Observed Vertical Structure of Precipitation over the Southeastern Tibetan Plateau in Summer 2021

Precipitation plays a key role in the earth’s climate system. Understanding the microphysical and structural characteristics of precipitation is important for precipitation prediction and radar-based quantitative precipitation estimation (QPE) (Zhang et al., 2001; Gilmore et al., 2004; Lam et al., 2015). Because raindrops are affected by microphysical processes during their descent, such as collision, breakup, and evaporation, the vertical distribution characteristics of precipitation are usually heterogeneous (Peters et al., 2005; Zhao and Garrett, 2008; Das and Maitra, 2016; Song et al., 2019).

Radar is an important instrument to monitor the three-dimensional structure and evolution of precipitation. However, due to the curvature of the earth, as the weather radar range increases, so does the height of the precipitation observations (Peters et al., 2005). Details of the precipitation vertical structure are critical for accurate radar-based QPE (Das and Maitra, 2016). Surface radar reflectivity factor could be corrected by the vertical profile of reflectivity (VPR) to reduce the bias of radar-based QPE (Qi et al., 2013). Wind profile radar (WPR) and micro rain radar (MRR) can obtain vertical profiles of raindrop size distribution (DSD) and integral rain parameters [e.g., rainfall rate (R), liquid water content (LWC), and radar reflectivity (Z)] from the power spectral density of the descent rates of raindrops at different elevations (Peters et al., 2005; He et al., 2013; He et al., 2014; Das and Maitra, 2016; Song et al., 2019).

With the advantages of small size, low cost, and high precipitation sensitivity, the vertically pointing K-band MRR has become an important instrument to study the vertical evolution of precipitation and has been widely used in comprehensive rainfall observations (Peters et al., 2005; Das et al., 2010; Wen et al., 2015; Das and Maitra, 2016; Song et al., 2019; Ramadhan et al., 2020). The measurements collected from MRRs located around the Baltic Sea showed a strong height dependence of DSD characteristics for high rainfall rates, and that weather radar tends to underestimate the high rainfall rates when using the ground-based Z–R relations (Peters et al., 2005). Convective clouds and precipitation exhibited more inhomogeneous vertical microstructures than stratiform clouds and precipitation (Das et al., 2010; Ma et al., 2018; Li D. J. et al., 2022). Das and Maitra (2016) explored the features of vertical precipitation structure at three tropical locations in India by using MRR observations and reported that the mean DSD near ground is dominated by smaller raindrops during light rainfall, but obvious increase in raindrop size in heavy rainfall. The diurnal variation in the vertical profile of DSD associated with stratiform precipitation at Kototabang, West Sumatra, was explored based on MRR measurements and showed that nocturnal stratiform precipitation is characterized by a higher concentration of mid- and large-sized raindrops (Ramadhan et al., 2020). Studies on the vertical precipitation structure using MRR data have been conducted in East and North China, and indicated that evaporation (coalescence) is evident during the fallout of raindrops for the weak (heavy) precipitation (Wen et al., 2015; Song et al., 2019).

The Tibetan Plateau (TP), known as the roof of the world and the third pole due to its being the highest plateau in the world, is also referred to as the Asian water tower as it is the source of seven important rivers in Asia, such as the Yangtze River, the Yellow River, the Yarlung Zangbo River (Xu et al., 2008), and so on. Clouds and precipitation over the TP are important components of global hydrological cycles and energy budgets, and play a significant role in atmospheric circulation and climate change (Xu et al., 2008; Kang et al., 2010; Zhao et al., 2016; Li, 2018). Recently, more attentions have also been paid to the aerosol characteristics and their impacts on the weather and climate system over the TP (Zhao et al., 2020). Observations of the microphysical characteristics of clouds and precipitation over the TP generally originate from satellite data (Yue et al., 2019; Lin et al., 2022). In order to improve understanding of the physical characteristics of cloud and precipitation over the TP, ground-based radar observations were carried out during the periods of three Tibetan Plateau Atmospheric Scientific Experiments (TIPEXs). Particularly, Ka-band cloud radar, X-band dual polarization radar, C-band frequency-modulated continuous-wave (C-FMCW), and MRR were deployed in Naqu during the third TIPEX (Liu et al., 2002, 2015; Chang and Guo, 2016; Zhao et al., 2016; Ma et al., 2018).

The Yarlung Zangbo Grand Canyon (YZGC), located in the southeastern TP, is a vital conduit for the water vapor transfer from the Indian Ocean to the TP (Gao et al., 1985; Gao, 2008). This area has attracted the attention of meteorological scientists, and most of related studies have focused on hydroclimatic characteristics (Gao et al., 1985; Zhang et al., 2016; Liu et al., 2018; Sun et al., 2020; Liu et al., 2022). For example, Gao et al. (1985) proposed that the rainfall totals of automatic weather stations in the YZGC primarily depend on the amount of water vapor from upstream. Due to the extensive influence of the topography and climate of the YZGC, the spatial distribution of extreme precipitation indicators in the YZGC indicates that rainstorms may occur more frequently in the eastern humid areas and flood disasters may be more serious (Liu et al., 2018). The rainfall totals decrease with increasing elevation in the Yarlung Zangbo River basin due to the height dependent reduction of total column water vapor (Sun et al., 2020). However, due to the lack of in situ observations, few studies have been conducted on the vertical structures and microphysical characteristics of clouds and precipitation.

The region of interaction between the westerlies and monsoons encompasses the TP and surrounding areas. Climate warming leads to westerly and monsoon anomalies and an imbalance of the Asian water tower. Mêdog, located at the entrance to the water vapor conduit of the YZGC, is the key valley region for the Asian water tower. Cloud and precipitation variability in Mêdog is vital to the investigation of multiscale water vapor transfer and the mechanisms that influence water resources. However, observations of clouds and precipitation in Mêdog have been impeded by the complex topography, transportation difficulties, and especially, frequent landslides and debris flows during the rainy season. To explore the reasons and mechanisms related to water resource changes in the YZGC pertaining to the interaction between westerlies and monsoons, a comprehensive cloud and precipitation observation platform was established in 2019 at the Mêdog National Climate Observatory (MNCO) [29.31°N/95.32°E, 1305 m above sea level (ASL)]. The field observation campaign was supported by the Second Tibetan Plateau Scientific Expedition and Research (STEP) Program and the “Earth-Atmosphere Interaction in the TP and Its Influence on the Weather and Climate in Lower Reaches of the Yangtze River” project.

The advanced X-band dual polarization phased array radar, Ka-band cloud radar, K-band MRR, particle size and velocity (PARSIVEL) disdrometer, and microwave radiometer were deployed at the MNCO to understand the vertical structure and microphysical processes of clouds and precipitation at the entrance of the water vapor conduit of the YZGC. Based on the observations of these remote instruments, some macro- and microphysical characteristics of clouds and precipitation were studied. Wang et al. (2021, 2022) proposed that the DSD characteristics during the summer monsoon season of this area were characterized by abundant small drops, and convective rain should be classified as maritime-like. The vertical structure and diurnal variation of clouds in Mêdog were examined by Zhou et al. (2021), who reported that clouds tended to form frequently at night and dissipate gradually in the daytime. However, the rain vertical profiles in this area have not been explored by in situ measurements.

Based on the measurements collected by the collocated MRR, PARSIVEL disdrometer, and rain gauge from June to September 2021 at the MNCO, the objectives of this study are to (i) evaluate the accuracy of MRR observation data at the MNCO by the intercomparison of measurements of MRR, PARSIVEL disdrometer, and rain gauge data; and (ii) explore the details of rain vertical structure in Mêdog, and if any distinct discrepancies in different rain rate categories are identified, those will be compared to observations from other regions.

The location of the MNCO, topography of the TP, and the MRR is shown in Fig. 1. Mêdog, located at the front of the water vapor conduit, has a mean elevation of 1200 m (ASL) and a subtropical climate. Under the interaction of westerlies and monsoons, a large amount of water vapor ascends from the southern foothills of the Himalayas to Mêdog and is then transferred throughout the TP. Therefore, Mêdog has become the key “entrance” of the water vapor transfer conduit in the southeastern TP. Due to the abundant warm and humid air flows brought by the Indian Ocean monsoon, the average annual rainfall in Mêdog usually exceeds 2000 mm (Chen and Li, 2018). In particular, rainfall during the summer monsoon period from June to September accounts for about 60% of the annual total rainfall (Chen and Li, 2018; Li R. et al., 2022; Wang et al., 2022).

The instruments and methodology adopted in this study are described in Section 2. The evaluation of the MRR measurements is described in Section 3. The details of the vertical rain structure at the entrance of the water vapor conduit of YZGC are given in Section 4. Conclusions are summarized in the final section.

Figure 1.  (a) The locations of the Mêdog National Climate Observatory (MNCO; black solid dot), Yarlung Zangbo Grand Canyon (YZGC; black rectangle), and topography (m; shaded) of the Tibetan Plateau (TP) and (b) the micro rain radar (MRR) at the MNCO.

Vertical rain profile data were collected from a vertically pointing MRR, which is a frequency modulated continuous wave (FMCW) Doppler radar and operates at 24.1 GHz. The MRR was produced by Meteorologische mess Technik GmbH (METEK) company, Germany, and uses common transmitting and receiving antennas due to the low transmitter power (e.g., 50 mW), which eliminates problems of the beam overlap (Peters et al., 2005). The performance parameters of the MRR at the MNCO are given in Table 1. In this study, the temporal resolution of the MRR measurements was 1 min, and the range resolution was set to 30 m.

The MRR measures the fall speed of raindrops based on the Doppler principle and receives raw spectral power. The techniques for the derivation of drop size distributions, including converting raw spectra into calibrated reflectivity spectra and attenuation correction, were described in detail by Peters and Fischer (Peters and Fischer, 2002). Briefly, the raindrop concentration $ N(D,h) $ (m⁻³ mm⁻¹) for a drop diameter D (mm) at a height h can be related to the reflectivity spectra $ \eta (D,h) $:

where $ \sigma (D) $ represents the backscattering cross section of a single hydrometer; $ \eta (v,h) $, which is directly obtained from the Doppler spectra measured by MRR, can be converted to $ \eta (D,h) $ by using the empirical relationship of terminal fall speed v and raindrop diameter D (Atlas et al., 1973):

where $ v(D,h) $ (m s⁻¹) is dependent on the height due to air density variation. The $ {\delta _\rho }(h) $ is the approximative height correction factor and is given by Foote and Du Toit (Foote and Du Toit, 1969):

The derivative of Eq. (2) with respect to D is:

and

Therefore, $ N(D,h) $ can be obtained by substituting Eqs. (4) and (5) into Eq. (1):

where $ \eta (v,h) $ can be obtained from the Doppler spectra of MRR, and $ \sigma (D) $ is calculated from Mie scattering. Once $ N(D,h) $ is known, the rainfall integral parameters, such as rainfall rate (R), liquid water content (LWC), and radar reflectivity (Z) at height h, can be calculated by using the following equations:

where ${\rho _{\rm{w}}}$ is the density of water.

The mean fall speed (${v_{\rm{m}}}$) at height h can be calculated directly from $ \eta (v,h) $:

The mass-weighted mean diameter D_m (mm) and the normalized intercept parameter N_w (m⁻³ mm⁻¹) can be calculated as follows:

where M_n is the nth-order moment of DSD. It is defined as:

The range gate and gate resolutions of the MRR at the MNCO are optional and were set to 128 and 30 m, respectively. This configuration could effectively monitor the region during the summer monsoon season because the 0°C isotherm is approximately 4 km AGL in Mêdog during summer monsoon period (Wang et al., 2022). Due to noise and ground clutter issues (Peters et al., 2005; Wen G. et al., 2017), the first two range gates were excluded. The MRR observations, at a 1-min temporal resolution, were collected from the summer monsoon season of 2021 (1 June–30 September).

The collocated PARSIVEL disdrometer and tipping-bucket rain gauge observations during June–September 2021 were also collected for the comparison and evaluation of MRR measurements. The PARSIVEL disdrometer has been widely used to conduct DSD studies over various regions (Tokay and Short, 1996; Bringi et al., 2003; Lam et al., 2015; Chen et al., 2017; Ji et al., 2019). The PARSIVEL disdrometer at the MNCO, constructed by the China Huayun (Beijing) Meteorological Technology Co., Ltd., is similar to that produced by the OTT company in Germany. The sampling area is 54 cm² (18 cm × 3 cm), and the sampling interval is 1 min. Measured hydrometer particles are divided into 32 diameter bins and 32 terminal velocity bins, ranging from 0.062 to 24.5 mm and from 0.05 to 20.8 m s⁻¹, respectively. The quality control procedures used for the disdrometer to minimize the measurement errors are not described in this work because the details can be found in many previous studies (Yuter et al., 2006; Friedrich et al., 2013; Tokay et al., 2013; Wang et al., 2022). Wang et al. (2022) reported that the PARSIVEL disdrometer at the MNCO has a high degree of agreement with the collocated rain gauge as evidenced by both the 5-min rainfall trends and the scatterplot of hourly cumulative rainfall, although it somewhat underestimates the gauge rainfall. The tipping-bucket rain gauge, at a 0.1-mm sensitivity and a 1-min sampling interval, is typically used as the benchmark of surface rainfall.

To explore the characteristics of the precipitation profiles in Mêdog, it is necessary to assess the accuracy of the MRR measurements. Therefore, intercomparisons were performed between the collocated MRR, PARSIVEL disdrometer, and rain gauge measurements collected at the MNCO. First, two rain events are analyzed in detail to demonstrate the accuracy of the MRR measurements: one stratiform event that occurred on 8 June 2021, and a mixed rain event observed on 29 September 2021. Then, statistical results from the complete datasets during the summer monsoon period of 2021 are discussed.

An example of a stratiform rain event that occurred on 8 June 2021, is given in Fig. 2. This event was characterized by a reflectivity factor mostly within 10–35 dBZ and a mean fall speed not exceeding 10 m s⁻¹. A clear bright band (BB) can be seen above a height of 3.5 km. Below the BB, the radar reflectivity exhibited some uniformity, although there was a minor vertical tilt. The mean fall speed at BB shows a small value of less than 4 m s⁻¹ and then increases sharply to a maximum value of 10 m s⁻¹ below the melting level. This implies that the conversion of ice particles into rain increased the mean fall speed. However, as the height decreased, the larger mean fall speeds (e.g., v_m > 7 m s⁻¹) also decreased while corresponding reflectivity factors increased, which might be associated with the breakups of larger drops and the increase of raindrop concentration during their descent. The slower fall speeds (e.g., v_m < 4 m s⁻¹) increased as height decreased, this probably indicates the coalescence and/or evaporation of small drops during their descent.

Figure 2.  An example of a stratiform rain event observed by MRR at the MNCO between 1700 and 2300 Beijing Time (BT, BT = UTC + 8 h) on 8 June 2021. (a) Reflectivity factor and (b) mean fall speed.

The intercomparison between the MRR, PARSIVEL disdrometer, and rain gauge observations for stratiform rain is shown in Fig. 3. The time series of DSD from the PARSIVEL disdrometer at a 1-min sampling interval can be used to examine those from the MRR. This study evaluated the DSD and rainfall rates from the MRR at heights of 90 m (MRR_90m) AGL and 180 m (MRR_180m) AGL. It can be seen that the diameter of drops does not exceed 3 mm, and the DSD change trends with time exhibit good agreements among the DSD values obtained from MRR_90m, MRR_180m, and the PARSIVEL disdrometer (Figs. 3a–c). However, it is also noted that the MRR overestimated the concentration of smaller drops (e.g., D < 1 mm) and underestimated the concentration of larger drops (e.g., D > 2 mm) compared to the PARSIVEL disdrometer. This is also evidenced by the mean DSD (Fig. 3d). However, the PARSIVEL disdrometer tends to underestimate the concentration of small drops due to the one-dimensional laser signal measurements with the “one drop at a time” assumption (Yuter et al., 2006). With the exception that the concentration of particles less than 0.5 mm was occasionally underestimated, the DSD values from MRR_180m were very similar to those of MRR_90m. This indicates that the DSD at different heights might be nearly uniform during the stratiform rainfall event.

Figure 3.  Intercomparison among the MRR, PARSIVEL disdrometer, and rain gauge for a stratiform rain event observed on 8 June 2021. Time series of DSD from (a) MRR_90m, (b) MRR_180m, and (c) PARSIVEL disdrometer. (d) Comparison of mean DSD from the MRR with that from the PARSIVEL disdrometer. Intercomparison of rainfall trends among the MRR, PARSIVEL disdrometer, and rain gauge: (e) 5-min rain rate and (f) hourly rainfall.

Regarding the 5-min mean rainfall rate, the correlation between the MRR (including MRR_90m and MRR_180m) and the disdrometer is better than that between the MRR and the rain gauge. This is evidenced by correlation coefficients (CCs) exceeding 0.95 between the MRR and disdrometer, though between the MRR and the rain gauge the coefficient was approximately 0.8 (cf. Fig. 3e). Significantly, the trend of the 5-min mean rainfall rate from the rain gauge shows more frequent fluctuations than the rain rate determined from the MRR and disdrometer. This might be due to the coarse sensitivity of 0.1 mm for the tipping-bucket. A previous study reported that tipping-bucket rain gauges usually have significant errors if timescales are less than 10–15 min (Habib et al., 2001). Taking hourly rainfall into account, CCs exceeding 0.98 are evident among the three datasets, such as the general rainfall trends and timings of rainfall peaks (cf. Fig. 3f), indicating good agreement. However, MRR_90m, MRR_180m, and the PARSIVEL disdrometers all underestimated the gauge rainfall by 10.1%, 15.7%, and 25.8%, respectively. We can also see that both the 5-min rainfall rate and hourly rainfall of MRR_90m are closer to the rainfall rates recorded by the gauge than those of MRR_180m. Furthermore, the MRR tended to overestimate the PARSIVEL disdrometer rainfall rates.

Figure 4 shows an example of a mixed rainfall event that occurred on the early morning of 29 September 2021. The mixed rainfall event experienced convective, shallow, and stratiform rain. During the convective rain period of 0300–0340 BT, the radar reflectivity and mean fall speed exhibited a significant variation in vertical structure. For the strong convective cores, reflectivity exceeded 45 dBZ, and the maximum mean fall speed was up to 11 m s⁻¹. The layer exceeding 45 dBZ, at heights of 2.0–2.5 km AGL at 0340 BT, indicated the development of intense updrafts in the middle troposphere (Dolan et al., 2018). Shallow precipitation followed from 0410 to 0450 BT, which was characterized by a lower cloud-top height below the 0°C isotherm level, generally remaining below 3 km, and a weak radar reflectivity below 25 dBZ. It is evident that stratiform precipitation occurred during 0530–0630 BT, with a BB above 3.5 km AGL and a reflectivity factor no more than 35 dBZ.

Figure 4.  An example of a mixed rain event observed by MRR at the MNCO on the early morning of 29 September 2021. (a) Reflectivity factor and (b) mean fall speed.

For the mixed precipitation event, the intercomparison between the MRR, PARSIVEL disdrometer, and rain gauge observations is given in Fig. 5. It is evident that the DSD trends also exhibit good agreement among the three datasets, such as the diameter of drops up to 5 mm during the convective precipitation period, raindrop sizes no more than 1.5 mm in shallow rainfall, and generally below 3 mm for the stratiform precipitation type. Nevertheless, the MRR tended to overestimate the concentration of small drops (e.g., D < 1 mm) and underestimate the concentration of raindrops with sizes larger than 2 mm (cf. Figs. 5a–d). Compared to MRR_90m, MRR_180m shows a significant decrease in the concentration of smaller drops during the convective rainfall period, while remaining relatively consistent for the shallow and stratiform types. This indicates that the DSD during the convective precipitation period was characterized by a significant variance in vertical structure. In terms of rainfall rate, good agreements are evident between MRR_90m, the PARSIVEL disdrometer, and rain gauge, such as high CCs above 0.95, the rainfall trends, and times of rainfall peaks (Figs. 5e, f). However, MRR_90m (the PARSIVEL disdrometer) slightly overestimated (underestimated) the gauge rainfall by 6% (6%). In addition, MRR_180m significantly underestimated the gauge rainfall during the convective rainfall period, with a mean negative bias of approximately 30%.

Figure 5.  As in Fig. 3, but for a mixed rain event observed on the early morning of 29 September 2021.

The monsoon precipitation in Mêdog contributes approximately 60% to the annual rainfall totals due to abundant warm and humid air flow from the Indian Ocean during this period (Chen and Li, 2018; Li R. et al., 2022; Wang et al., 2022). Therefore, the study focused on precipitation vertical structures in Mêdog during the summer monsoon season. Considering the MRR deployed at MNCO in July 2020, simultaneous observations of the MRR, PARSIVEL disdrometer, and rain gauge during the summer monsoon season (June–September) of 2021 were collected to statistically evaluate the accuracy of the MRR measurements at the MNCO. After eliminating the missing data due to a power failure caused by landslides and debris flows, the simultaneous 169,200 1-min samples from the three instruments were included in this study.

The rainfall rate measured from the MRR at a height of 90 m AGL was closer to that from the rain gauge based on the two precipitation events. In addition, the tipping-bucket rain gauge has reliable rainfall totals, although it suffers from significant errors at shorter timescales (e.g., less than 15 min) (Habib et al., 2001; Tokay and Bashor, 2010). Therefore, the hourly rainfall from the MRR at 90 m AGL was compared with that from the rain gauge and disdrometer.

In general, good agreements are noticed among the three instruments, evidenced by higher CCs, although the MRR slightly overestimated the gauge rainfall by 5.0% and the PARSIVEL disdrometer underestimated the gauge rainfall by 16.6% (Fig. 6). The MRR tended to overestimate weak rainfall (e.g., R < 5 mm h⁻¹) while underestimating heavy precipitation. The MRR shows a higher standard deviation (SD), indicating a larger dispersion in the MRR rainfall measurements. In contrast, the PARSIVEL disdrometer exhibits a smaller SD, although its underestimation of rainfall is fairly distinct. The distinct negative bias of the PARSIVEL disdrometer may be attributed to underestimating the concentration of the small and medium drops that are predominate in Mêdog during the summer monsoon period (Wen L. et al., 2017; Wang et al., 2022). Compared to the PARSIVEL disdrometer, the MRR overestimated rainfall by 25.9% but had a high CC of 0.942, indicating a general agreement between the two instruments.

Figure 6.  Intercomparison of hourly rainfall between the MRR, rain gauge, and PARSIVEL disdrometer during the summer monsoon season of 2021. The correlation coefficient (CC), bias, and standard deviation (SD) are also given.

The mean DSD from the MRR at 90 m at the MNCO during the summer monsoon season of 2021 was also compared to that from the PARSIVEL disdrometer (Fig. 7). The MRR records a significantly higher concentration of very small drops (e.g., D < 0.5 mm) as opposed to the PARSIVEL disdrometer, which tends to underestimate the concentration of small drops due to the assumption of “one drop at a time” (Wen L. et al., 2017). However, the PARSIVEL disdrometer can observe more large drops with D > 1.8 mm than the MRR. This might be attributed to the Mie scattering correction of the MRR spectral power, resulting in a decrease in the observation of larger raindrops (Löffler-Mang et al., 1999). The good agreement between the two instruments is evident for raindrops with 0.5 < D < 1.8 mm.

Figure 7.  Comparison of mean DSD from the MRR at 90 m AGL and that from the PARSIVEL disdrometer at the MNCO during the summer monsoon season of 2021.

The probability density of the normalized intercept parameter N_w, the mass-weighted mean diameter D_m, and the integral rain parameter LWC calculated from MRR measurements was also compared with those from the PARSIVEL disdrometer during the monsoon period of 2021 (Fig. 8). Taking D_m into account (Fig. 8a), the MRR has a peak occurrence frequency at 0.5 mm, while the PARSIVEL disdrometer exhibits a peak occurrence frequency of 0.9 mm. Meanwhile, the MRR probability density curve shows a larger (lower) probability for drops with D_m < 0.75 mm (D_m > 0.75 mm) than the PARSIVEL disdrometer. This result may be partly attributed to the fact that the PARSIVEL disdrometer tends to underestimate the concentration of small and medium drops and partly due to the MRR’s inherent limitation of the Mie scattering correction (Peters et al., 2005; Wen L. et al., 2017). As shown in Fig. 8b, the MRR peaks at a higher lgN_w value than the PARSIVEL disdrometer and has a relatively higher (lower) probability when the lgN_w value is larger (lower) than 3.9, indicating that the MRR can detect more drops. Lower LWC values below 0.1 g mm⁻³ are dominant. The LWC probability distribution of the MRR peaks at a slightly higher LWC value and has a higher probability than the PARSIVEL disdrometer when the LWC is larger than 0.1 g mm⁻³, which is probably due to the MRR detecting more drop counts (cf. Fig. 8c).

Figure 8.  Probability density distribution of (a) D_m (mm), (b) lgN_w, and (c) LWC (g m⁻³) computed from the MRR and PARSIVEL disdrometer during the summer monsoon season of 2021.

To better understand the vertical structures of precipitation parameters at the entrance of the water vapor conduit in the YZGC, the mean profiles of rainfall rate, LWC, mean fall speed, and radar reflectivity are discussed under four rain rate (R) categories: R1: 0.02 ≤ R < 0.2 mm h⁻¹, R2: 0.2 ≤ R < 2.0 mm h⁻¹, R3: 2.0 ≤ R < 20.0 mm h⁻¹, and R4: 20.0 ≤ R < 200.0 mm h⁻¹. The simple classification method is used because the ground rainfall rate is closely associated with rain types (cf. shallow, stratiform, and convective rain). The MRR rain rate at 90 m was used as a classification criterion due to its good consistency with the rain gauge and the nonuniformity of the rain rate vertical structure. The rain samples and accumulated rain amounts of the four categories based on the MRR observations at the MNCO during the summer monsoon season of 2021 are given in Table 2. Weak rainfall with rain rates between 0.2 and 2.0 mm h⁻¹ occurred frequently in Mêdog, whereas the third rain rate category (2.0 ≤ R < 20.0 mm h⁻¹) was the largest contributor to the rainfall amount.

The mean profiles of rain rate, LWC, radar reflectivity, and mean fall speed for different rain rate categories are shown in Fig. 9. Based on Ka band cloud radar and the ECMWF Reanalysis version 5 (ERA5) data, the mean height of the melting isothermal layer of the summer monsoon season in Mêdog is approximately 4.0 km AGL (Li R. et al., 2022; Wang et al., 2022; Zhang et al., 2022). To avoid the influence of ice particles, we focus on the rain profile analysis below 3.0 km AGL in this study. For the lightest rain rate category (R1), the profiles of the four parameters are almost uniform between 2 and 3 km AGL. However, the mean profiles of radar reflectivity, rain rate, and LWC exhibit a significant positive gradient below the height of 2 km AGL. Similar profiles can be found in different regions, such as Xingtai, North China (Song et al., 2019), Shillong, Northeast India (Das and Maitra, 2016), and Arctic (Zhao and Garrett, 2008). The lightest precipitation is usually related to drizzle or shallow rain. Considering the negligible rain attenuation in the lightest rainfall rate category, the behavior might be attributed to evaporation. In the case of very weak precipitation, the atmospheric environment is usually drier, and evaporation is increased. Higher evaporation rates reduce the concentration of raindrops reaching the ground, especially small raindrops. Evaporation usually occurs at lower heights, and hence, radar reflectivity, rain rate, and LWC exhibit positive gradients only near the ground. Furthermore, the apparent negative gradient is observed in the mean fall speed profile below a height of 2 km AGL, which indicates that the DSD peak tends toward larger raindrops at decreasing height. The reduction in small raindrops due to the enhancement of evaporation at lower heights may be responsible for the DSD peak shift.

The rain category R2 with rainfall rates between 0.2 and 2.0 mm h⁻¹ is generally associated with stratiform rain. The profiles of rain rate and LWC are almost uniform, and radar reflectivity exhibits an evident negative gradient with increasing height. Due to insensitivity to rain attenuation, the mean fall speed is useful to indicate the variation in raindrop size. The mean fall speed profile at the R2 category shows a slight negative slope below 1.5 km AGL, indicating that the DSD peak tends toward larger raindrops below the boundary layer. Therefore, the negative gradient in the radar reflectivity profile in this rainfall rate category is probably due in part to both rain attenuation and the shift of the DSD peak.

In rainfall rate category of R3, the profiles of rainfall rate, LWC, and radar reflectivity exhibit negative gradients. Furthermore, the magnitudes of the slopes of these profiles are apparently different. The rainfall rate profile shows a larger negative slope than the LWC profile, which could be attributed to the effect of air density (Peters et al., 2005). The radar reflectivity profile exhibits the maximum negative gradient, while the mean fall speed is almost independent of height in the rainfall rate category of R3. This may indicate that the breakups of large drops during their fallout increase the drop amounts. Rain attenuation is another reason for the significant negative gradient of radar reflectivity. Radar reflectivity decreases more than LWC if the raindrop size decreases with increasing height. Because radar reflectivity and LWC are proportional to the sixth and third power of raindrop size, respectively, radar reflectivity with a larger negative slope than LWC implies the coalescence of larger drops during their descent.

Radar reflectivity shows a steep negative gradient with height in the heaviest rainfall rate category, R4, than in the two medium rainfall rate categories (e.g., 0.2 < R < 20.0 mm h⁻¹). Precipitation with a rainfall rate above 20 mm h⁻¹ is usually associated with convective rain (Kirankumar and Kunhikrishnan, 2013; Das and Maitra, 2016). Such a steep slope of the radar reflectivity profile is probably attributed partly to the heavy rain attenuation at the MRR operational frequency and partly to the mixture of water and ice particles in the convective rain (Das and Maitra, 2016). The profiles of rainfall rate, LWC, and mean fall speed have a disturbance between 1.5 and 2.0 km AGL, which is not a general feature but may be the result of a single turbulent vertical wind profile. A local updraft reduces the fall speed, and the retrievals of LWC and rain rate are then positively biased. This explanation is supported by the observation that the relative anomaly of LWC is larger than that of the rainfall rate. This is expected if vertical wind is the cause. Upward wind causes a positive bias in the LWC. The rain flux is reduced by vertical wind, which mitigates the rain bias. The MNCO is located on the southern slope of the Himalayas and at the lower reaches of the Yarlung Zangbo River, with steep mountain and valley topography. The interaction of steep terrain and mesoscale convective systems tends to result in a local updraft at heights of 1–2 km (Barros et al., 2000).

The DSD plays an important role in understanding the fundamental microphysical rainfall characteristics (Rosenfeld and Ulbrich, 2003). Many ground-based DSD measurements have been analyzed to explore the variation due to different precipitation types or climate regimes (Zhang et al., 2003; Wen L. et al., 2017; Wang et al., 2022). However, the vertical structure of DSD could improve the understanding of the evolution of microphysical rainfall characteristics during drop descent. In addition, the vertical structure and evolution of DSD in the YZGC region are first explored in this study. According to previous studies, small (large) raindrops are defined as sizes below (above) 1 mm (3 mm), and raindrops with sizes of 1–3 mm are considered medium drops (Tokay et al., 2008; Rao et al., 2009; Krishna et al., 2016).

Figure 10 shows the mean DSD profiles for different rainfall rate categories. To avoid the influence of the melting layer, we also limit the region to below 3 km AGL for this discussion. The lightest rainfall rate category shows a narrow spectrum width with a raindrop size no more than 3 mm. The DSD vertical structure is almost uniform above 2 km AGL, below which the concentration of raindrops decreases with decreasing height, especially near the ground. This behavior corresponds to uniform profiles of rainfall rate, LWC, and radar reflectivity above the height of 2 km AGL and the positive vertical gradients below (cf. Fig. 9). As mentioned earlier, the reduction in the smaller raindrop concentration with decreasing height near the ground may be attributed to evaporation. Additionally, it can be seen that the concentration of larger raindrops decreases rapidly near the ground, which is probably due to the larger raindrop breakup process. It seems that the coalescence, breakup, and evaporation processes of raindrops reach near equilibrium at heights of 2–3 km AGL (Hu and Srivastava, 1995; Das and Maitra, 2016; Wu et al., 2019).

Figure 9.  Mean vertical profiles of (a) rain rate, (b) LWC, (c) radar reflectivity, and (d) mean fall velocity for different rain rate categories. R1: 0.02 ≤ R < 0.2 mm h⁻¹, R2: 0.2 ≤ R < 2.0 mm h⁻¹, R3: 2.0 ≤ R < 20.0 mm h⁻¹, and R4: 20.0 ≤ R < 200.0 mm h⁻¹.

Figure 10.  Mean vertical structures of DSDs for different rain rate categories during the summer monsoon season of 2021. (a) 0.02 ≤ R < 0.2 mm h⁻¹, (b) R2: 0.2 ≤ R < 2.0 mm h⁻¹, (c) R3: 2.0 ≤ R < 20.0 mm h⁻¹, and (d) R4: 20.0 ≤ R < 200.0 mm h⁻¹.

The R2 rainfall rate category usually represents stratiform rain. The small and medium raindrops show a uniform DSD profile. This suggests that the DSD may maintain a near equilibrium condition, in which the coalescence, breakup, and evaporation processes of raindrops are just in balance. However, the number of large raindrops decreases slightly with decreasing height, indicating the dominance of the large drop breakup process in their fallout, especially at lower heights. The decrease in large raindrop amounts without causing a noticeable increase in small drops indicates the role of evaporation.

For the rainfall rate category of 2–20 mm h⁻¹, the spectral width is broader than that of the two weak rainfall rate categories (e.g., R < 2.0 mm h⁻¹), and the largest raindrop size exceeds 4 mm near the ground. The concentration of small raindrops with sizes smaller than 0.7 mm is almost uniform, while the concentration of raindrops with sizes between 0.7 and 3.0 mm increases slightly as height decreases. This indicates that the droplet coalescence process is slightly dominant. Large raindrops are almost independent of height, suggesting a near steady state caused by a balance between the coalescence and breakup processes. The vertical structures are different from those in Xingtai and Shillong, in which coalescence prevails due to medium and large drops displaying a larger negative relationship with height (Das and Maitra, 2016; Song et al., 2019).

The R4 rainfall rate category represents convective rain and further broadens the spectral width with a large raindrop size up to 6 mm. All the raindrop concentrations rapidly increase with decreasing height, and the DSD shows more significant variation in height, especially for large raindrops. The collision and coalescence in the falling process of raindrops may be the reason for the increase in medium and large raindrops. During heavy precipitation, a sufficiently humid environment may be conducive to the coalescence process (Song et al., 2019). Breakups of large raindrops are probably responsible for the increase in smaller drop counts, especially near the ground (Morrison et al., 2012). The interpretation of the DSD vertical structure for this rainfall rate category is quite difficult due to the effect of strong rain attenuation and updraft in convective precipitation (Peters et al., 2005; Das and Maitra, 2016). For example, a local updraft could be responsible for the bulges of the DSD profile at heights of 1.5–2 km. Before heavy precipitation occurs in Mêdog, strong updrafts at heights of 1.5–2 km AGL are recorded by collocated wind profile radar. An updraft would shift spectral reflectivity to smaller fall speeds, resulting in an increase in the concentration of raindrops. The effect of mean vertical wind on the DSD and various integral parameters has been detailed by Peters et al. (2005). Better methods to improve the understanding of the DSD vertical structure in the case of this rainfall rate category should utilize other equipment, such as X-band dual polarization radar and wind profile radar.

In summary, the breakups of larger drops in Mêdog precipitation during their fallouts are more significant. This may be a major microphysical factor indicating why Mêdog precipitation was characterized by numerous size-limited raindrops (Wang et al., 2022). In addition, the mean values of fall speed and radar reflectivity of each rain rate category in Mêdog are lower than those in other regions (e.g., Xingtai and Shillong), which indicates the smaller mean drop size in Mêdog.

Mêdog, located in the lower reaches of the Yarlung Zangbo River, is the entrance of the water vapor conduit and the key region for water resources in Asian water towers. Based on cloud and precipitation observations made at the MNCO during the summer monsoon season of 2021, this study first verified the performance of the MRR at the MNCO by comparing it with the collocated rain gauge and PARSIVEL disdrometer. Then, the vertical structures of integral rain parameters and DSD in different rainfall rate categories were statistically analyzed.

Good agreement between the MRR, rain gauge, and disdrometer measurements is evidenced by the high CCs (exceeding 0.93) for the rainfall trends and hourly rainfall observations. However, the MRR tends to overestimate the gauge and PARSIVEL disdrometer rain rates by 5.0% and 25.9%, respectively. Compared with the PARSIVEL disdrometer, the MRR in general overestimates (underestimates) the concentration of smaller (larger) raindrops, and there is reasonable agreement for raindrops between 0.5 and 1.8 mm in diameter. However, it is well known that the PARSIVEL disdrometer usually underestimates the concentration of small raindrops and rainfall rate due to the assumption of “one drop at a time” (Wen L. et al., 2017). This is partly responsible for the MRR overestimation of small raindrop amounts.

The MRR measurements in Mêdog were classified into four classes in terms of rain rate: 0.02 < R ≤ 0.2 mm h⁻¹; 0.2 < R ≤ 2.0 mm h⁻¹; 2.0 < R ≤ 20.0 mm h⁻¹; and 20.0 < R ≤ 200.0 mm h⁻¹. The characteristics of precipitation vertical structure in Mêdog exhibit a distinct discrepancy for different rainfall rate categories. Strong evaporation occurs in the weakest rainfall regime, leading to radar reflectivity, rainfall rate, LWC, and the concentration of raindrops decreasing at near the ground. In terms of rainfall rate category of 0.2–2.0 mm h⁻¹, the profiles of rainfall rate and LWC (small- and mid-sized drops) show almost uniform vertical structure, while radar reflectivity (large drops) shows an apparent negative (positive) gradient. The coalescence process of raindrops is slightly dominant in rain rate category of 2.0–20.0 mm h⁻¹, resulting in the increase of integral rain parameters with decreasing height. The coalescence mechanism prevails during heavy rainfall (e.g., R > 20.0 mm h⁻¹), and steep negative gradients can be observed, although a disturbance appears at a height of 1.5–2 km AGL due to local strong updrafts associated with the interaction of the steep terrain and convective rain.

The profiles of integral rain parameters for the four different rain rate categories in Mêdog are similar to those in other regions such as Xingtai, North China and Shillong, Northeast India. However, Mêdog precipitation has a lower mean fall speed and lower radar reflectivity for each rain rate category. This indicates that Mêdog precipitation is characterized by size-limited drops during summer monsoon season, which is consistent with previous studies (Wang et al., 2021, 2022; Li R. et al., 2022). The breakup of large raindrops during their falling process is more significant in Mêdog, which may be partly responsible for the predominant size-limited drops in this region.

The DSD are retrieved from the Doppler spectra of the MRR following the hypothesis of a static atmospheric environment. Although the hypothesis is generally unjustified, the vertical structures of DSD and the integral rain parameters observed in stratiform precipitation may be reasonably accepted considering the very weak vertical wind in stratiform precipitation. In terms of convective precipitation (e.g., R > 20 mm h⁻¹), strong updrafts may cause large errors in the retrieval procedures of DSD and integral rain parameters. However, the mean vertical structures of DSD and rain parameters can, to a certain extent, reflect microphysical processes. Recently, raindrop spectrum measurements have been used to eliminate the influence of vertical wind on the performance of MRR (Adirosi et al., 2016). Jameson et al. (2021) proposed a method in which the relationship between the measured reflectivity and the calculated rain rates conforms to the physical theory in order to remove the effects of mean vertical drafts. A wind profile radar and a PARSIVEL disdrometer were deployed close to each other at the MNCO, and the estimation of the rain vertical structure was improved by combining the K-band MRR, PARSIVEL disdrometer, and wind profile radar observations.

We thank the Mêdog Meteorological Bureau of Xizang Autonomous Region, for maintenance of the remote sensors deployed at the Mêdog National Climate Observatory, including the K-band micro rain radar, Ka-band cloud radar, PARSIVEL disdrome-ter, etc. Special thanks go to Suolang Zhaxi and Ting Wang for their patience and conscientiousness.