective, (b) weak convective, and (c) stratiform precipitation during two sub-periods of the day, i.e., 0000-1200 and 1200-2400 LST. Shadings in light blue, dark blue, green, yellow, and red represent radar reflectivity of -10, 20, 25, 30,and 35 dBZ, respectively, with the corresponding median values of their top heights (Mean Sea Level: MSL) labeled. The median heights of the bright band top (htop) and bottom (hbot) for the stratiform precipitation are also labeled in (c). The percentages at the top of each panel indicate the occurrence frequency of the classified precipitation type (out of the parentheses) and the corresponding fraction in all precipitation profiles (in the parentheses).

In summary, analysis of the C-FMCW radar observations at Naqu during the 2014-IOP of TIPEX-III generally corroborates findings from previous studies about the diurnal variations of precipitation over the central TP.However, it is important to note that the C-FMCW radar,with its unique capability to identify the bright band in stratiform precipitation over the TP with very high resolutions in the vertical and temporal, presents a new opportunity to better classify the precipitation, and the findings herein add to our knowledge of the precipitation and associated convective activity in this region.