Precipitation Microphysical Characteristics of Typhoon Ewiniar (2018) before and after Its Final Landfall over Southern China

Tropical cyclone (TC) disasters threaten lives and cause economic losses. To accurately forecast the intensity and track of a TC, it is essential to understand the microphysical processes of the clouds and precipitation (Didlake and Kumjian, 2017; Park et al., 2020).

In recent years, in situ disdrometer data, satellite observations, and radar observations have been widely used to investigate the microphysical structure of TC precipitation. In particular, the development of polarimetric radar has led to valuable new insights into the microphysical characteristics of TCs (Brown et al., 2016). The parameters obtained from polarimetric radar measurements such as the radar reflectivity (Z_H, dBZ), the differential reflectivity (Z_DR, dB), and the specific differential phase shift (K_DP, ° km^–1) are related to the microphysical characteristics such as the shape, type, and size of the hydrometeor (Seliga and Bringi, 1976; Kumjian, 2018). Due to this unique superiority, polarimetric radar has been widely used to study cloud precipitation microphysics, for radar quantitative precipitation estimation (QPE), and for numerical modeling evaluation (Wen et al., 2017a; Chen et al., 2019; Wang et al., 2019, 2020; Xia et al., 2020; Wu et al., 2021a; Zhang et al., 2021).

The raindrop size distribution (DSD) in a typical TC exhibits high concentrations of small and medium raindrops (diameter < 3 mm), which dominate the precipitation caused by the TC (Dehart and Bell, 2020; Feng et al., 2020; Zheng et al., 2021). Brauer et al. (2021) found that during landfall, the microphysical characteristics of Tropical Storm Bill included warm rain signatures and collision–coalescence processes. In contrast, Wu et al. (2018b) reported that the ice-phase microphysical process dominated the heavy surface rainfall in the outer rainband of Typhoon Nida (2016). Wang et al. (2016) found that the concentration of small raindrops in Typhoon Matmo (2014) was higher than those of other TCs formed in the same ocean, which may be the result of a higher concentration of condensation nuclei. Huang et al. (2022) found that more graupel melted into raindrops, accompanied by a significant accretion process, resulting in heavy rainfall with a large raindrop size in the outer eyewall of Typhoon Lekima (2019).

Due to the different dynamic and thermodynamic mechanisms of rain formation, the microphysical characteristics of a TC are significantly variable in different rain regions and rain types (Seliga and Bringi, 1976; Park et al., 2020). Bao et al. (2020) found that the average mass-weighted mean diameter (D_m, mm) decreases and the average logarithmic normalized intercept parameter (lgN_w, mm⁻¹ m⁻³) increases radially from the center of the TC to the outer rainbands. Cecil and Zipser (2002) studied the reflectivity, ice scattering, and lightning characteristics in the eyewall, inner rainbands, and outer rainbands and found that more supercooled cloud water occurred in the outer rainbands. Compared with the stratiform precipitation in the typhoon systems in the western Pacific, the convective precipitation had a higher D_m and higher concentration of small and medium raindrops (Chang et al., 2009; Seela et al., 2022). Zhang et al. (2022) found that landfalling TCs in northeast China had smaller N_w and D_m than landfalling TCs in other parts of China.

Moreover, due to the asymmetric dynamic, thermodynamic, and microphysical processes in a TC, not only the rainfall but also the wind and lightning activity exhibited non-symmetric signatures (Corbosiero and Molinari, 2003; Braun et al., 2006; Chen et al., 2006; Huang and Liang, 2010). Wen et al. (2017b) found that the maximum rainfall mainly occurred in the southwestern quadrant (i.e., the downshear of the vertical wind shear) of the TC’s center, which was connected with the deep vertical wind shear (200–850 hPa) (Xu et al., 2014). Lonfat et al. (2004) found that the maximum rainfall of tropical storms mainly occurred in the front-left quadrant and that of hurricanes or typhoons mainly occurred in the front-right quadrant. Compared with the eyewall and inner rainbands, the outer rainbands had the largest average flash densities (Molinari et al., 1994; Zhang et al., 2012). In addition, the ratio of the flash density in the outer rainbands was asymmetric and almost occurred in the deep convective region (Qie et al., 2014).

Although TCs mostly form over oceans, many studies in China have focused on the precipitation microphysical characteristics of TCs after landfall (Yu et al., 2015; Wang et al., 2016, 2018b; Wu et al., 2021b; Zheng et al., 2021). These studies analyze the precipitation microphysical characteristics (e.g., the raindrop size distribution) and microphysical processes (e.g., the ice-phase process) in the eyewall, inner rainbands, and outer rainbands. However, the microphysical characteristics of TC precipitation are different before and after landfall due to the change in underlying surfaces (Cecil and Zipser, 2002; Rosenfeld et al., 2012).

Few studies have investigated the variations in the DSD before and after landfall (Chen et al., 2012; Janapati et al., 2017; Lin et al., 2020; Feng et al., 2021) and the variations in the DSD in different quadrants (Didlake and Kumjian, 2018; Feng and Bell, 2019; Huang et al., 2022). Janapati et al. (2017) found that compared with before landfall, the after landfall mean D_m values of cyclones Thane and Nilam were higher, and the after landfall mean D_m value of cyclone Jal was smaller. Feng and Bell (2019) found that compared with other quadrants, larger size of raindrops occurred in the downshear left quadrant in the eyewall, which was related to the greater amount of collisions and coalescence. Didlake and Kumjian (2018) found that the signature of hydrometeor size sorting is obvious in the primary eyewall of Hurricane Irma. Fewer studies have focused on the microphysical non-symmetric signatures in the different regions (eyewall, inner core, and outer rainbands) and different quadrants (downshear left, downshear right, upshear left, and upshear right) before landfall and after landfall.

In this study, we documented the microphysical non-symmetric characteristics and microphysical processes during the final landfall of Typhoon Ewiniar (2018) based on 2D video disdrometer (2DVD) and S-band polarimetric radar data. Due to the different types of underlying surfaces, the differences in these characteristics and processes during the before landfall and after landfall periods were also a focus. The goals of this study were to deepen our understanding of the microphysical structures and microphysical processes of TCs and to provide a reference for microphysical parameterization schemes in numerical models.

Typhoon Ewiniar (2018) was the first TC to make landfall in China in 2018, and it occurred 21 days earlier than the usual first landfall date (27 June) (Li et al., 2021). It caused heavy rainfall in Guangdong Province and resulted in five deaths. Figure 1 shows the track of Typhoon Ewiniar (2018) recorded by the National Meteorological Center (the website is https://typhoon.nmc.cn/web.html). First, it made landfall in Xuwen, Guangdong, at 2225 UTC on 5 June 2018. Its second landfall was on Hainan Island at 0650 UTC on 6 June 2018. Finally, it made landfall in Yangjiang, Guangdong, at 1230 UTC on 7 June 2018. For each landfall, the associated time was when the center of the TC first crossed into the coastal area (Janapati et al., 2017).

Figure 1.  Track of Typhoon Ewiniar (2018) and accumulated rainfall from 2200 UTC 6 June 2018, to 0800 UTC 8 June 2018. The blue stars represent the locations of the GZ S-band polarimetric radar and the YJ S-band polarimetric radar. The blue square represents the HLD 2DVD. The black line represents the track from 2200 UTC 6 June 2018 to 0800 UTC 8 June 2018.

In this study, we mainly focused on the microphysical characteristics during the period of the final landfall (from 2200 UTC 6 June 2018, to 0800 UTC 8 June 2018). Thus, in this study, the period of before landfall was from 2200 UTC 6 June 2018, to 1230 UTC 7 June 2018, and that after landfall was from 1230 UTC 7 June 2018, to 0800 UTC 8 June 2018. During the period of the final landfall, the accumulated rainfall was more than 250 mm over the coast from the eastern part of western Guangdong Province to the western part of eastern Guangdong Province (Fig. 1).

During the period of the final landfall, the S-band polarimetric radar instruments were located at the Yangjiang (YJ) site (21.85°N, 111.98°E) and Guangzhou (GZ) site (23.00°N, 113.36°E), and the data collected by these instruments were analyzed in this study. The instruments have nine elevation angles (0.5°, 1.5°, 2.4°, 3.3°, 4.3°, 6.0°, 9.9°, 14.6°, and 19.5°), a high resolution of range (250 m), and a 6-min time interval. In this study, the radar observation data and inversion data were interpolated to horizontal and vertical resolutions of 1 km and 0.5 km, respectively.

Polarimetric radar data cannot be directly used for data analysis. Before data analysis, it is necessary to conduct a series of quality control procedures to remove the influence of non-meteorological echoes. First, we used the method of Tang et al. (2014) to remove non-standard blockage mitigation (e.g., wood and building). Then, the hydrometeor classification method (Park et al., 2009; Wu et al., 2018a) was used to remove ground clutter and biological echoes. The threshold checks of the co-polar cross-correlation coefficient (ρ_HV), K_DP, and signal-to-noise ratio (SNR) were used to distinguish between meteorological echoes and non-meteorological echoes. On this basis, Liu et al. (2018) added the threshold check of Z_DR. Detailed quality control procedures are shown in the studies of Wang et al. (2018a) and Liu et al. (2018).

The methods of Liu et al. (2018) and Li et al. (2020) were used to retrieve the DSD parameters from the radar data.

The DSD is often fitted by a three-parameter gamma function:

where $ N\left(D\right) $ (mm^–1 m^–3) is the raindrop concentration of diameter $ D $ (mm), $ \mu $ is the shape, $ \gamma $ (mm^–1) is the slope, and $ {N}_{0} $ (mm^–1 m^–3) is the intercept.

Based on a large amount of 2DVD data from the Longmen Cloud Physics Field Experiment Base, China Meteorological Administration, Liu et al. (2018) found the following relationships:

when $ {Z}_{DR} $ is given, the three parameters ($ \mu $, $ \gamma $, and $ {N}_{0} $) can be determined. Then, the liquid water content $ \mathrm{L}\mathrm{W}\mathrm{C} $ (g m^–3), the nth-order moment of DSD $ {M}_{n} $, the mass-weighted mean diameter $ {D}_{m} $ (mm), and the generalized intercept parameter $ {N}_{w} $ (mm^–1 m^–3) can be calculated by the following equations:

where $ {\rho }_{\omega } $ (g cm^–3) represents the water density.

Li et al. (2020) pointed out that for pure rain, $ \mathrm{L}\mathrm{W}\mathrm{C} $ is calculated by Eq. (5), while for mixed-phase rain, $ \mathrm{L}\mathrm{W}\mathrm{C} $ and ice water content ($ \mathrm{I}\mathrm{W}\mathrm{C}, $ g m^–3) are calculated by the following equations:

where $ {Z}_{DP} $ (dB) represents the difference reflectivity, ${Z}^{{\rm{rain}}}$ (dBZ) represents the radar reflectivity of rain, ${Z}^{{\rm{ice}}}$ (dBZ) represents the radar reflectivity of ice, and $ {\rho }_{i} $ (g cm^–3) represents the ice density.

The method of Wu et al. (2018a) was used for hydrometeor classification. Based on radar parameters (e.g., Z_H, Z_DR, ρ_HV, K_DP, and SNR), hydrometeors such as dry aggregated snow, wet snow, crystal, graupel, and heavy rain were identified. Moreover, the 2DVD data of the Hailingdao (HLD) site (21.572°N, 111.859°E) were analyzed in this study. Station HLD was the nearest observation station to the typhoon’s landfall site.

In this study, to better understand the microphysical structures and processes of the TC’s precipitation, a method similar to that of Didlake and Kumjian (2017) was used to investigate the evolution of the dual-polarimetric variables such as Z_H, Z_DR, and K_DP. We also focused on the DSD parameters, such as D_m and lgN_w, and the hydrometeor types retrieved from the polarimetric radar data.

Based on the radius of the maximum wind (RMW) obtained from the Joint Typhoon Warning Center (JTWC), the distance ranges (from the TC’s center) of 0–57 km, 57–110 km, and 110–200 km were defined as the eyewall, inner core, and outer rainbands, respectively (Weatherford and Gray, 1988; Kimball and Mulekar, 2004; Abarca and Montgomery, 2015). Moreover, based on the 850–200-hPa environmental wind shear vector (VWS) from the National Centers for Environmental Prediction (NECP) operational Global Forecast System analysis data during the time span (Gao et al., 2009), the TC was divided into four quadrants: the downshear left (DL), downshear right (DR), upshear left (UL), and upshear right (UR) quadrants. The left side of the VWS is the left quadrant, while the right side of the VWS is the right quadrant. The left side of the approximate storm motion direction which is perpendicular to the VWS is the upshear quadrant, while the right side of the approximate storm motion direction is the downshear quadrant (Feng and Bell, 2019).

The time series of the logarithmic raindrop concentration [lgN(D), mm^–1 m^–3] obtained from the 2DVD data from station HLD during the period of the final landfall is shown in Fig. 2a. In addition, the distance (km) between the TC’s center and station HLD and rain rate (R, mm h^–1) is shown in Fig. 2b. Station HLD captured the characteristics of the DSD in both the eyewall and inner core.

Figure 2.  (a) Time series of the logarithmic raindrop concentration in each size bin [lgN(D), mm^–1 m^–3]. (b) The distance (km) between the TC’s center and station HLD and rain rate (R, mm h^–1). (c) Raindrop spectrum. (d) Average D_m and lgN_w. The vertical dotted black line in Fig. 2b represents the landfall time. The colored squares in Fig. 2d are the mean values of D_m and lgN_w from Feng et al. (2020, 2021) and Zheng et al. (2021).

During the period before landfall, the DSD exhibited a high concentration of small raindrops (diameter < 1 mm), with a maximum diameter of 4.9 mm, and the maximum rain rate was 58 mm h^–1 (Figs. 2a and b). As Typhoon Ewiniar (2018) was located close to station HLD, the values of lgN(D) and R decreased and reached zero. The concentration of medium raindrops (diameter of ~1–2 mm) was higher, the maximum raindrop size was smaller, and the maximum rain rate was smaller after landfall than before landfall. Figure 2c shows that compared with after landfall, the raindrop spectrum before landfall was wider and the concentration of large raindrops was higher (diameter > 3 mm). In addition, the slope of the raindrop spectrum was quite different before landfall and after landfall. The average D_m was similar before and after landfall (Fig. 2d). The average lgN_w was higher after landfall than before landfall. Moreover, the mean D_m and lgN_w values of the landfalls of other TCs in the same region (southern China) are shown in Fig. 2d, which are quite different from the mean D_m and lgN_w values obtained in this study. It is interesting that even within the same landfall region, the microphysical characteristics of the TC precipitation still exhibited great differences. This suggests that in-depth study of the microphysical structure and processes of each TC case is very important.

In general, the DSD characteristics based on the data from the 2DVD station before landfall and after landfall are quite different. However, the in situ 2DVD station has a fixed location and cannot represent the entire region. Thus, observations with high temporal and spatial resolutions, such as polarimetric radar observations, are required. To further investigate the characteristics of the DSD, the DSD parameters retrieved from the S-band polarimetric radar were analyzed.

The frequencies of occurrence of D_m and lgN_w from 0 km to 2 km in the eyewall, inner core, and outer rainbands retrieved from the S-band polarimetric radar data are shown in Fig. 3. In addition, the mean D_m and lgN_w values from 0 km to 2 km are listed in Table 1.

Figure 3.  Frequencies of the occurrence of D_m and lgN_w from 0 km to 2 km (a–c) before landfall and (d–f) after landfall; and (g–i) the difference between the after landfall and before landfall periods. The grey star represents the mean values of D_m and lgN _w from 0 km to 2 km.

Compared with the other regions, the eyewall had a narrower distribution of D_m (Fig. 3). Before landfall, from the eyewall to the outer rainbands, the mean lgN_w value increased, the mean D_m value initially decreased and then increased (Table 1), and the frequency of occurrence of a high concentration of medium raindrops (diameter of ~1–1.5 mm) initially increased and then decreased. After landfall, from the eyewall to the outer rainbands, the mean D_m value increased, the mean lgN_w value decreased, and the frequency of occurrence of a high concentration of medium raindrops (diameter of ~1–1.5 mm) decreased, especially from the eyewall to the inner core. When Typhoon Ewiniar (2018) made landfall, the mean D_m (lgN_w) value decreased (increased) in the eyewall, remained unchanged (decreased) in the inner core, and decreased (decreased) in the outer rainbands. In addition, the frequency of occurrence of a high concentration of medium raindrops (diameter of ~1–1.5 mm) increased in the eyewall and decreased in both the inner core and outer rainbands. Although the distributions of D_m and lgN_w in the eyewall, inner core, and outer rainbands did not noticeably change, they did noticeably change in the four quadrants (the relevant figure is omitted). This is an interesting feature and indicates that in-depth study of the microphysical structure in the four quadrants is essential.

The Z_H, Z_DR, and K_DP at an altitude of 2 km are shown in Fig. 4. The three time periods shown in Figs. 4a–c, d–f, and g–i represent the time periods before landfall, during landfall, and after landfall, respectively. In these three time periods, Z_H, Z_DR, and K_DP at an altitude of 2 km exhibited obvious asymmetry. The radar reflectivity at an altitude of 2 km mainly occurred in the DL and DR quadrants, while the maximum Z_H (>45 dBZ) at an altitude of 2 km mainly occurred in the eyewall and outer rainbands (Figs. 4a, d, g). When Typhoon Ewiniar (2018) made landfall, the range of the maximum Z_H (>45 dBZ) broadened, which resulted in heavy rainfall in the coastal areas of Guangdong Province.

Figure 4.  The Z_H, Z_DR, and K_DP at an altitude of 2 km at (a–c) 0800 UTC on 7 June 2018, (d–f) 1230 UTC on 7 June 2018, and (g–i) 1600 UTC on 7 June 2018. The three time periods represent the time before landfall, during landfall, and after landfall, respectively. The three circles represent the boundaries of the eyewall, inner core, and outer rainbands. DL, DR, UL, and UR denote the downshear left, downshear right, upshear left, and upshear right quadrants, respectively. The dark solid circle denotes the location of radar site YJ. The intersection point of the two lines is the TC’s center. The dark gray arrow in Fig. 4a denotes the VWS. The pale arrow in Fig. 4a denotes the approximate storm motion direction.

The Z_DR is the horizontal polarization to vertical polarization of the reflectivity factor ratio, which is closely related to the hydrometeor shape. In general, for small raindrops, the Z_DR is roughly equivalent to 0 dB. The value of Z_DR increases as the raindrop size increases. Z_DR values > 1.2 dB mainly occurred in the DL and DR quadrants; and Z_DR values < 1 dB mainly occurred in the UL and UR quadrants.

The K_DP is positive in rain which is particularly useful for rainfall estimation (Kumjian, 2013). The distribution of K_DP was consistent with that of Z_H, especially the maximum value. It is interesting that the distribution of the maximum Z_DR was completely inconsistent with those of the maximum Z_H and K_DP. This means that in this study, a high concentration of small and medium raindrops, rather than large raindrops, dominated the heavy rainfall. Moreover, the large Z_H, Z_DR, and K_DP values all occurred in the downshear quadrants, while the low Z_H, Z_DR, and K_DP values mainly occurred in the upshear quadrants, exhibiting obvious hydrometeor size sorting.

To further investigate the microphysical characteristics, the vertical profiles of the average Z_H, Z_DR, and K_DP values in the four quadrants (DL, DR, UL, and UR) in the eyewall, inner core, and outer rainbands were analyzed (Figs. 5–7). The vertical structures of the average radar parameters exhibited obvious asymmetry.

Figure 5.  Vertical profiles of the mean Z_H in the four quadrants (DL, DR, UL, and UR) in the eyewall, inner core, and outer rainbands and the average vertical profiles in the eyewall, inner core, and outer rainbands (a–d) before landfall and (e–h) after landfall.

Figure 7.  As in Fig. 5 but for K_DP.

Figure 6.  As in Fig. 5 but for Z_DR.

Before landfall (Fig. 5d), compared with the other regions, the outer rainbands had the largest mean Z_H above −10°C, while the eyewall had the largest mean Z_H below −10°C. In the eyewall (Fig. 5a), below 2 km, the DL quadrant had the largest mean Z_H, while the UR quadrant had the lowest mean Z_H. In the inner core (Fig. 5b), below 2 km, the DR quadrant had the largest mean Z_H, while the UR quadrant had the lowest mean Z_H. In the DL and DR quadrants, Z_H changed slightly with height; while in the UL and UR quadrants, it changed significantly. In addition, the same phenomenon occurred in the outer rainbands (Fig. 5c).

After landfall (Fig. 5h), compared with the other regions, the outer rainbands had the largest mean Z_H above 3 km; while the eyewall had the largest mean Z_H below 3 km. In the eyewall (Fig. 5e), the DL quadrant had the largest mean Z_H throughout the entire layer, while the UR quadrant had the lowest mean Z_H, indicating that convective precipitation dominated the DL quadrant and stratiform precipitation dominated the UR quadrant due to the VWS, which is consistent with the results of previous studies (Hence and Houze, 2011; Feng and Bell, 2019; Homeyer et al., 2021). In the inner core (Fig. 5f), below 8 km, the DL quadrant had the largest mean Z_H and the UR quadrant had the lowest mean Z_H. In the outer rainbands (Fig. 5g), below 2 km, the DL quadrant had the largest mean Z_H and the UR quadrant had the lowest mean Z_H. Moreover, in the inner core and outer rainbands, the Z_H in the DL and DR quadrants changed slightly with height; while in the UL and UR quadrants, it changed significantly. The same phenomenon occurred before landfall.

The vertical profiles of the mean Z_DR and K_DP (Figs. 6 and 7) have some features similar to those of the Z_H. During the period before landfall, Z_H, Z_DR, and K_DP in the eyewall had similar quadrant profiles due to the small region and the strong azimuthal wind, which is consistent with the results of Didlake and Kumjian (2017). Moreover, due to abundant moisture from the sea, a mass of water vapor condensed in the eyewall, resulting in rapid increases in Z_H and Z_DR (Figs. 5d and 6d). Below 2 km, the eyewall had larger Z_H, Z_DR, and K_DP values than the inner core (Figs. 5d, 6d, and 7d).

During the period after landfall, in the eyewall and inner core, the DL and DR quadrants had larger Z_H, Z_DR, and K_DP values than the UL and UR quadrants below the melting layer (Figs. 5e, f; Figs. 6e, f; and Figs. 7e, f). This means that a high concentration of large raindrops fell in the downshear quadrants and more small raindrops fell in the upshear quadrants. The hydrometeor size sorting was widespread (especially in the eyewall), which is consistent with the result of Homeyer et al. (2021). Moreover, the DL and DR quadrants had larger Z_H and smaller Z_DR values than the UL and UR quadrants above the melting layer. This means that compared with the other quadrants, the DL and DR quadrants had more graupel. In addition, falling ice phase particles (such as graupel) and the collision-coalescence growth and aggregation processes in the melting layer may have contributed to the larger Z_H and Z_DR values and lower correlation coefficient (CC) due to the variation of hydrometeor shapes, orientations, and relative permittivity (Zrnić and Ryzhkov, 1999; Kumjian, 2013; Didlake and Kumjian, 2018).

Both before landfall and after landfall, above the melting layer, the outer rainbands had larger Z_H values and lower Z_DR values than the inner core (Figs. 5d, h; Figs. 6d, h). This means that compared with the inner core, the outer rainbands had more graupel, which resulted in larger Z_H, Z_DR, and K_DP values below the melting layer. When Typhoon Ewiniar (2018) made landfall, below 2 km, the average Z_H and Z_DR values decreased in the eyewall, while the Z_H, Z_DR, and K_DP values increased in the inner core (Figs. 5d, h; Figs. 6d, h; and Figs. 7d, h). Moreover, the Z_H, Z_DR, and K_DP values increased somewhat in the DL quadrant.

In addition, some of the conclusions above are confirmed by the vertical profiles of the average D_m and lgN_w values of liquid phase particles below the melting layer (Figs. 8 and 9). Before landfall, especially in the eyewall, D_m increased rapidly, which was related to a mass of water vapor condensing and liquid water aggregating due to abundant moisture from the sea. After landfall, D_m increased slowly because the raindrops continually collected, coalesced, and broke-up due to the friction with and topography of the land. Compared with the UL and UR quadrants, the DL and DR quadrants had larger D_m and lgN_w values, which resulted in larger Z_H values (Fig. 5). In addition, the hydrometeor size sorting was obvious, which is consistent with the conclusion drawn from Figs. 5–7. Compared with the inner core, the outer rainbands in the four quadrants had larger D_m and lower lgN_w values both before landfall and after landfall.

Figure 8.  Vertical profiles of average D_m of liquid phase particles in the eyewall, inner core, and outer rainbands (a–c) before landfall and (d–f) after landfall.

Figure 9.  As in Fig. 8 but for lgN_w.

Above the melting level, deposition, riming, and aggregation of ice phase particles are the main microphysical processes in TC precipitation, while condensation, collision–coalescence, and break-up are the main microphysical processes for raindrops below the melting level (Houze, 2010). In this study, the ice-phase processes and warm rain processes were both determined to be important.

Above the melting level, the dry aggregated snow, wet snow, crystals, and graupel were present in all of the regions and quadrants (the relevant figure is omitted). Different types of hydrometeors have different sizes, concentrations, and falling velocities, and they melt into raindrops of different sizes, resulting in different rainfall intensities (Brown and Swann, 1997; Houze, 2014). In particular, graupel is important in the formation of TC precipitation. The UR quadrant had the largest proportion of graupel before landfall, while the DR quadrant had the largest proportion of graupel after landfall (Fig. 10). This is consistent with the ice water content, that is, the UR quadrant had the largest ice water content before landfall, while the DR quadrant had the largest ice water content after landfall (Fig. 11). When Typhoon Ewiniar (2018) made landfall, the ice water contents in the four quadrants (especially the UR quadrant) decreased rapidly in the eyewall. This occurred because, before landfall, a mass of moisture from the sea was transported upward, which was conducive to the ice-phase processes and the growth of ice particles. Then, after landfall, less moisture was provided by the land, which led to a lower ice water content. Moreover, in the inner core and outer rainbands, the ice water content decreased in the UL and UR quadrants and increased in the DL and DR quadrants, which is consistent with the proportion of graupel. Compared with the inner core, the outer rainbands had a larger proportion of graupel before landfall and after landfall, which is consistent with the conclusion based on Figs. 5 and 6.

Figure 10.  Proportion of graupel above the melting level in the (a) eyewall, (b) inner core, and (c) outer rainbands.

Figure 11.  Vertical profiles of the average liquid water content (g m^–3) and ice water content (g m^–3) in the eyewall, inner core, and outer rainbands (a–c) before landfall and (d–f) after landfall.

Below the melting level, raindrops and ice phase particles were present in all of the regions and quadrants. Below 2 km, the liquid water content increased rapidly because a large amount of water vapor from the sea condensed before landfall; whereas after landfall, it increased slowly due to the lower amount of water vapor provided by the land (Fig. 11). In the eyewall, the DL quadrant had the largest liquid water content both before landfall and after landfall. Due to the strong updraft in the DL quadrant (Kepert, 2001), a mass of moisture was transported upward, which was conducive to the transformation of cloud droplets into raindrops and to the growth of raindrops. This resulted in the largest liquid water content occurring in the DL quadrant. In the inner core and outer rainbands, the DR and UR quadrants had the largest liquid water contents before landfall, while the DL and DR quadrants had the largest liquid water contents after landfall. It is interesting that, after landfall, the quadrant that had the largest liquid water content was the quadrant that had the largest ice water content, but it was different before landfall. This means that compared with each other, the warm rain processes of raindrop condensation, collision, and coalescence contributed more liquid water before landfall (especially in the eyewall), and the ice-phase process of ice phase particles (such as graupel) melting into raindrops contributed more liquid water after landfall.

The 3D wind field is calculated from the Custom Editing and Display of Reduced Information in Cartesian space (CEDRIC) method based on the Zhaoqing radar and GZ radar. However, the resulting area of 3D wind field is relatively limited. The Z_H and wind field at an altitude of 2 km at 0430 UTC on 8 June 2018, which is within the period after landfall, are shown in Fig. 12. Unfortunately, during the period before landfall, we could not resolve the complete wind field of Typhoon Ewiniar (2018). As shown in Fig. 12a, the cross sections along A–B1 and A–B7 occurred in DR quadrant and the convective rainband mainly occurred in the inner core. The Z_H of the inner core at each altitude was higher than that of the eyewall, and the inner core had stronger ascending motion (between 55 km and 65 km, between low-level southwesterly and southeasterly winds) than the eyewall (Figs. 12b, c). Along with the strong ascending motion, graupel particles are found above the freezing level and there is heavy rain near the surface. Additionally, above the freezing level, ice crystals mainly occurred in the eyewall, while dry snow mainly occurred in the inner core (Figs. 12d, e). Compared with the area of A–B1, the area of A–B7 had obviously smaller ascending motion, less heavy rain, and lower Z_H values. Moreover, the area of A–B7 had no graupel particles above the freezing level. These observations suggest that graupel particles above the freezing level and heavy rain below had good correspondence with strong ascending motion (Wang et al., 2018b), and the ice-phase process played a key role in the surface precipitation.

Figure 12.  (a) The Z_H and wind field at an altitude of 2 km at 0430 UTC on 8 June 2018. The black stars represent the Zhaoqing (ZQ) radar and GZ radar. The letter “A” represents the TC’s center. Cross sections along (b), (d) A–B1 and (c), (e) A–B7 of the Z_H (second row) and hydrometeor classification (bottom row). RH, HR, RA, BD, GR, CR, WS, and DS represent mixture of rain and hail, heavy rain, light and moderate rain, big drops, graupel, crystals of various orientations, wet snow, and dry aggregated snow, respectively. The gray line is the boundary between they eyewall and inner core.

In this study, in order to understand the influence of different underlying surfaces (land/sea), the microphysical asymmetric characteristics and microphysical processes before and after the final landfall of Typhoon Ewiniar (2018) were analyzed based on 2D video disdrometer and S-band polarimetric radar data. The microphysical characteristics in the different regions (eyewall, inner core, and outer rainbands) and four quadrants (DL, DR, UL, and UR) were revealed. The main conclusions are as follows.

Both before landfall and after landfall, in the eyewall, Z_H, Z_DR, and K_DP exhibited similar quadrant profiles due to the small region and the strong azimuthal wind, which is consistent with the results of Didlake and Kumjian (2017). In addition, the outer rainbands had larger Z_H values and lower Z_DR values than the inner core above the melting layer, that is, the outer rainbands had more graupel, which resulted in larger Z_H, Z_DR, and K_DP values below the melting layer.

During the period before landfall, due to the abundant moisture from the sea, a mass of water vapor condensed and raindrops collected and coalesced, resulting in rapid increases in Z_H, K_DP, D_m, lgN_w, and the liquid water content. Moreover, due to the strong updraft in the DL quadrant (Kepert, 2001), more moisture was transported upward, which was conducive to the transformation of cloud droplets into raindrops and the growth of raindrops. This resulted in a larger liquid water content in the DL quadrant compared to the other quadrants.

During the period after landfall, the Z_H, Z_DR, D_m, and lgN_w values increased slowly, which was related to the lower amount of moisture provided by the land and the friction caused by the topography of the land. The eyewall and inner core (especially the eyewall) exhibited hydrometeor size sorting, that is, a high concentration of large raindrops fell in the DL quadrant, while more small raindrops fell in the UR quadrant. Moreover, the concentration of medium raindrops (diameter of ~1–2 mm) was higher, the maximum raindrop size was smaller, and the raindrop spectrum was narrower after landfall than before landfall.

The ice-phase processes and warm rain processes were both important in the formation of the TC precipitation. However, the comparison of the before landfall and after landfall periods revealed that the warm rain processes of raindrop condensation, collection, and coalescence contributed more liquid water before landfall (especially in the eyewall), while the ice-phase process of ice-phase particles (such as graupel) melting into raindrops contributed more liquid water after landfall. Moreover, graupel particles above the freezing level had a good correspondence with strong ascending motion (Wang et al., 2018b).

The interactions of the different TC structures with different environments and underlying surfaces result in different microphysical characteristics. Each TC case is worth studying in detail. In the future, we plan to study more TCs to obtain more useful information to improve microphysical parameterization schemes for numerical simulations.

Acknowledgements. We thank Professor Kun ZHAO, Qing LIN, and Xiaona RAO for providing the 3D wind field based on dual-doppler radar data. We thank the Longmen Cloud Physics Field Experiment Base, China Meteorological Administration for providing the 2DVD data. This research was jointly supported by Guangdong Basic and Applied Basic Research Foundation (2021A1515011415), the National Natural Science Foundation of China (Grant Nos. 42075086, 41975138, and 42005062) and the Natural Science Foundation of Guangdong Province, China (2019A1515010814).