1 Introduction

Raindrop size distribution (DSD) plays a pivotal role in rainfall research and is used for quantitative precipitation estimation (QPE) in meteorology and hydrology as well as to understand uncertainty in cloud microphysical processes (CMPs) during climate change. Studies on DSD have been conducted in different locations worldwide (Bringi et al. 2003, 2009). The characteristics of DSDs are greatly affected by the geographical location, climate type, and seasonal variation (Tokay and Short 1996; Seela et al. 2017; Ji et al. 2018). Second, we analyzed the DSD characteristics, and calculated the normalized gamma distribution for the DSD at a mountainous site in Korea using ground-based disdrometer data. Third, we applied the results of the DSD parameters of the normalized gamma distribution to estimate the CMP as case studies in Korea and East Asia using the conceptual model of the CMPs and in situ observation DSD data.

2 Data and Methods

The PARSIVEL (OTT MESSTECHNIK GmbH & Co., Kempten, Germany) is a laser-based optical instrument that measures all types of precipitation. The laser sensor of the PARSIVEL produces a horizontal strip of light. The emitter and receiver are integrated into a single protective housing. To determine the particle speed, the duration of the signal is measured. A signal is generated immediately after a precipitation particle enters the light strip and ends when the particle has completely left the light strip. The PARSIVEL was used to collect raindrops at Daegawllyeong (site I in Fig. 1) (37°41′N, 128°45′E; at 845 m asl) in 2010 and 2011. Quality control procedures based on Cha and Yum (2021) method were applied to the PARSIVEL data used to analyze the DSD in this study as follows:

  1. 1)

    The total number of particles was more than 10 drops.

  2. 2)

    The DSD spectra for which non-zero drop counts were recorded in a series of consecutive channels were selected and spectra in which the drop diameter was < 0.56 mm were excluded from the analysis.

  3. 3)

    To remove the wind-effect error, winds with speeds of < 10 m s.−1 were considered (Friedrich et al. 2013)

  4. 4)

    To remove raindrop contamination by mixed-type precipitation, rain and snow data were selected when the temperature was warmer than 5 °C.

Fig. 1
figure 1

Map of rain drop size distribution (DSD) observation sites in East Asia. (A: Taoyuan, B: Kaohsiung, C: Naqu, D: Bei**g, E: Motuo, F: Palau, G: Busan, H: Seoul, I: Daegawllyeong)

Full size image

The DSD characteristics obtained in this study were compared with those obtained in previous studies in East Asia (Fig. 1) as case studies. The Köppen-Geiger climate classification (Peel et al. 2007) was used to define the climate at each site in East Asia. As shown in Fig. 1, A (Taoyuan) and B (Kaohsiung) have a humid subtropical climate; F (Palau) has a tropical rainforest climate; D (Bei**g) has a hot summer continental climate; C (Naqu) and E (Motuo) have a subarctic climate in a high-altitude region; H (Seoul) has a hot summer continental climate; I (Daegawllyeong) is a mountainous region with a warm summer continental climate; and G (Busan) has a dry winter and humid subtropical climate. Sites D and H are megacity regions affected by the urban climate (Hong et al. 2013, 2019; Ma et al. 2019). To determine the differences in the DSD parameters of different climate zones in Korea, the DSD parameters at site G were compared with those at site I, as site I is in a mountainous region with a warm summer continental climate. Table 1 summaries the characteristics of the DSD observational sites, including the measurement period, site information, and equipment at each observation site.

Table 1 Summary of characteristics of the rain drop size distribution (DSD) observational sites
Full size table

2.1 Calculation of Normalized Gamma Distribution for Estimating DSD Parameters

Most studies of the characteristics of DSDs employed a normalized gamma distribution-based statistical model (Testud et al. 2000; Seela et al. 2017; Ma et al. 2019). The same model was used in the present study because its parameters provide a physical meaning for DSDs.

$$N(D)={N}_{w}f(\mu ){\left(\frac{D}{{D}_{m}}\right)}^{\mu }\mathrm{exp}[-\left(4+\mu \right)\frac{D}{{D}_{m}}]$$
(1)

where D is the drop diameter (mm), \(N(D)\) (m−3 mm−1) is the number of drops per unit volume per unit size interval, \({N}_{w}\) (m−3 mm−1) is the normalized intercept parameter, \(\upmu\) (unitless) is the shape parameter, and \({D}_{m}\)(mm) is the mass-weighted mean drop diameter. \({D}_{m}\) is used to describe the DSD and is estimated using the following equation:

$${D}_{m}=\frac{{\int }_{0}^{\infty }N\left(D\right){D}^{4}dD}{{\int }_{0}^{\infty }N\left(D\right){D}^{3}dD}$$
(2)

To estimate CMPs, as reported by Dolan et al. (2018), the median volume diameter (\({D}_{o},\) mm) was calculated using the following equation:

$${D}_{0}= \frac{3.67+\mu }{4+\mu }{D}_{m}$$
(3)

Two parameters are considered when characterizing rain intensity: the liquid water content (LWC, g m−3) and rainfall rate (R, mm h−1) (Testud et al. 2000). Testud et al. (2000) found that the LWC has a clearer physical meaning compared with R (as altitude increases, R is subject to vertical air motion and its terminal fall velocity changes with air density).

$$LWC=\frac{\pi {\rho }_{\omega }}{6}{\int }_{{D}_{min}}^{{D}_{max}}{D}^{3}N\left(D\right)d(D)$$
(4)

\({N}_{w}\) (m−3 mm−1) is estimated using the following equation:

$${N}_{w}= \frac{{4}^{4}LWC}{\pi {\rho }_{\omega }{D}_{m}^{4}}$$
(5)

where, ρw is the water density.

Moment-based DSD estimation of meteorological products is a general method used in radar meteorology. An optical disdrometer is used for the products. The nth-order moment is determined as:

$${M}_{n}={\int }_{{D}_{min}}^{{D}_{max}}{D}^{n}N\left(D\right)dD=\sum \nolimits_{{{D}_{i}=D}_{min}}^{{D}_{max}}{D}_{i}^{n}N({D}_{i})\Delta {D}_{i}$$
(6)

The parameters use to describe the DSD were calculated as follows:

$${D}_{m}= \frac{{M}_{4}}{{M}_{3}}$$
(7)
$${N}_{w}= \frac{{4}^{4}LWC}{\pi {\rho }_{\omega }{\left(\frac{{M}_{4}}{{M}_{3}}\right)}^{4}}$$
(8)
$$\mu =\frac{(3{M}_{4}{M}_{2}-4{M}_{3}^{2})}{{M}_{3}^{2}-{M}_{4}{M}_{2}}$$
(9)

2.2 Conceptual Model of DSD Observation-Based Microphysical Process

To identify the cause of changes in the DSD parameters, the CMPs were estimated using the conceptual model (DCPM) of the DSD observation-based microphysical process suggested by Dolan et al. (2018). DSD parameters were classified into six groups using PCA, and vertical observation data from the dual-polarization radar were compared with the DSD parameters of the six groups to generate the DCPM. Dolan et al. (2018) described a DCPM based on group determination in log10Nw–D0 space (Fig. 12 in Dolan et al. (2018)). Groups 1, 3, 5, and 6 are characterized by convective precipitation processes, where particles are assumed to increase in size as they are lifted by a convective updraft; after sufficient growth, the particles fall back through the updraft and grow to even larger sizes. Groups 2 and 4 are stratiform precipitation processes, with increasing D0 and decreasing log10Nw correlated with a bright band intensity and representing an evolution from melted vapor-grown particles to aggregation and riming processes. Especially, the light rain rates and small drop diameter associate with group 2 by vapor deposition. We used the DCPM to identify the dominant mechanisms in the six groups objectively determined from the surface disdrometers using PCA in log10Nw–D0 space.

2.2.1 Results and Discussion

2.3 DSD Parameter Characteristics in Korea

To estimate the DSD parameters at site I, PARSIVEL data from Daegawllyeong (site I in Fig. 1) collected in 2010–2011 were used. The total observation time of rain events was 1,303 h. To evaluate the quality of the PARSIVEL data, we compared the rain rate of the PARSIVEL with that of the Tip** Bucket (TB) rain gauge at site I; the correlation coefficient was calculated as 0.91 (Fig. 2). The standard deviations of the collected observational data using the PARSIVEL and TB rain gauge were 2.07 and 1.93 mm h−1, respectively. The correlation coefficient calculated by You and Lee (2015) in Korea was 0.99 between the PARSIVEL and TB rain gauge. Our results and those of previous studies in Korea show that PARSIVEL is suitable for studying the DSD.

Fig. 2
figure 2

Comparison of the rain rate determined using a particle size velocity disdrometer (PARSIVEL) with that obtained using the tip** bucket (TB) rain gauge from hourly data at site I (Daegwallyeong) from 2010 to 2011 after data quality control (Cha and Yum 2021)

Full size image

As shown in Fig. 3a, Dm decreases with an increase in μ, indicating that the breadth of the DSD decreases when the number of small raindrops increases. Dm and log10Nw exhibit an inverse relationship (Fig. 3b). At the same water content, the total drop concentration decreases when the mean volume weighted diameter size increases. The total drop concentration at the same water content increases with increasing μ (Fig. 3c). Most previous studies showed similar results for the relationship of Dm, log10Nw, and μ (Chen et al. 2016; Ma et al. 2019; Suh et al. 2021; Wang et al. 2021). Thus, the PARSIVEL data were of suitable. Figure 4a, 5a, and 6a show the relative frequencies (%) of the parameters at site I. The mean (standard deviation) values of Dm, μ, and log10Nw were estimated as 1.40 (± 0.43) mm, 3.71 (± 3.46), and 3.62 (± 0.60) mm−1 m−3. The values of the parameters calculated in this study showed similar ranges as those reported previously (You et al. 2004; You and Lee 2015; Bang et al. 2017).

Fig. 3
figure 3

a Comparison with a mass-weighted mean drop diameter (Dm) and shape (μ) parameter of raindrop size distribution at site I (Daegwallyeong) in Korea from 2010 to 2011. b Same as Fig. 3a except a normalized intercept (Nw) and mass-weighted mean drop diameter (Dm) parameter of raindrop size distribution. c Same as Fig. 3a except a normalized intercept (Nw) and shape (μ) parameter of raindrop size distribution

Full size image
Fig. 4
figure 4

Relative histograms of (a) a mass-weighted mean drop diameter (Dm) parameter of rain drop size distribution at site I (Daegwallyeong) from 2010 to 2011 and (b) Dm of You and Lee (2015) at site G (Busan) in 2012 (Avg: mean; Std: standard deviation)

Full size image
Fig. 5
figure 5

Same as Fig. 4 except a shape (μ) parameter of rain drop size distribution (Avg: mean; Std: standard deviation)

Full size image
Fig. 6
figure 6

Same as Fig. 4 except a normalized intercept (Nw) parameter of rain drop size distribution (Avg: mean; Std: standard deviation)

Full size image

Figure 4a shows the relative frequency histograms of Dm at site I. You and Lee (2015) studied DSD parameters at site G using a Precipitation Occurrence Sensor System (POSS), which is a low-power X-band bi-static radar system, in 2002 and a PARSIVEL in 2012. We compared the DSD parameters obtained in this study with the PARSIVEL DSD data collected at site G in 2012 by You and Lee (2015). The mean (standard deviation) value of Dm at site G was 1.45 (± 0.40) mm (Fig. 4b) and mean (standard deviation) value of Dm at site I calculated in this study was 1.40 (± 0.43) mm (Fig. 4a). The maximum values of Dm were approximately 3.8 and 2.6 mm at sites G and I, respectively. The average precipitation particle size was larger in coastal areas than in mountainous areas in Korea (Cha and Yum 2021). As shown in Fig. 4a and b, the relative frequency of Dm at site I in the mountainous region had a wider distribution than that at site G in the coastal region because of differences in the CMPs of precipitation drop generation. Figure 5 shows a histogram of μ, the shape of the DSD at both sites. As shown in Fig. 5a and b, the mean μ (standard deviation) values ​​at sites I and G were 3.71 (± 3.46) and 5.99 (± 3.75), respectively. A larger μ value indicates a wider DSD and vice versa (Testud et al. 2000). In Korea, continental mountain clouds at site I tend to generate a large number of small particles, and oceanic clouds at site G generate larger rain drops (Cha and Yum 2021). Figure 6a and b show a histogram of the relative frequencies of log10Nw at sites I and G. The mean (standard deviation) values of log10Nw at site I and G were 3.62 (± 0.60) and 3.51 (± 0.45) mm−1 m−3, respectively.

2.3.1 DSD Parameter Characteristics in East Asia

The mean (standard deviation) values of Dm, log10Nw, and μ in East Asian regions were investigated based on ground observations (Table 2). The DSD parameters at site I were evaluated as precipitation parameters in the mountainous region in the hot summer humid continental climate. The values of Dm (1.40 (± 0.43) mm) and log10Nw (3.62 (± 0.60) m−3 mm−1) at site I were smaller than those at site H; Dm = 1.54 (± 0.32) mm and log10Nw = 3.98 (± 0.42). That is, mountainous terrain near site I forces airflow to ascend and then initiates the formation of precipitation, as described by Cosma et al. (2002). Therefore, there was a much larger number of smaller particles at site I than at site G; additionally, these particles were smaller at site Dm than at site H (Seoul). The value of μ at site G [5.99 (± 3.75)] was higher than that at site I [3.71 (± 3.46)] because the convective rainfall at site G increases the value of μ (You and Lee 2015). The value of μ at site A [6.72 (± 7.11)] and at site B [6.93 (± 7.93)] at the Taiwan observation sites were two-fold higher than those at site I. The breadth of the DSD at sites A and B was smaller than that at site I. If the value of μ is > 0, the DSD curve is concave downward; if the value of μ is < 0, the DSD curve is concave upward; and if it is equal to 0, the DSD is exponential (Ulbrich 2001). The value of μ at site F [8.37 (± 8.94)] was approximately two-fold higher than that at site I, [3.71 (± 3.46)]. The breadth of the DSD was narrow because of the convective rainfall in the tropical rainforest climate. Similarly, for sites C and E, which are at high-altitude regions in China, the values of μ were approximately three-fold lower than that at site I because of the higher total aerosol concentration, resulting in a wider breadth of DSD in the high-altitude regions with a continental climate. Although the total aerosol concentration is small in marine climates, many particles can act as cloud condensation nuclei (Kim et al. 2012). Therefore, the DSD varies widely, and the value of μ was relatively low at sites C and E.

Table 2 Mean (standard deviation) rain drop size distribution (DSD) parameter values of Dm, log10Nw, and μ in East Asia; Dm is a mass-weighted mean drop diameter, μ is a shape, and Nw is a normalized intercept parameter of the DSD
Full size table

2.4 Comparison of DSD Parameters in Different Disdrometers

Figure 7 shows a comparison of the parameters of the DSD statistical model based on the rainfall intensity data obtained using different disdrometers, including PARSIVEL. The rainfall intensity was classified based on the rain rate into the following six categories: 1) 0.1–1 mm h−1 (very light rain); 2) 1–2 mm h−1 (light rain); 3) 2–5 mm h−1 (moderate rain); 4) 5–10 mm h−1 (heavy rain); 5) 10–20 mm h−1 (very heavy rain); and 6) > 20 mm h−1 (extreme rain). To compare the results using different equipment, the two-dimensional video disdrometer (2DVD) and Joss Waldrogel Disdrometer (JWD) observation data reported by Tokay et al. (2001) and POSS observation data reported by You et al. (2004) were used. Comparison of Dm by rainfall intensity measured using different instruments (Fig. 7a) revealed similar trends in which Dm increased with increasing rainfall intensity. The difference between the 2DVD and other instruments tended to decrease with increasing rainfall intensity, except that the 2DVD showed the highest Dm value in very light rain. As observed using the POSS, a sudden decrease in extreme rainfall intensity occurred because of the attenuation effect of strong precipitation (Park et al. 2005; Kestwal et al. 2014). Figure 7b shows the μ values for different instruments. The μ value of POSS is relatively low and that of 2DVD is high, ​​except for extreme rainfall intensity. The mean μ values of PARSIVEL, JWD, and 2DVD are similar. As shown in Fig. 7c, the mean log10Nw value of 2DVD is high and that of POSS is low. The JWD and PARSIVEL showed similar log10Nw values with a similar trend according to the change in rainfall intensity for all instruments, except for very light and light rain.

Fig. 7
figure 7

a Comparison of mass-weighted mean drop diameter (Dm) parameters of rain drop size distribution for different rainfall intensities observed using different instruments (PARSIVEL: particle size velocity disdrometer, 2DVD: 2-dimensional video disdrometer, and POSS: Precipitation Occurrence Sensor System disdrometer). The rainfall intensities were divided into six rain rate categories: 1) 0.1–1 mm h−1 (very light); 2) 1–2 mm h−1 (light); 3) 2–5 mm h−1 (moderate); 4) 5–10 mm h−1 (heavy); 5) 10–20 mm h−1 (very heavy); and 6) > 20 mm h−1 (extreme). b Same as Fig. 7a except the shape (μ) parameters of the rain drop size distribution. c Same as Fig. 7a except the normalized intercept (Nw) parameters of the rain drop size distribution

Full size image

2.5 Case Studies for Estimating CMPs in Korea

The CMPs at site I (Daegawllyeong) in Korea were estimated using the DCPM based on the calculated DSD parameters (Fig. 8). To conveniently explain the CMPs used by Dolan et al. (2018), vapor deposition is represented as VD, weak convection as WC, collision and coalescence as CC, ice-based and warm rain growth as ICWRG, ice-based growth as ICG, and aggregation and riming as AR. Figure 8 shows the results of applying D0 and log10Nw, observed in this study, to DCPM. The classification ranges of these values were approximated based on DCPM to understand the CMPs. As shown in Fig. 8, the convective precipitation is a more dominant mechanism than stratiform precipitation at site I in Korea. Estimates of the major CMPs at site I were 30.3% for AR, 23% for ICWRG, 12.7% for WC, 11.6% for ‘mixed ICG and AR’, and 8.5% for ‘mixed ICWRG and ICG’, as shown in Table 3. The VD were classified as having little effect at site I. Analysis of CMPs at site I showed that most of cold cloud formation processes, such as AR and ICWRG, contributed to precipitation. In further studies, the classification of CMPs at site I should be improved using more objective and specific statistics methods, as the overlap** CMPs in Fig. 8 are unclear.

Fig. 8
figure 8

Estimation of dominant cloud microphysical processes using a conception model described by Dolan et al. (2018) at site I (Daegwallyeong) in Korea from 2010 to 2011; Do is the median volume diameter and Nw is the normalized intercept parameter of the rain drop size distribution

Full size image
Table 3 Percentage of estimation of dominant cloud microphysical processes using a conception model by Dolan et al. (2018) at site I (Daegwallyeong) from 2010 to 2011 in Fig. 8
Full size table

2.6 Case Studies for Estimating CMPs in East Asia

To estimate the CMPs at other sites in East Asia as case studies, we selected various sites and analyzed the DSD parameter data from previous DSD studies. We could not analyze all CMPs at the selected sites because some DSD parameter data from previous DSD studies were unavailable. Therefore, we approximated one major CMP for case studies using mean DSD parameters such as D0 and log10Nw for rainfall intensities reported previously.

Figure 9 and Table 4 show the results of classification of the major CMPs based on rainfall intensity at sites A (Taoyuan) and B (Kaohsiung). At site A, the one major precipitation-generating CMP was a VD of < 1 mm h−1 (A1-A4) and ‘mixed WC, CC, and ICWRG’ of > 2 mm h−1(A6-A12). Using the criteria described by Janapati et al. (2020), the major CMP at site B was ambiguous at < 1 mm h−1(B1), as observed at site A. Thus, a large amount of water vapor leads to cloud growth and rain via VD of < 1 mm h−1 (A1-A4), and many complex CMPs affect the precipitation generation at sites A and B in humid subtropical climates.

Fig. 9
figure 9

Case study to estimate major cloud microphysical processes using the conception model described by Dolan et al. (2018) at sites A and B; VD (red), ambiguous (brown), and mixed WC + CC + ICWRG (blue). Diamonds and dots represent sites A and B, respectively. The ranges of the rain rate (mm h−1) at A1–A11 and B1–B6 are shown in Table 4 (VD: vapor deposition, WC: weak convection, CC: collision and coalescence, and ICWRG: ice-based and warm rain growth)

Full size image
Table 4 Case study of mean drop size distribution (DSD) parameter values of Do and log10Nw by rain rate in Taiwan in East Asia; Do is the median volume diameter and Nw is the normalized intercept parameter of the DSD
Full size table

Figure 10 and Table 5 show the major CMPs by rainfall intensity in a subarctic climate in high-altitude regions, where the average elevations of sites C (Naqu) and E (Motuo) are 4,508 and 1,200 m, respectively. The major CMP in these alpine climates was caused by VD < 5 mm h−1 (C1-C3 and E1-E3) and by ICWRG > 5 mm h−1 (C4-C5 and E4-E5). At site D (Bei**g, China), VD was the only major CMP at approximately ≤ 1 mm h−1 (D1-D2), and the ICWRG was dominant above this level (D3-D8). At sites C and E, VD was the only major CMP at < 5 mm h−1(C1-C3 and E1-E3). Therefore, the VD is the most important CMP < 5 mm h−1 at sites C and E based on the presence of numerous supersaturated clouds in alpine climate regions (Hammer et al. 2013). Thus, alpine regions with higher elevation have lower the temperatures, where the air is likely to be oversaturated.

Fig. 10
figure 10

Same as Fig. 9 except at sites C, D, and E; VD (red) and ICWRG (blue). Squares, diamonds, and dots represent sites C, D, and E, respectively. The ranges of the rain rate (mm h−1) at C1–C5, D1–D8, and E1–E5 are shown in Table 5 (VD: vapor deposition and ICWRG: ice-based and warm rain growth)

Full size image
Table 5 Case study of mean drop size distribution (DSD) parameter values of Do and log10Nw by rain rate in China in East Asia; Do is the median volume diameter and Nw is the normalized intercept parameter of the DSD
Full size table

Figure 11 and Table 6 show the major CMPs based on rainfall intensity at site F (Palau), a tropical rainforest region. Precipitation in this region is mainly formed through CMPs such as WC, CC, and ICWRG. The results suggest that WC and CC are the major precipitation-generating microphysical processes at site F because of the influence of the tropical climate. The major CMPs by rainfall intensity at site F were WC at 0.1–0.7 mm h−1(F1-F3), ‘mixed WC, CC, and ICWRG’ at 0.7–5 mm h−1 (F4-F6), ‘mixed WC and CC’ at 5–18 mm h−1 (F7-F9), CC at 18–40 mm h−1 (F10-F11), and ICWRG at > 40 mm h−1 (F12). These results demonstrate that the collision-coalescence process is the primary mechanism responsible for producing precipitation and that heavy rain is mainly generated by ice-based and warm rain growth in climates with tropical rainfall.

Fig. 11
figure 11

Same as Fig. 9 except at site F; WC (red), mixed WC + CC + ICWRG (blue), mixed WC + CC (gray), CC (green), and ICWRG (black). The range of the rain rate (mm h−1) at F1–F12 is shown in Table 6 (WC: weak convection, CC: collision and coalescence, and ICWRG: ice-based and warm rain growth)

Full size image
Table 6 Case study of mean drop size distribution (DSD) parameter values of Do and log10Nw by rain rate in Palau in East Asia; Do is the median volume diameter and Nw is the normalized intercept parameter of the DSD
Full size table

Figure 12 and Table 7 show the major CMPs based on rainfall intensity at site H (Seoul), G (Busan), and I (Daegawllyeong) in Korea. At mountainous site I in a warm summer continental climate, ‘mixed ICG and ICWRG’ was the major CMP at > 1 mm h−1 (I3-I5), whereas the CMP was ambiguous at < 1 mm h−1 (I1-I2). The megacity’s site H, Seoul in Korea, in a hot summer continental climate case such as site D, Bei**g in China, was mainly affected by VD at < 5 mm h−1 (H1-H4) and ICWRG at > 5 mm h−1 (H5-H8). The major CMP was ICWRG at > 5 mm h−1 (G2-G4) and ambiguous at < 5 mm h−1 (G1) at the harbor city’s site G (Busan) in dry winter humid subtropical climates. In particular, the major CMP of megacities was vapor deposition at site D in Bei**g (< 1 mm h−1 (D1-D2)) and site H in Seoul (< 5 mm h−1(H1-H4)) because the strong updraft of the urban heat island effect in megacities increased the LWC (Han et al. 2012), with many supersaturated clouds formed at higher altitudes.

Fig. 12
figure 12

Same as Fig. 9 except at sites G, H, and I; VD (red), mixed ICG + ICWRG (purple), ICWRG (blue), and ambiguous (brown). Dots, diamonds, and squares represent sites G, H, and I, respectively. The ranges of rain rate (mm h−1), G1–G4, H1–H6, and I1–I6, are shown in Table 7 (VD: vapor deposition, ICWRG: ice-based and warm rain growth, and ICG: ice-based growth)

Full size image
Table 7 Case study of mean drop size distribution (DSD) parameter values of Do and log10Nw by rain rate in South Korea in East Asia; Do is the median volume diameter and Nw is the normalized intercept parameter of the DSD
Full size table

3 Summary and Conclusion

We analyzed the DSD observation data using PARSIVEL in the mountainous regions of Korea in 2010 and 2011. We also determined the characteristics of DSD parameters based on rainfall intensity observed in previous DSD studies by comparing the PARSIVEL data used in this study with those obtained using other disdrometer equipment. We developed a simple method for understanding the CMPs of precipitation generation using a conceptual model of CMPs and in situ observation DSD data. We used observed DSD parameters and the CMP conceptual model of DSD observation-based microphysical processes suggested by Dolan et al. (2018). As case studies, we applied DSD observation data from Korea and East Asia to estimate the CMP.

At sites A (Tauyuan) and B (Kaohsiung) in a humid subtropical climate region, the major CMPs were vapor deposition at less than a light rain rate; mixed weak convection, collision and coalescence; and ice-based and warm rain growth at more than moderate rain rates. At site F Palau) in a tropical rainforest climate region, the major CMPs were weak convection, collision and coalescence, ice-based and warm rain growth, and their mixed CMPs. Thus, a larger number of CMPs at site F (Palau) are involved in generating precipitation compared to those at other sites. At sites C (Naqu) and E(Motuo) in high-altitude regions with a subarctic climate, vapor deposition was the major CMP at < 5 mm h−1 because of the presence of many supersaturated clouds in alpine climate regions. Sites D (Bei**g) and H (Seoul) have a hot summer continental climate and are megacity sites. The major CMPs at sites D (Bei**g) and H (Seoul) were similar to those of the subarctic climate in high-altitude regions in East Asia, such as vapor deposition and ice-based and warm rain growth. Particularly, vapor deposition is the major CMP in megacities such as Bei**g and Seoul because the strong updraft of the urban heat island effect in these regions increases the LWC, after which large numbers of supersaturated clouds form at higher altitudes. Ice-based growth is an important CMP at site I (Daegawllyeong), which is in a mountainous area. The major CMP at site G (Busan) is ice-based and warm rain growth, as a maritime cloud affects this site in dry winter humid subtropical climate regions.

Our case studies revealed that CMPs can be estimated using DSD observation data in different climate regions and can improve the classification of CMPs using new techniques such as machine and deep learning. In our further studies, we will apply the conceptual model of CMPs by Dolan et al. (2018) to vertical rain radar data, such as the Micro Rain Radar data, to calculate vertical DSD parameters. These calculated parameters will be compared with ground-based DSD observation data to evaluate the accuracy of the vertical DSD distribution, and the estimated cloud microphysical model data will be compared with S-band dual-polarization radar observation data in Korea. These studies may improve the estimation accuracy of radar-based precipitation and evaluation of precipitation-generating microphysical processes by using an objective numerical weather model.