Comparison of Microphysical Characteristics of Warm-sector, Frontal and Shear-line Heavy Rainfall During the Pre-summer Rainy Season in South China

South China is located in the low latitude region, with intense solar radiation, strong East Asian summer monsoon, complex atmospheric circulation, and special coastal terrain, and the rainfalls in this region are changeable, instantaneous, intense, and lengthy (Zhang et al. ^[1]). The precipitation amount between the period from April to June (known as the pre-summer rainy season) accounts for 40%-50% of the total precipitation each year in south China (Wu et al. ^[2]). Before and after the onset of the summer monsoon, due to the particular configurations of atmospheric circulation, the rainfalls in south China gradually change from frontal rainstorms to monsoon rainstorms (Liang et al. ^[3]). In general, the rains are diverse and complex, and the coexistence of shear-line heavy rain (SR), warm-sector heavy rain (WR), and frontal heavy rain (FR) frequently occur (Luo et al. ^[4]).

The definition of the WR was proposed by Huang et al. ^[5] in 1986 when they classified the pre-summer rainy season rainfall in south China. It refers to rainstorms that occur on the side of a warm sector 200-300 km away from a surface front, or in the absence of a significant shear at the confluence of westerly and easterly winds. The triggering factors of the WR are complex (He et al. ^[6]; Sun et al. ^[7]), warm and humid atmosphere (Wu et al. ^[8], Zheng et al. ^[9]), surface sensible / latent heating (Yin et al. ^[10]), special topographies (Wu et al. ^[8]; Zhong et al. ^[11]; Wu et al. ^[12]), thermal differences between the land and sea (He et al. ^[6]; Sun et al. ^[7]), low-level jet (Sun et al. ^[7]; Wu et al. ^[12]; Liang et al. ^[13]), the release of condensation latent heat of water vapor (He et al. ^[6]), surface cooling effect (Wu et al. ^[14]), favorable circulation structure (Zheng et al. ^[9]; Zhao et al. ^[15]), and weak cold air intrusion (Sun et al. ^[7]; Zheng et al. ^[9]; Wang et al. ^[16]) may be related. Typically, the precipitation intensity of the WR is 3-5 times that of the SR, for which the area is smaller, and the scale is only tens to hundreds of kilometers (Lin ^[17]).

For the FR, when the East Asian summer monsoon arrives, the cold air originating from northwestern China and the water vapor from the western Pacific Ocean collide over southern China (Zheng et al. ^[18]). The temperature difference between the two types of air masses is large (Liu and He ^[19]), lighter warm and moist air is lifted above the cold air on a large scale, and the water vapor is cooled and condensed to form frontal rain (Zheng et al. ^[18]).

For the SR, the temperature difference between the two sides of a shear line near the low altitude (700-850 hPa) is not obvious, but the horizontal shear of the wind is large. When the northern side of the shear line is dominated by easterly winds and the southern side by southerly winds, water vapor content in the flow will be abundant and a shear vortex is likely to occur. At this time, the airflow will converge to facilitate the accumulation and uplift of water vapor, resulting in the occurrence of heavy rainfall on a large scale (Larson ^[20]). Since the atmospheric circulation and microphysical properties of SR are different from those of WR and FR, we decided to perform some additional comparative analysis of the microphysical properties of these three types of precipitation.

The differences in the structure of atmospheric thermodynamics and dynamics will lead to different types of rainfall, as well as different development and evolution processes of mesoscale convective systems (MCS) and microphysical characteristics of rainfalls (Wen et al. ^[21]; Han et al. ^[22]; Lai et al. ^[23]). Based on the measurements of polarization radar, Li et al. ^[24] identified that the record-breaking May 7 extremely heavy rainfall process in Guangzhou in the pre-summer rainy season of 2017 was primarily produced by a warm rain process. The sufficient supply of warm and moist airflow makes the rainfalls in south China in the pre-summer rainy season have the features of a low mass center and warm rain process, resulting in higher precipitation efficiency (Fu et al. ^[25]). Based on the polarization radar, Huang et al. ^[26] found that in the extreme rainstorms in south China, the liquid water content under the 0 ℃ level increased rapidly, mainly through the warm rain process, and the raindrops mainly grew through the collision process. Han et al. ^[22] studied the microphysical characteristics of two types of rainfall in south China and obtained that the coastal WR has stronger convective intensity and larger raindrop size than the inland FR.

During 10-15 May 2022 (Local Standard Time (LST), the same hereafter), a persistent rainfall event occurred in south China, in which the SR, WR, and FR were observed successively. As this event had additional types of rainfall and longer duration than in Han's research, this precipitation process was selected for additional comparative analysis of the microphysical properties of different types of precipitation in southern China. Based on the detection of 11 S-band polarimetric radars in Guangdong province, this study compared the microphysical characteristics of three types of rainfall events: SR, WR, and FR, in an attempt to reveal the different microphysical characteristics of those rainfall events during the pre-summer rainy season in south China.

By 2022, all S-band Doppler weather radars in Guangdong have upgraded their polarization functions. The major radar parameters, such as the radar reflectivity (Z_H, dBZ), differential reflectivity (Z_DR, dB), and the specific differential phase shift (K_DP, ° km^-1), have been discussed in related studies (Chen et al. ^{[27, 28]}; Xia et al. ^[29]; Liu et al. ^[30]). The yellow dots in Fig. 1 represent the spatial distribution of the 11 S-band polarimetric radars. The observation range of each radar is 230 km. When all radars are networked, they can cover the entire territory of Guangdong province and the coastal areas, forming multiple coverage for various types of rainfall (Li et al. ^[31]).

Figure 1.  Spatial distribution of S-band polarimetric radars (yellow dots) and automatic weather station rain gauges (red dots).

To obtain high-reliability radar data, quality control of radar echoes was carried out to eliminate non-precipitation echoes referring to previous schemes (Liu et al. ^[30]; Wu et al. ^[32]; Wang et al. ^[33]), including ground clutter, clear sky echo, and abnormal beam. The Z_DR system error of each radar was corrected by the micro raindrop method (Liu et al. ^[30]), then the radar data were gridded and interpolated into 3-D Cartesian coordinates. The horizontal resolution of the interpolated data is 250 m, and the vertical resolution is 500 m (Han et al. ^[22]; Yu et al. ^[34]).

According to the convective cell (CC) identification method (Han et al. ^[22]; Wang et al. ^[33]), if the Z_H exceeds 40 dBZ in more than one-third of the grid points, it was considered a CC. The echoes were counted from 500 m, taking into account that some radars lacked scans at low altitudes.

Hourly precipitation data were detected from 477 national automatic weather stations (AWSs) in south China and 3388 regional AWSs in Guangdong province from May 10-15, 2022 (red dots in Fig. 1). The quality of precipitation data was controlled by excluding accumulated observation time that was less than 95%. The lightning data was extracted from the Guangdong-Hongkong-Macao lighting location system (GHMLLS), which is the most influential professional lightning location system in Guangdong, recording the time, location, strength, cloud flash, and ground flash of lightning. The environmental conditions of this rainfall event were analyzed by the European Center for Medium-Range Weather Forecasts (ECMWF) fifth-generation hourly reanalysis data (ERA5) with a spatial resolution of 0.25 ° × 0.25 °.

Since 2016, seven 2D-Video Distrometers (2DVDs) have been deployed by the Longmen Field Experiment Base for Cloud Physics of China Meteorological Administration in the coastal and inland rainstorm-prone areas of Guangdong (Liu et al. ^[35]). By scanning the shape, size, and final velocity of raindrops, the raindrop size distribution (RSD) of different types of rainfall can be obtained (Lai et al. ^[23]; Feng et al. ^{[36, 37]}; Ding et al. ^[38]). With the long-term RSD data and the S-band polarimetric radar variables, the inversion relation between them can be obtained. In this study, the constrained Gamma distribution model was applied to retrieve the RSD data based on the S-band polarimetric radar observations (Liu et al. ^[30]). The parameters of raindrop mass-weighted mean diameter (D_m, mm) and the logarithmic-normalized intercept (log₁₀N_w, the unit of N_w is mm^-1 m^-3) were calculated from the retrieved Gamma raindrop size distribution. The Z_DP-R method was adopted to retrieve liquid water content (LWC) and ice water content (IWC) in this study (Li et al. ^[39]). The hydrometeors were classified by the algorithm described by Park et al. ^[40] and optimized by Wu et al. ^[32]. Ultimately, the radar echoes were divided into ten categories: (a) ground clutter (GC); (b) biological scatters (BS); (c) dry snow (DS); (d) wet snow (WS); (e) ice crystal (CR); (f) graupel (GR); (g) big drops (BD); (h) light and moderate rain (RA); (i) heavy rain (HR); and (j) a mixture of rain and hail (RH), only meteorological radar echoes (excluding GC and BS) were used in this study.

At 1400 LST on May 10, at 850 hPa (Fig. 2a), southwesterly-westerly winds dominated over Guangdong. A shear line was located in the southern province of Hunan. A surface front was located in northern Guangdong, and Guangdong was mainly affected by the southerly winds near the surface (Fig. 2d). At 2100 LST, as the low equivalent potential temperature (θ_e) air in northern Guangdong gradually advanced southwards, the 850 hPa shear line (Fig. 2b) and the surface front (Fig. 2f) moved southwards. It is worth noting that the shear line is one of the major precipitation weather systems, and it frequently cooperates with the surface frontal systems. On the shear line, there is commonly a horizontal convergence and upward motion of the airflow, providing dynamical conditions for the development of precipitation. In this event, at 850 hPa (Fig. 2a-2b), the temperature difference of the airflow on both sides of the shear line was slight, but there was evidently opposite wind speed (northerly winds on the north side and westerly-southwesterly winds on the south side), which caused the convergence of the cold and warm air on the ground in northern Guangdong, and finally, the FR occurred near the surface front and the south of the shear line.

Figure 2.  (a-d) Winds, equivalent potential temperature (colored), and geopotential height (blue contours) at 850 hPa. The purple line indicates the location of the shear line. (e-h) 10-m winds, 2-m equivalent potential temperature (colored), and mean sea level pressure (blue contours). White lines with triangles indicate the location of the surface front. The first column is 1400 LST on May 10; the second column is 2100 LST on May 10; the third column is 1200 LST on May 11; and the fourth column is 1200 LST on May 13.

From 2100 LST, it can also be seen that the low-level winds over the coast of Guangdong and the sea adjacent to it gradually strengthened (Fig. 2c-2d, 2f-2g). The winds near the Pearl River Estuary changed to be southerly (Fig. 2f-2g). The intensification of the southerly component enhanced the wind speed convergence and orographic uplift near the coast (Figs. 2f and 3b). At this time, more convection was initiated near the Pearl River Estuary. During this period, coastal precipitation was more than 200 km away from the surface front (Figs. 2f and 3b), having typical WR precipitation characteristics (Huang et al. ^[5]; Wu et al. ^[12]). Therefore, the period from 2100 LST on May 10 to 1100 LST on May 11 was chosen as the WR period.

Figure 3.  Cross-section along 112°E of equivalent potential temperature (colored), meridional wind, and vertical wind (vertical wind speed multiplied by 100). (a) 1400 LST on May 10; (b) 2100 LST on May 10; (c) 1200 LST on May 11; and (d) 1200 LST on May 13.

After that, the low θ_e air in northern Guangdong further moved southwards (Fig. 2g-2h). At 1200 LST on May 13, the front moved to central and southern Guangdong (Fig. 2h). From the meridional cross-section of the equivalent potential temperature and winds, the southward movement of the front can be distinctly seen (Fig. 3). At 1200 LST on May 11, the distance between the front and the coastal precipitation area was less than 200 km. The convection was concentrated in the east of the Pearl River Estuary and was governed by the FR.

Figure 4a shows the distribution of total precipitation determined from the AWSs in south China for the entire precipitation event. Rainfall was mainly in coastal and central inland areas over Guangdong, where accumulations reached 200-800 mm and 400-600 mm, respectively.

Figure 4.  (a) Total precipitation from 10-15 on May 2022 and accumulated precipitation for the three types of rainfall of (b) SR, (c) WR, and (d) FR.

According to the synoptic situation (Figs. 2 and 3), accumulated precipitation of AWSs (Fig. 4), and reflectivity observations (Fig. 5), the SR region was mainly distributed in central inland Guangdong (red box in Fig. 4b). On both sides of the SR, the wind direction and velocity at 850 hPa were relatively different. The regional MCS governed the rainfall, and the rain bands were elliptical. The duration of the rain was mainly from 1400 LST on May 10 to 0200 LST on May 11, with a maximum accumulation of 150-200 mm. The WR mainly occurred on the west coast of Guangdong (red box in Fig. 4c) and featured a linear mesoscale convective system (MCS) parallel to the coastline. The main duration of the rainfall was from 2100 LST to 1100 LST on May 10, with accumulations reaching 200-300 mm. The FR region was close to the inland area of the coast (red box in Fig. 4d), the main duration was from 1200 LST to 2400 LST on May 13, and the maximum accumulated precipitation was 100-200 mm.

Figure 5.  The radar reflectivity data from a 3-km constant altitude plan position indicator (CAPPI) of (a) SR, (b) WR, and (c) FR. The red boxes indicate the corresponding rainfall areas.

The time-height distributions of the mean Z_H, Z_DR and K_DP of the convective cells (CCs) for the three types of rainfall (WR, SR, and FR) are shown in Fig. 6. The mean Z_H of CCs for three types of rainfall were all higher than 40 dBZ below the 0 ℃ level. However, the distribution range of high Z_H values was found to be most widely distributed in the WR (Fig. 6b), less in the FR (Fig. 6a), and least in the SR (Fig. 6c). These indicate that Z_H is the strongest in the WR, followed by the FR and the weakest in the SR. The Z_DR increased as the height decreased in the WR, and the averaged Z_DR in the WR (Fig. 6d) reached a maximum of 0.5-1.0 dB below the 0 ℃ level. To liquid raindrops, a larger Z_DR suggests a larger average raindrop diameter. This shows that the average size of raindrops in WR CCs increases during the rainfall. The average K_DP in the WR (Fig. 6h) was the largest, reaching a maximum of 0.5-1.5° km^-1 below the 0 ℃ level, obviously larger than those in the SR and FR. The S-band polarimetric radar quantitative precipitation estimation (QPE) algorithm (Guo et al. ^[41]; Zhou et al. ^[42]) shows that K_DP is closely related to rainfall intensity. A larger K_DP indicates larger precipitation. Thus, the rainfall intensity in the WR was significantly stronger than that in the SR.

Figure 6.  Time-height distributions of the mean Z_H, Z_DR, and K_DP of CCs of three types of rainfall. (a, d, g): SR, (b, e, h): WR, and (c, f, i): FR.

Figure 7a-7c shows the maximum height of Z_H for the three types of rainfall with different intensities. The red curve represents the Z_H for 40 dBZ, and the blue curve represents the Z_H for 20 dBZ. As shown in Fig. 7, the Z_H for 20 dBZ and 40 dBZ reached up to 16-20 km and 10-14 km in the WR, respectively, followed by 14-18 km and 8-12 km in the FR alone. The maximum Z_H heights in the SR were the lowest at 20 dBZ and 40 dBZ. These indicate that the convection in the WR was most intense during this rainfall event.

Figure 7.  The Z_H heights at 20 dBZ and 40 dBZ for the three types of rainfall of (a) SR, (b) WR, and (c) FR, and the ratio of CCs area versus rainfall area (units: %) and the lightning density for the three types of rainfall (units: 10^-4 times min^-1 km^-2) of (d) SR, (e) WR, and (f) FR.

Figure 7d-7f shows the proportion of the CCs area versus the rainfall area and the lightning density in the SR, WR, and FR, and the proportion of the CCs area was calculated as follows: CCs areas versus areas with Z_H > 20 dBZ. As shown in Fig. 7, the area of CCs in the WR was the largest. The averaged (maximum) ratio was 6.2% (13.1%) in the WR, 4.6% (6.9%) in the FR, and 3.8% (7.5%) in the SR. Meanwhile, the lightning density in the WR was higher than that in the FR and SR. The averaged (maximum) lightning density was 3.0 (11.7) × 10^-4 times min^-1 km^-2 in the WR, and was 1.5(5.3)×10^-4 times min^-1 km^-2 in the FR, and only 0.6(2.4) × 10^-4 times min^-1 km^-2 in the SR. It can be seen from CCs ratio and lightning density that the convective updrafts in the WR was the most vigorous, followed by the FR, and the SR was the weakest.

Furthermore, we compared the rainfall rates of different Z_H (Fig. 8a-8c) and the probability density distributions of 20 dBZ (solid lines) and 40 dBZ (dashed lines) radar echo heights in the CCs for the three types of rainfall (Fig. 8d-8f). The maximum Z_H of the CCs was classified into three grades (40-45 dBZ, 45-50 dBZ, and > 50 dBZ). For all the three types of rainfall, a higher Z_H value corresponds to higher precipitation intensity. In addition, with identical echo classification, the precipitation intensity in the WR was larger than in the other two types of rainfall. As can be seen from Fig. 8d-8f, the height of the probability density distribution of 20 dBZ and 40 dBZ radar echo heights in the WR was the highest, followed by the FR, and the height of the SR was the lowest.

Figure 8.  Rainfall rates of different Z_H (a, b, c), probability density distribution (units: %) of 20 dBZ (solid lines) and 40 dBZ (dashed lines) radar echo height (d, e, f), and the maximum Z_H of CCs were divided into three grades (40-45 dBZ, 45-50 dBZ, and > 50 dBZ). The three types of rainfall are SR, WR, and FR.

To exhibit the differences in microphysical features of the SR, WR, and FR more distinctly, the contoured frequency by altitude diagrams (CFADs) (Fig. 9a-9c, 9e-9f, and 9i-9k) and their profile distributions (Fig. 9d, 9h, and 9l) of radar variables were analyzed. The Z_H of the WR (Fig. 9a-9c) was mainly distributed between 40-50 dBZ below the 0 ℃ level, and the Z_H of the SR and FR was mainly distributed between 40-46 dBZ. The vertical profile of Z_H is closely related to the storm intensity in the previous analysis (Han et al. ^[22]; Li et al. ^[24]; Wang et al. ^[33]). As shown in Fig. 9d, the Z_H in the WR was larger than that in the SR and FR at all altitudes, indicating that the convection intensity of the WR was most vigorous in this event. The vertical profile values of Z_H in the FR were slightly higher than in the SR.

Figure 9.  The contoured frequency by altitude diagrams (CFADs) (units: %) of Z_H, Z_DR, and K_DP of the CCs for the three types of rainfall (a-c, e-f, i-k), and corresponding vertical profile distributions (d, h, l) for the three types of rainfall of SR, WR, and FR.

As can be seen from the CFADs of Z_DR (Fig. 9e-9g) and vertical profile (Fig. 9h), the Z_DR of the three types of rainfall increased rapidly as the height decreased from 7 km to 4 km. Below 3 km, the mean Z_DR changed little in the SR, while it continued to increase in the WR and FR, from 0.7 dB to 0.9 dB and from 0.65 dB to 0.78 dB, respectively, as the altitude decreased. This means that the average size of the raindrops increased at heights below 3 km in the WR and FR. Combined with the variation of the mean value of the Z_H, it can be found that there were clear collision-coalescence processes in the WR and FR below the 3-km altitude, especially for the WR. For SR, however, it is likely that the raindrop breakup and coalescence processes tended to be equilibrated. This may be related to more abundant water vapor and stronger updrafts in the WR and FR where raindrops become larger with the support of stronger air currents.

Below the 0 ℃ level, the K_DP was mainly distributed between 0-1.6 ° km^-1 in the WR (Fig. 9j), and between 0-1.0 ° km^-1 in the SR (Fig. 9i) and FR (Fig. 9k). The mean K_DP for the three types of rainfall increased rapidly with the height decreasing below the 0 ℃ level. The K_DP increased from 0.6 ° km^-1 to 1.05 ° km^-1 in the WR (Fig. 9l), and from 0.37 ° km^-1 to 0.47 ° km^-1 in the SR (Fig. 9l), and from 0.37 ° km^-1 to 0.54 ° km^-1 in the FR (Fig. 9l). From 1 to 6 km in height, the vertical profile values of the K_DP in the WR were significantly larger than those in the SR and FR. For pure liquid precipitation, a larger K_DP indicates stronger rainfall intensity. It can be concluded that the intensity of rain in the WR was considerably larger than that in the SR and FR.

The RSD is directly related to the condensation, coalescence, evaporation, and falling velocities of hydrometeors and is the most influential parameter affecting microphysical processes in clouds. The joint probability density distributions of the D_m and log₁₀N_w parameters for the three types of precipitations retrieved from the S-band dual-polarization radars at altitudes below 2 km are shown in Fig. 10. Here, D_m denotes the average raindrop size, and N_w is related to the concentration of raindrops. The statistical results of D_m and log₁₀N_w show that the three types of rainfall belonged to marine-type precipitation (Feng et al. ^{[36, 37]}). Specifically, the average value of D_m and N_w in the WR were the largest (1.6 mm, 4.28 mm^-1 m^-3), the SR (1.53 mm, 4.23 mm^-1 m^-3) and the FR were close (1.55 mm, 4.22 mm^-1 m^-3). These suggest that raindrop sizes and number concentrations were larger in WR than in SR and FR.

Figure 10.  Joint probability density distribution of D_m (units: mm) and log₁₀N_w (units: mm^-1 m^-3) of the CCs for the three types of rainfall of (a) SR, (b) WR, and (c) FR.

Figure 11 shows the average distribution of the vertical profiles of D_m and N_w retrieved from radar observations for the three types of heavy rainfall below 4 km in height. The vertical profiles of D_m and N_w were considerably different for the three types of rainfall. The mean D_m of WR and FR increased with decreasing altitude, especially for WR. In contrast, the mean values of log₁₀N_w for WR and FR decreased with the height decreasing. These confirm again that for both WR and FR, and especially for WR, there was a clear coalescence process during the falling process. The mean value of D_m in the SR showed a decreasing trend from the height of 4 km to 2.5 km, but the transition was less pronounced below 2.5 km. The mean value of log₁₀N_w for SR showed an increasing trend from 4 km to 1.5 km and then decreased. This also suggests that for SR below 3 km, fragmentation and coalescence processes might tend to be in equilibrium.

Figure 11.  Mean profiles of the retrieved (a) D_m (units: mm) and (b) log₁₀N_w (units: mm^-1 m^-3) of the CCs. Solid blue, red, and green lines represent the SR, WR, and FR, respectively.

As shown in Fig. 12, the vertical profile of the LWC (units: g m^-3) and IWC (units: g m^-3) of CCs for the three types of rainfall were determined according to different maximum Z_H classification, (a) 40-45 dBZ, (b) 45-50 dBZ, and (c) > 50 dBZ. For each Z_H classification, the LWC and IWC in the WR had the highest vertical content at all height levels, followed by the FR and SR. In particular, when maximum Z_H > 50 dBZ (Fig. 12c), the IWC in the WR reached a maximum of 1.3 g m^-3 above the 0 ℃ level, suggesting that the ice-phase process was active. The LWC in the WR reached a maximum of 4.8 g m^-3 below the 0 ℃ level, indicating that the IWC accounted for about 27% of LWC during the falling process for the WR. Correspondingly, when maximum Z_H > 50 dBZ, the IWC accounted for about 22% and 20% of LWC during the rain process for the SR and FR, respectively. These suggest that the ice-phase process played a more prominent role in the WR systems than in the SR and FR, and can produce heavier rainfall. The IWC in the SR was the lowest of the three heavy rainfall systems, indicating that the ice-phase process was the least active.

Figure 12.  The vertical profile of LWC (units: g m^-3) and IWC (units: g m^-3) of CCs for the three types of rainfall according to different maximum Z_H classification for (a) 40-45 dBZ, (b) 45-50 dBZ, and (c) >50 dBZ. The blue, red, and green solid lines represent the LWC of SR, WR, and FR, respectively, and the solid black line indicates the 0 ℃ level.

The proportion of heavy rains in the WR (Fig. 13b) was obviously larger than in the FR and SR. Hence, the WR had the highest rainfall intensity. From 4 to 8 km in height, the average graupel ratio reached up to 26% in the WR (Fig. 13b), followed by the FR (Fig. 13c) (~20%), while the SR (Fig. 13a) was only 17%. Graupel occupied the maximum fraction in the WR, suggesting that its riming process was likely to be the most active. In general, more proportion of graupel in the rain, stronger updrafts, and more abundant water vapor supply might lead to larger-sized raindrops during the melting and collision-coalescence processes.

Figure 13.  The stacked frequency (units: %) of each identified hydrometeor by altitude of the CCs for the three types of rainfall.

During 10-15 May 2022, a heavy rainfall event occurred in south China, in which three types of rainfall (SR, WR, and FR) happened successively. In this study, based on the combined observations from 11 polarimetric radars, AWSs observations, ERA5 reanalysis, lightning observations, and the corresponding retrieval algorithms of microphysical characteristics of rainfall, we investigated the differences in the microphysical characteristics of these three rainfalls. The conclusions are as follows:

(1) Throughout the entire altitude range, the mean Z_H values of the CCs in the WR were higher than those in the SR and FR. The Z_H values of 20 dBZ reached the height of 18 km in the WR, followed by the FR and SR. The average area ratio of CCs was 6.2%, 4.6%, and 3.8% in the WR, FR, and SR, respectively. Meanwhile, the lightning densities in the WR were higher than those in the SR and FR. These indicate that the convection motion in the WR was the most vigorous, followed by the FR and SR. The mean K_DP in the WR was the largest below the 0 ℃ level, indicating that its rainfall intensity was also the largest. Below 3 km, there was likely an obvious coalescence process in the WR and FR. The breaking process in the SR balanced with the coalescence process below 3 km.

(2) The raindrop size distribution characteristics of the three types of heavy rainfall all belong to marine-type precipitation. The WR process had the highest proportion of graupel (26%), the most active riming and coalescence process. In addition, WR raindrops had the highest mean values of D_m and N_w compared to SR and FR. Moreover, the mean values of D_m and N_w for FR and SR were close to each other.

(3) The IWC of the WR was considerably higher than that of the SR and FR at all heights, and the lightning density was also the largest, indicating that the ice-phase process in the WR was likely the most active. The ice-phase process in the SR was the least active. The IWC in the WR accounted for about 27% of LWC when Z_H exceeded 50 dBZ, which is significantly higher than those in the SR (22%) and FR (20%). These suggest that the ice-phase process in the WR contributed higher liquid rainfall than in the SR and FR.

Based on the observations from multiple S-band polarimetric radars, we presented a detailed comparison of the vertical structures and microphysical characteristics of the WR, SR, and FR in south China in this study. However, this analysis only focuses on the process of heavy rainfall that occurred in May 2022, and a more in-depth study based on additional case statistics is needed.