Characteristics of Raindrop Size Distribution in Typhoon Nida (2016) before and after Landfall in Southern China from 2D Video Disdrometer Data

During the passage of Typhoon Nida, the raindrop size distribution parameters, the raindrop spectra, the shape and slope (μ–Λ) relationship, the radar reflectivity factor, and rain rate (Z–R) relationship were investigated based on a two-dimensional (2D) video disdrometer in Guangdong, China, from August 1 to 2, 2016. Due to the underlying surface difference between the ocean and land, this process was divided into two distinct periods (before landfall and after landfall). The characteristics of raindrop size distribution between the period before landfall and the period after landfall were quite distinct. The period after landfall exhibited higher concentrations of each size bin (particularly small drops) and wider raindrop spectral width than the period before landfall. Compared with the period before landfall, the period after landfall had a higher average mass-weighted mean diameter Dm that was smaller than those of other TCs from the same ocean (the Pacific). The μ–Λ relationship and Z–R relationship in this study were also compared with other TCs from the same ocean (the Pacific). This investigation of the microphysical characteristics of Typhoon Nida before landfall and after landfall may improve radar quantitative precipitation estimation (QPE) products and microphysical schemes by providing useful information.


Introduction
Tropical cyclones (TCs) usually bring heavy precipitation, strong winds, lightning strikes, and storm surge, which cause economic loss and endanger human security. Southern China is the most TC-prone area within China. An average of two to three TCs make landfall in southern China each year. is indicates the necessity of investigating TC precipitation characteristics in southern China to improve TC precipitation forecasts.
With the development of technology, an increasing number of instruments are being used to measure DSD parameters. In situ disdrometer devices, such as the laser disdrometer like OTT Parsivel and the 2D video disdrometer (2DVD), are now used extensively. e DSD characteristics of precipitation systems have been investigated using in situ disdrometers [8][9][10][11][12]. Raupach and Berne [13] indicated that methods using the DSD parameters measured by a Parsivel disdrometer should be corrected using a 2DVD as a reference instrument. Kruger and Krajewski [14] and Liu et al. [15] compared 2DVD with the Parsivel disdrometer and found that because of the orthogonality of scanning, 2DVD can effectually prevent particle superposition errors. Adirosi et al. [16] evaluated the Micro Rain Radar (MRR) performance for rainfall and DSD parameters, considering 2DVD as a primary reference.
Moreover, remote sensing instruments, such as polarization radar and satellites, are widely used to obtain a wide range of DSDs [17][18][19][20][21]. Zhang and Fu [22] examined the microphysical processes in raindrops during each stage in the life cycle of clouds by satellite (Global Precipitation Measurement (GPM) mission). Chen et al. [23] revealed the relationship between droplet growth modes and cloud particle microphysics using FY-4A satellite and GPM DPR data.
Previous studies have revealed that TCs feature higher concentrations of small and midsize drops, compared with non-TCs [24][25][26][27]. Radhakrishna and Rao [27] found that large drops in cyclonic rain were negligible, and few small to midsize drops were present in non-TC rain, possibly due to a large amount of evaporation. Deo and Walsh [28] also implied that the DSDs of TCs varied with distance from the TC center.
In addition, the DSDs of TCs vary with different rain region and rain type, which is associated with different rain formation in terms of dynamic and thermodynamic mechanisms [29,30]. TCs from different oceans were examined by Janapati et al. [31]. is study indicated that the Z-R relationship in the Indian Ocean was distinctly different from that in the Pacific. Wen et al. [25] analyzed the DSDs of TCs making landfall in China. ey found that when Λ was given, the value of μ was lower than those typhoons making landfall in the western Pacific, which were studied by Chang et al. [26]. Chang et al. [26] focused on different rain types and implied that the DSDs of TCs over the ocean were mainly maritime-like convection, while those over land fell between continental-like and maritime-like convection, as defined by Bringi et al. [4]. Other studies [31][32][33][34][35] focused on the differences in the DSD characteristics of the inner core (eyewall) and outer rainbands. Feng et al. [33] found that the concentrations of each size bin in the inner core were higher than those in the outer rainbands.
Even though many scientists have investigated the DSDs of TC systems, there have been few documented studies in southern China and still fewer focusing on the difference between DSDs before landfall and after landfall over southern China. erefore, on the basis of microphysical information provided by 2DVD, this study reported the differences between DSD characteristics in Typhoon Nida before landfall and after landfall in southern China.

2D Video Disdrometer.
e 2D Video Disdrometer (2DVD) is the third-version disdrometer manufactured by Joanneum Research in Graz, Austria, which performs the orthogonality of scanning (55 kHz) and measures the particle sizes (0.1-8.1 mm). e website (https://www. distrometer.at) provides details about the 2DVD. In this study, the temporal resolution of this data was preprocessed into 1 minute and then adopted using the quality control method of Tokay [36] to minimize the measurement errors. e processed 2DVD data have been assessed with rain gauge data by Feng et al. [37]. During Typhoon Nida's passage over Guangdong Province on August 1-2, 2016, the Longmen Field Experiment Base for Cloud Physics of the China Meteorological Administration in Guangdong, China, at the Longmen site (23.781°N, 114.248°E) obtained the observation of 2DVD. e raindrop concentration (N(D i )) from 2DVD is calculated as follows: where D i (mm) represents the raindrop diameter for the i th size bin; ΔD i (mm) represents the interval of diameter (0.2 mm); n(i) represents the number of drops for the i th size bin; V j (m s −1 ) represents the fall speed for the j th size bin drops; A represents the sampling area (10 × 10 cm 2 ); and Δt represents the sampling time (60 s).
If N(D i ) is given, the parameters such as rainfall rate R (mm h −1 ), radar reflectivity Z (mm 6 m −3 ), liquid water content LWC (g m −3 ), and the total raindrop number concentration N t (m −3 ) can be calculated as follows: e DSD expression for the n th order moment is defined as follows: e mass-weighted mean diameter D m (mm) can be calculated as follows: where L represents the total number of bins (41) and M 4 and M 3 represent the DSD for the fourth and third moments, respectively. e expression of standardized intercept parameter N w (mm −1 m −3 ) can be calculated as follows: where ρ ω (10 6 g m −3 ) represents the water density and W (g m −3 ) represents the rainwater content.

S-Band Polarimetric Radar.
e Guangzhou S-band polarimetric radar (22.962°N, 113.363°E) belongs to Guangzhou Meteorological Bureau, China Meteorological Administration, which has been upgraded from CINRAD/ SA radar. Based on the dual-polarization/double-emission and double-receiving model, the S-band polarimetric radar can achieve double-polarization parameters, which reflect the size, density, and deformation of precipitation particles in the precipitation cloud system. e S-band polarimetric radar has a resolution range of 1000 to 250 m, an elevation angle of 0.5°to 6.0°, and a temporal resolution of 6 min. In this study, the S-band polarimetric radar data have been adopted by the quality control method of Liu et al. [15] for system noise and hardware issues.

Overview of Typhoon Nida.
In the summer of 2016, Typhoon Nida caused heavy rainfall and losses to the economy of southern China because it lasted for an extended period of time and grew rapidly. Moreover, due to Typhoon Nida, 1 person died and at least 12 people in southern China were injured. For Typhoon Nida, Wu et al. [38] and Feng et al. [39] focused on the microphysical convective cells in an outer rainband and the effect of drag coefficient parameterizations on air-sea coupled simulations, respectively. In this study, the environmental condition and microphysical characteristics from the period before landfall to the period after landfall have been analyzed.
Typhoon Nida (initially classified as a tropical storm) formed over the ocean east of the Philippines at 17:00 Beijing Standard Time (BST) on July 30, 2016, and then moved to the northwest (Figure 1(a)). At 03:35 BST on August 2, 2016, Typhoon Nida (the grade of strong typhoon) made landfall in Shenzhen, Guangdong Province. e intensity of Typhoon Nida then diminished, with its grade dropping from typhoon to tropical depression, and it finally dissipated at 08: 00 BST on August 3, 2016. e maximum wind speeds (central pressure) of Typhoon Nida increased (decreased) steadily and then decreased (increased) rapidly (Figure 1(b)). When Typhoon Nida crossed over the coastal area (around 03:00 BST), the maximum wind speed reached its peak (42 m s −1 ), while the central pressure reached its minimum (960 hPa).
e National Meteorological Center (https://typhoon.nmc.cn/web.html) provided the observed track of the TC and the standard used to distinguish different grades of TC based on the maximum average wind speed of the TC center in the surface layer. According to the definition proposed by Janapati et al. [34], the time of landfall is the time when the center of the TC first lands in the coastal area. Additionally, in this study, the rain rate at a given location was considered to be influenced by the TC when the distance between that location and the TC center was <500 km [28,31,40]. us, for Longmen, the period before landfall is from 17:00 BST August 1, 2016, to 03:35 BST August 2, 2016, and that after landfall is from 03:35 BST August 2, 2016, to 18:00 BST August 2, 2016. During the period before landfall (Figure 1(c)), Longmen was in the right front quadrant of the outer rainbands. en during the period after landfall (Figure 1(d)), Typhoon Nida continued moving northwest; Longmen was in the rear side of the outer rainbands.
e reanalysis data of the National Centers for Environmental Prediction's (NECP) were used to analyze the source of water vapors. Figure 2 shows the periods before landfall and after landfall, which indicated that the moisture flux divergence from the Bay of Bengal and the South China Sea provided a sea of moisture to Typhoon Nida. From 08:00 to 14:00 BST August 1, 2016 (Figures 2(a) and 2(b)), there was no water vapor at Longmen. Subsequently, the moisture flux divergence moved northwest and strengthened. At 20:00 BST on August 1, 2016 ( Figure 2(c)), a small amount of moisture began to occur in Longmen. From 02:00 to 08:00 BST on August 2, 2016 (Figures 2(d) and 2(e)), the moisture flux divergence at Longmen enhanced rapidly and reached more than 50 g cm −1 hPa −1 s −1 . Finally, Longmen still maintained sufficient vapors with the moisture flux divergence about 25 g cm −1 hPa −1 s −1 at 14:00 ( Figure 2(f )). In brief, a larger duration and intensity of the moisture flux divergence at Longmen were found after landfall than before rainfall.
Next, we examined the sounding at Heyuan station around Longmen, which was used to analyze atmospheric stratification. Heyuan station (114.733°E, 23.804°N) is located 50 km to the northeast of Longmen. At 20:00 BST on July 31, 2016 (Figure 3(a)), before the approach of Typhoon Nida, it was dry below 300 hPa with a convective available potential energy (CAPE) value of 3434 J kg −1 , the lifting condensation level (LCL) value of 880 hPa, and the total precipitable water (TPW) value of 50 mm. en at 08:00 BST on August 1, 2016 ( Figure 3(b)), it showed unstable statues with a "horn-type" sounding that it was dry above 700 hPa, relatively wet below 700 hPa, and a CAPE value of 932 J kg −1 (the blue line and black line form a horn shape in Figure 3(b)).
is indicated that thunderstorms and hail seemed to happen at Heyuan station. en at 20:00 BST (Figure 3(c)), the dew-point temperature profile of air was close to the temperature profile of air above 500 hPa. After 12 hours (Figure 3(d)), the dew-point temperature profile of air was close to the temperature profile of air at each layer of atmosphere, and the value of CAPE was very low (6 J kg −1 ), and the value of TPW was very large (70 mm). is indicated that after landfall, the warm and moist air in the surface layer mainly resulted in the development of an unstable status of the atmosphere. Moreover, compared with the period before landfall, the period after landfall had a deep wet layer, which was in accord with the result of Figure 2. e freezing level was about 5.2 km (Figure 3(d)).
Furthermore, to gain a better understanding of the vertical microphysical structures of typhoon precipitation, we examined the occurrence frequency and averaged profiles of radar reflectivity (Z H ), differential reflectivity (Z DR ), and the differential specific (K DP ) in the vertical direction before landfall and after landfall (Figure 4). During the period before landfall, below the freezing level (0°C), the Z H was mainly between 10 and 25 dBZ (Figure 4(a)) and the Z DR was mainly between 0 and 0.5 dB (Figure 4(a)). Above the freezing level, the Z H decreasing with the increasing altitude Advances in Meteorology indicated that the supercooled water or large frozen hydrometeors were finite [19,25]. e K DP of the whole layer was concentrated around 0°/km, which indicated that the water content of the cloud was low. en during the period after landfall, below the freezing level, the Z H , Z DR, and K DP were higher than those before landfall (Figure 4(d)). is indicated that, compared with the period before landfall, the period after landfall had a high concentration of rain particles and raindrop size. Around the level of −10°C, compared with the period before landfall, the period after landfall had a similar value of Z DR and high value of K DP , which indicated a uniform raindrop size and a higher concentration of raindrop.

Characteristics of Raindrop Size Distribution.
In this study, we defined small drops (diameter < 1 mm), midsize drops (diameter between 1 and 3 mm), and large drops (diameter > 3 mm) following definitions given by Tokay et al. [36]. During the period before landfall, there was persistent precipitation from 01:00 to 03:35 BST with a peak rain rate of 29 mm h −1 , a maximum raindrop diameter of 3.5 mm, and a small drop concentration of <4.3 mm −1 m −3 ( Figure 5(a)). During the period after landfall, there was persistent heavy precipitation and the distance of the center of Typhoon Nida from the Longmen was <150 km from 04:00 to 12:00 BST August 2, 2016. In addition, the maximum rain rate (63 mm h −1 ) occurred at 07: 30     After this period, there was persistent precipitation from 11:30 to 18:00 BST on August 2, 2016, with increasing distance from the center, while rain rate exhibited several peak values, and the concentration of small drops and the maximum drop diameter increased first and then decreased. e DSD parameters before landfall and after landfall are presented in Figure 6. Compared with the period before landfall, the maximum raindrop diameter was 4.3 mm, the raindrop spectra were wider, and the number of concentration was high within each size bin in the period after landfall (Figure 6(a)), which was in accord with the result of Figure 4. From the period before landfall to that after landfall, the mean values of D m and N w increased (Figure 6(b)). Previous studies [25,31] have demonstrated that the DSDs of different TCs are distinct. Hence, we compared the D m -N w pairs with those of other TCs from the same ocean (the Pacific) ( Figure 6(b)). Our D m was smaller than the results of Bao et al. [41], Janapati et al. [31], and Zheng et al. [42], while our N w was smaller than two of the results. is is probably because the results have higher rain rates and higher concentration of large drops than Typhoon Nida in this study. e shape and slope (μ-λ) relationship was described with quadratic polynomial with one variable, and the period before landfall and the period after landfall had broadly similar coefficients (Figure 6(c)). is indicated that the shape and slope of Gamma raindrop size distribution were similar before landfall and after landfall, which was in accord with Figure 6 relationship in this study is different from those of other TCs from the same ocean (the Pacific), which is consistent with the result of Zhang et al. [43] that the μ-λ relationships in different regions or different rain types are distinct due to different microphysical process and formation mechanism. Previous studies [44][45][46][47] have investigated that the relationship between radar factor and rain rate (Z-R) is expressed by the form of Z � A * R b , where A and b vary with the raindrop size distribution and spatial scale. Appropriate coefficients (A and b) can improve the precision of local precipitation estimation. Compared with the period before landfall, the period after landfall had a small value of A and a large value of b (Figure 6(d)). e Z-R relationship is significantly distinct between before landfall and after landfall. Moreover, the Z-R relationship in this study is different from the results of Bao et al. [41] and Janapati et al. [31]. Moreover, the mean values of the above parameters (R, Z, D m , and lgN w ) before and after landfall are listed in Table 1.
To make it clear, we also examined the DSD characteristics of different rain rate (mm h −1 ) classes (first class: 0.1-1 mm h −1 , second class: 1-5 mm h −1 , third class: 5-10 mm h −1 , fourth class: 10-20 mm h −1 , fifth class: 20-30 mm h −1 , sixth class: 30-40 mm h −1 , seventh class: 40-50 mm h −1 , and eighth class: >50 mm h −1 ) before landfall and after landfall. e precipitation particles mainly consist of the drops of R <10 mm h −1 before landfall and the drops of R <20 mm h −1 after landfall (Table 1).    Advances in Meteorology e distribution of mass-weighted mean diameter D m and logarithmic normalized intercept parameter lgN w at different rain rate classes is shown in Figure 7. In the period both before landfall and after landfall, the distribution of D m and lgN w was concentrated and the value of D m and lgN w almost increased with rain intensity (Figure 7). During the period before landfall, the D m was concentrated between 0.5 mm and 2.5 mm and lgN w was concentrated between 1.9 and 4.5 mm −1 m −3 (Figure 7(a)). e maximum and minimum values of D m were the drops of R <1 mm h −1 . e maximum value of lgN w was the drop of R∼20-30 mm h −1 , while the minimum value of lgN w was the drop of R <1 mm h −1 . en during the period after landfall, the D m was concentrated between 0.5 and 2.1 mm and lgN w was concentrated between 2.6 and 4.7 mm −1 m −3 , which was more concentrated than that before landfall (Figure 7(b)). e maximum value of D m was the drop of R∼5-10 mm h −1 , while the minimum value of D m was the drop of R <1 mm h −1 . e maximum value of lgN w was the drop of R∼5-10 mm h −1 , while the minimum value of lgN w was the drop of R <1 mm h −1 . When R <20 mm h −1 , the mean value of D m (lgN w ) was almost smaller (larger) than the period before landfall. Figure 8 shows the average raindrop spectra of different rain rate (mm h −1 ) classes. During the period before landfall, the raindrops spectral width was narrower with rain intensity, while the maximum diameter (D max ) decreased with the rain intensity (Table 1). e concentration of small drops (<1 mm) increased first and then decreased gradually with rain intensity. ere were a few large drops with a concentration of 0.01-0.15 mm −1 m −3 . en during the period after landfall, the concentration of each size bin was almost higher than that before landfall in each rain rate class. e raindrops spectral width was wider first and then narrower gradually with rain intensity. ere were a few large drops with a concentration of  [41], Janapati et al. [31], and Zheng et al. [42]. Table 1: Mean of rain rate R (mm h −1 ), radar reflectivity Z (dBZ), mass-weighted mean diameter D m (mm), logarithmic normalized intercept parameter lgN w (mm −1 m −3 ), and the maximum diameter D max (mm) at different rain rate classes before and after landfall.

Segment Parameters
Rain rate classes (mm h −1 ) All classes 0.    <1.5 mm −1 m −3 . When R∼0.1-1 mm h −1 , the raindrop concentration difference between the period before landfall and after landfall was large, and particularly the small drop concentration difference was less than 3600 mm −1 m −3 . When R∼20-30 mm h −1 , the raindrop concentration difference between the period before landfall and after landfall was small (less than 600 mm −1 m −3 ).
In the period both before landfall and after landfall, the maximum contribution to R at each rain rate classes was midsize drops (1-2 mm), while the maximum contribution to N t at each rain rate classes was small drops (Figure 9). is indicated that the small and midsize drops (<2 mm) dominated typhoon precipitation in the period both before landfall and after landfall. From the period before landfall to the period after landfall, the contribution of midsize drops (1-2 mm) to R decreased when R∼1-30 mm h −1 and increased when R >30 mm h −1 , while the contribution of midsize drops (2-3 mm) to R increased at each rain rate class. is also implied that the rain rate difference between the period before landfall and the period after landfall was mainly connected with midsize drops (1-3 mm).

Summary
During the passage of Typhoon Nida, the environmental condition was analyzed from NECP reanalysis data, sounding data at Heyuan station, and S-band polarimetric radar data at Guangzhou station, and the raindrop size distribution and raindrop shape relation characteristics were investigated based on the 2DVD. e DSD characteristics of different periods and rain rate classes were revealed. A summary of the conclusions is as follows:   (1) e Bay of Bengal and the South China Sea provided a sea of moisture to the formation of the typhoon precipitation. During the period before landfall, the environment with large CAPE, high TPW, and "horn" type (the atmosphere with upper-layer dry and lower-layer wet) strengthened the development of typhoon precipitation. en during the period after landfall, the whole layer of atmosphere became wet and the CAPE was small.
(2) e DSD characteristics before landfall and after landfall were found to be different. e period after landfall exhibited higher concentrations of each size bin (particularly small drops) and wider raindrop spectral width than the period before landfall did. Compared with the period before landfall, the period after landfall had a high D m that was smaller than those of other TCs from the same ocean (the Pacific). e relationship between the radar reflectivity factor and the rain rate (Z-R) was Z � 307 R 1.12 before landfall and Z � 173 R 1.37 after landfall. Moreover, there were some similar characteristics between the period before landfall and after landfall. In the period both before landfall and after landfall, the typhoon precipitation was mainly dominated by the midsize drops (1-2 mm). e shape and slope (μ-Λ) relationship were similar before landfall and after landfall, which were different from those of other TCs from the same ocean (the Pacific).
(3) e DSD characteristics in different rain rate classes before landfall and after landfall were also quite  Figure 9: e contribution (%) of particle size classes to rain rate R (mm h −1 ) and total raindrop number concentration N t (m −3 ) at different rain rate classes.
distinct. e precipitation particles mainly consist of the drops of R <10 mm h −1 before landfall, while the precipitation particles mainly consist of the drops of R <20 mm h −1 after landfall. In the period both before landfall and after landfall, the distribution of D m and lgN w was concentrated and the value of D m and lgN w almost increased with rain intensity. e rain rate difference between the period before landfall and the period after landfall was mainly connected with midsize drops.
In this study, even though the observations before landfall and after landfall were both from the outer rainbands of the TC, the DSD characteristics before landfall and after landfall differed considerably. us, the DSDs of TCs before and after landfall over southern China warrant further investigation, and this future research has important implications for the improvement of microphysical parameterization schemes during the simulation of TCs.

Data Availability
e 2DVD data used in this study were collected from the Longmen Field Experiment Base for Cloud Physics, China Meteorological Administration. e data are available upon request to the corresponding author at xiaoh@gd121.cn.

Conflicts of Interest
e authors declare no conflicts of interest.