Diurnal Variation Characteristics of Raindrop Size Distribution Observed by a Parsivel 2 Disdrometer in the Ili River Valley

Te diurnal variation characteristics of raindrop size distribution (RSD) in the Ili River Valley are investigated in this study, using the RSD data from May to September during 2020-2021 collected by a Parsivel 2 disdrometer in Zhaosu. Signifcant diurnal variations (02–07, 08–13, 14–19, and 20-01 local standard time (LST)) of precipitation and RSD in Zhaosu are revealed during the rainy seasons. Precipitation mainly occurs in the late afternoon and early evening. A higher concentration of small raindrops is observed in the morning, whereas more mid-size and large raindrops are observed in the afternoon. Te RSD exhibits diurnal diferences between diferent rainfall rate classes; the diurnal diference of RSD is more pronounced in the case of high rainfall rates. Stratiform precipitation can occur at any time of the day, yet convective precipitation mainly occurs during the late afternoon and early evening. Te RSD of stratiform rainfall shows a similar distribution over the four time periods. For convective rainfall, the concentration of small raindrops is the highest (lowest) over 02–07 (14–19) LST, while the highest (lowest) concentration of medium and large drops is observed over 14–19 (02–07) LST. Convective rain in the Ili River Valley over 14–19 LST can be characterized as the continental convective cluster, while in the rest time of the day, it is neither in the maritime cluster nor in the continental cluster. Te empirical relationships between the radar refectivity factor and rainfall rate ( Z - R ) for stratiform and convective rain types are also derived. Te purpose of this study is to advance our understanding of precipitation microphysics in arid mountainous region.

Many studies have shown that there are obvious temporal and spatial variations in RSD characteristics in terms of rainfall rate and type due to diferences in climate background, geographical location, and atmospheric condition [22][23][24][25][26][27][28][29].Based on observations of two diferent disdrometers and radar retrievals, Bringi et al. [1] analyzed RSD characteristics under various climate regimes and found that there are signifcant sea-land dissimilarities in convective rainfall.Te continental-like cluster is characterized by a higher mass-weighted mean diameter, while the maritime-like cluster is characterized by a lower mass-weighted mean diameter.Dolan et al. [30] employed principal component analysis (PCA) to reveal comprehensive modes of spatial and temporal variations of RSD spanning from the deep tropics to the high latitudes.Tey conceived that the physical processes responsible for shaping the RSD appear to vary as a function of latitude.Marzuki et al. [31] studied the regional variability of RSD along the Equator according to the data collected at four stations in Indonesia.Te results indicate that the regional variability of RSD is highly associated with oceanic and continental systems, topographic conditions, and horizontal scale of landmass.Wen et al. [32] investigated seasonal variations of rainfall properties in East China by using a two-dimensional video disdrometer and a vertically pointing micro rain radar.Tey concluded that summer rainfall is dominated by convective rain, while winter rainfall is completely composed of stratiform rain.
Diurnal variation is the most fundamental cycle of change driven by solar radiation in the Earth's climate system.So far, numerous studies have been conducted on the diurnal variation of the rain spectra worldwide [33][34][35][36][37]. Kozu et al. [33] examined the diurnal and seasonal RSD variations in the maritime continental (Kototabang and Singapore) and inland (Gadanki) stations in the Asian monsoon region.Tey found that the most pronounced diurnal variation of RSD occurs at Kototabang, which is related to the fact that Kototabang is afected by orographic efect and land-sea thermodynamic diferences.Chen et al. [34] studied the diurnal variation of RSD in summer over the Tibetan Plateau.Te results indicated that convective rain has large day-night diferences, while stratiform rain shows minimal diferences.As reported in previous studies, there are distinct diferences between the raindrop spectra of summer and winter rainfall over north Taiwan, and both exhibit distinct diurnal variations [35].RSD in Busan also demonstrates diurnal and seasonal variations [38,39].However, few studies have focused on the diurnal variation of RSD in semiarid area of China.
Xinjiang is located in the northwest of China, which is a typical arid and semiarid area.Te precipitation characteristics of Xinjiang are signifcantly diferent to that in the eastern monsoon region of China [40,41].Te Ili River Valley is located in the northwest of Xinjiang, where the annual average precipitation is 417.6 mm, which makes it the most humid area in Xinjiang.Te Ili River Valley is 360 km long from east to west and 275 km wide from north to south, surrounded by mountains to the north, east, and south with an open "trumpet-shaped" topography to the west.Because of its special geographical location, precipitation in the Ili River Valley presents remarkable diurnal variation.Although microphysical properties of precipitation have been studied in this area, researches about the diurnal variation of RSD are still scarce [42][43][44].In this paper, the continuous Parsivel 2 disdrometer measurements from 2020 to 2021 in Zhaosu are used to conduct a comprehensive study on the diurnal variation of RSD and rainfall integral parameters in the Ili River Valley during rainy seasons.Te present study aims to advance our understanding of rainfall microphysical processes in arid mountainous region.
Te manuscript is organized as follows.Section 2 describes the instruments, data, and methodology used in this study.Te diurnal variations of RSD and rainfall integral parameters for diferent rainfall rates and rain types are detailed in Section 3. Discussion is provided in Section 4. Summary and conclusions are provided in Section 5.

Data and Methodology
2.1.Instrument and Data.Te RSD data used in this study are collected by a Parsivel 2 disdrometer manufactured by OTT Messtechnik, Germany [45].Te instrument can emit a laser beam with a sampling area of 54 cm 2 and a time interval of 60 s.Te particle diameter is determined by the maximum attenuation of the signal, and the fall velocity is estimated from the transit time of the particles within the laser beam.Te particles are subdivided into 32 nonequidistant velocity bins ranging from 0.05 to 20.8 m•s − 1 and 32 nonequidistant size bins ranging from 0.062 to 24.5 mm.As an upgraded version of the frst-generation disdrometer Parsivel, the Parsivel 2 disdrometer uses a more expensive laser sensor for the raindrop size and rainfall measurements [46,47].
Tis work utilizes two years of disdrometer data during rainy seasons in the Ili River Valley (2020-2021).Te instrument is deployed at the meteorological station in Zhaosu (81.09 °E, 43.08 °N, 1851 m a.m.s.l.).Zhaosu is located in the southwest of the Ili River Valley.Te rainfall in the Ili River Valley during rainy seasons is about 200 to 250 mm, and the temperature is usually maintained at 17 ∼ 23 °C.Easterly and westerly winds occur most frequently in the Ili River Valley, and the wind direction is consistent with the direction of the valley.Te synoptic systems afecting the Ili River Valley during rainy seasons are mainly Central Asia low vortex (trough), and the mesoscale systems are mainly mesoscale shear line, mesoscale low pressure, and convective cells.Te location of the observation site and the topography of the Ili River Valley are shown in Figure 1.In addition, automatic weather station (AWS) data and ERA5 reanalysis data are also used.AWS can record surface meteorological parameters (temperature, relative humidity, and wind speed) at 1 h sampling intervals.Convective available potential energy (CAPE) is obtained from the ERA5 hourly data on single levels with a horizontal resolution of 0.25 °× 0.25 °.Based on the above data, this study attempts to illuminate the microphysical and thermodynamic characteristics in the Ili River Valley.

RSD.
Based on disdrometer data, the measured RSD can be calculated as follows: where From the raindrop concentration N(D i ), the integral rainfall parameters including rainfall rate R (mm•h − 1 ), radar refectivity factor Z (mm 6 •m − 3 ), liquid water content W (g•m − 3 ), and total number concentration N t (m − 3 ) can be calculated by the following equation: where ρ w (g•cm − 3 ) is the density of water.
Te three-parameter gamma distribution proposed by Ulbrich [48] has been proven to be a suitable representation of the raindrop spectra and is described as follows: where N 0 (mm − 1− μ m − 3 ), μ (dimensionless), and Λ (mm − 1 ) are the intercept, shape, and slope parameters of the gamma distribution, respectively.Te three parameters in equation ( 4) can be obtained by using the method of moments [49].
Te n-th moment of the RSD is expressed as follows: where the third, fourth, and sixth moments of the size distribution are considered in this study.Equations ( 6), (7), and (8) are the three parameters of gamma distribution (μ, Λ, and N 0 ).
where Γ(x) is the complete gamma function and is described as follows: Te normalized gamma distribution has been widely applied to analyze the properties of RSD and the relationships between rainfall integral parameters [50].In particular, the most important advantage of the normalized gamma distribution is that it is free of any assumption about the shape of the RSD or the relationships between moments of the RSD.In the meantime, it allows observational facts to decide what this shape is and what these relationships are.Here, the model is defned as follows [10]: where D m (mm) and N w (mm − 1 •m − 3 ) are the mass-weighted average diameter and the normalized intercept parameter, respectively.D m , N w , and f(μ) are given by the following equation: 2.3.Data Quality Control.Apart from the inherent limitations of instrument, several quality control (QC) procedures have been applied to the 1-minute disdrometer data to minimize the observation errors.First of all, the lowest two size bins are left empty because of low signal-to-noise ratio.Following Tokay et al. [46], samples with a total number of raindrops smaller than 10 or a disdrometer-derived rainfall rate less than 0.1 mm•h − 1 are discarded as noise.To focus on rainfall, all samples related to solid precipitation (hail, snow, etc.) are excluded, and raindrops with a diameter larger than 8 mm are also eliminated since they are unrealistic and probably breakup near the ground.As noted by Yuter et al. [51] and Friedrich et al. [52], margin drops caused by particles partially falling within the laser beam as well as unrealistically large and slow falling particles produced by strong winds and splashing efect should be treated as spurious raindrops.In order to further guarantee the reliability of the disdrometer data, raindrops outside ±60% of the empirical fall velocity-diameter relationship proposed by Atlas et al. [53] are discarded from the raw data.Considering the impact of sampling height, before quality control, air-density adjustments are implemented to modify the empirical relationship proposed by Atlas et al. [53] by multiplying the correction factor [34,54].Te modifed relationships are given by the following equations: where D is the particle diameter, V t is the fall velocity, h is the sampling height above sea level, and δ(h) is a correction factor for air-density changes with height.
As shown in Figure 2, the fltered raindrops are mainly below 3 m•s − 1 , mostly likely related to the "splashing efect."After QC, these unrealistic data are basically removed.Additionally, the defnition of an efective rain event proposed by Chen et al. [34] is adopted in this study.Considering most rain events occurring in the Ili River Valley are intermittent, rain events lasting less than 30 min will be excluded to avoid erratic measurements.In this research, May, June, July, August, and September are considered as rainy seasons, since most rainfall occurs from May to September every year.Ultimately, 56 rain events are selected, which include 20918 1 min RSD samples covering the two rainy seasons of 2020-2021 in the Ili River Valley.

Diurnal Variation of Rainfall Integral Parameters.
Diurnal variation is one of the most basic modes of global weather and climate system variations, and diurnal variation in precipitation is the most signifcant among variations of meteorological variables.Precipitation in Zhaosu exhibits an apparent diurnal cycle during rainy seasons.Precipitation mainly occurs in the late afternoon and early evening (Figure 3).Te largest contributor to the total rainfall is around 14-19 LST.
Figure 4 displays the time series of mean values of integral rainfall parameters derived from 1-minute disdrometer observations.Te time series of R, Z, and W present an obvious multipeak structure.Te mean values of R, Z, and W range from 0.67 to 2.64 mm•h − 1 , 17 to 24.4 dBZ, and 0.06 to 1.14 g•m − 3 , respectively.Te maximum and minimum values of R appear at 19 and 08 LST, respectively.A same variation pattern is found for Z and W. Analysis of the trends of D m and log 10 N w indicates that the D m value shows a fuctuating upward trend until 19 LST and then turns to a downward trend, while the log 10 N w value shows an opposite trend.From 14 to 20 LST, the result shows a higher mass-weighted mean diameter D m (>1.1 mm) and a lower normalized intercept parameter log 10 N w (<3.6).Te terminal velocity exhibits an upward trend between 08 and 14 LST, fuctuates around 3.8 m•s − 1 during 14-00 LSTperiod, and then decreases rapidly.Obviously, the variations in rainfall rate, radar refectivity, and liquid water content are controlled by the size of raindrops.Te terminal velocity of particles also matches with the diameter of raindrops.On the other hand, the signifcant increase (decrease) in D m (log 10 N w ) from afternoon to early evening is probably related to the intensifcation of collision-coalescence processes, which may be responsible for the rapid growth of raindrop size.

Diurnal Variation of RSD.
In order to explore the diurnal variation of RSD in the Ili River Valley during rainy seasons, each day is divided into four time periods (02-07, 08-13, 14-19, and 20-01 LST). Figure 5 shows the composite raindrop spectra corresponding to these four time periods.In this paper, the classifcation method proposed by previous researchers is adopted to divide raindrops into three categories: (a) small drops: <1 mm, (b) medium-sized drops: 1 ∼ 3 mm, and (c) large drops: >3 mm [55][56][57].Te RSD spectra are quite similar at diferent time periods, showing a unimodal distribution with the peak value appearing around the diameter D ∼ 0.5 mm.It can be seen that, except for 02-07 LST, the number of small raindrops is almost the same in all the other periods.As the raindrop diameter increases, the variations in RSD spectra start showing up.Te concentration of medium and large drops in the period of 14-01 LST is higher than that in the period of 02-13 LST.And this further confrms that the increase in rainfall intensity from the afternoon to early evening is mainly owing to the contribution of medium and large drops.
Figure 6 shows the histograms of diferent RSD parameters for each defned time period with respect to the entire data set.Te histogram of log 10 R displays a pronounced unimodal distribution with the peak located at approximately − 0.4 in all the periods except for the period of 20-01 LST (Figures 6(a-d)).Te statistical characteristics of log 10 R show a higher mean and standard deviation at 14-19 LST, indicating that the precipitation variability is stronger.Te log 10 W histograms in all the four time periods are relatively similar and meet a normal skewness distribution (Figures 6(e-h)).Te log 10 W frequency distributions for the four periods are concentrated between − 2.2 and 0.2 with an almost symmetrical shape.Figures 6(i-l) present a signifcant diurnal variability of D m .For the four time periods, the peak value of D m exists at around 0.8 mm and the frequency rapidly decreases from the peak value as D m increases.Compared with the other three time periods, the D m histogram in the 14-19 LST period shows a much wider distribution with a higher frequency when D m > 1.8 mm.Te log 10 N w histogram is negatively skewed throughout the day, which displays an inverse distribution to that of D m (Figures 6(m-p)).

RSD Properties for Diferent Rainfall Rates.
To better understand the performance of RSD spectra corresponding to diferent rain intensities from light to heavy, the processed data have been further stratifed into fve classes on the basis of rainfall rate: 0 4 Advances in Meteorology  6

Advances in Meteorology
Figure 7 shows the diurnal variations of the RSD spectra for the fve diferent rainfall rate classes.Te breadth of the RSD shape and the concentration of raindrops in the four time periods increase with increasing rainfall rate at diferent rainfall rates.In particular, the abundance of small raindrops can be noted in all the periods.Te disagreement among the four curves increases signifcantly when R ≥ 5 mm•h − 1 , indicating that the diurnal diference of RSD is more pronounced in the case of high rainfall rates.Te integral rainfall parameters for the fve rainfall rate classes at diferent time periods are summarized in Table 1.Te mean value of D m increases monotonically with increasing rainfall rate in all the four time periods.On the contrary, the shape parameter μ and slope parameter Λ tend to decrease with increasing rainfall intensity.Tere is no regularity in the variation of log 10 N w within each defned time period.In all rainfall rate classes, the maximum mean D m value occurs during 14-19 LST, followed by that during 20-01, 08-13, and 02-07 LST.On the contrary, the maximum mean log 10 N w value appears over 02-07 LST, followed by that over 08-13, 20-01, and 14-19 LST.Te larger μ values during 02-07 LST indicate that the growth of large diameter particles is suppressed, which may be attributed to weak convective activity in the early morning.

RSD of Diferent Rain
Types.Natural rainfall is generally divided into two essential types: stratiform and convective precipitation.To divide precipitation into these two types, diferent researchers have adopted diferent classifcation criteria [1,58].In this study, the classifcation procedure of Bringi et al. [1] that is based on the standard deviation of the consecutive rainfall rate (σ R ) is used to distinguish the two precipitation types.For 10 consecutive 1-min RSD samples, samples with R ≥ 0.5 mm•h − 1 and σ R ≤ 1.5 mm•h − 1 are classifed as stratiform rain and samples with R ≥ 5.0 mm•h − 1 and σ R ≥ 1.5 mm•h − 1 are classifed as convective rain.Te remaining samples are classifed as mixed rain and excluded from the investigation.Using the above classifcation scheme, 11098 stratiform samples and 1174 convective samples are identifed, and results are given in Table 2.In addition, there are 1441 (70), 2156 (32), 3238 (677), and 4263 (395) stratiform (convective) rainfall samples over 02-07, 08-13, 14-19, and 20-01 LST.
Figure 8 shows the diurnal variations of accumulated rain amount and rain duration for stratiform and convective rain types.During the observation period from 2020 to 2021, the accumulated rain amounts of stratiform and convective rain are 255.2 and 199.3 mm, respectively, while the occurrence times of stratiform and convective rain are 180.52 and 19.56 h, respectively.Despite the negligible contribution of convective precipitation to the total rain duration, its contribution to the accumulated rain amount is signifcant.Diferences in the diurnal cycle between the two rain types are also observed.Stratiform precipitation in Zhaosu can occur at any time of the day, and the maximum accumulated rain amount of stratiform precipitation occurs at 20-01 LST.Convective precipitation in Zhaosu mainly occurs during the late afternoon and early evening when convection is active, and the maximum accumulated rain amount of convective precipitation occurs at 14-19 LST.
To further understand the RSD characteristics of different rain types, Figure 9 shows the composite raindrop spectra for stratiform and convective rain types over different time periods.Although the RSDs in the four time periods exhibit a similar distribution for the same rain type,   Advances in Meteorology clear diferences in the RSDs can be found between stratiform and convective rain types.Te RSD in stratiform rainfall shows a steep slope, while that in convective rainfall shows a gentle slope.Compared to that for stratiform precipitation, the mean RSD for convective rain type has a wider and fatter spectral width.For both stratiform and convective precipitation, the maximum raindrop diameter appears over 14-19 LST.Te spectra indicate that convective precipitation has a higher number concentration of medium and large drops than stratiform precipitation, which may be attributed to the collisional breakup of large raindrops in convective rainfall [7].For stratiform precipitation, the four    Advances in Meteorology curves agree well with each other for raindrop sizes smaller than 4 mm.In contrast, the concentration of raindrops larger than 4 mm is signifcantly higher during 14-19 and 20-01 LST.For convective precipitation, the gap between the four curves gradually widens when the raindrop diameter is greater than 2 mm, and the raindrop number concentration in the afternoon is signifcantly higher than in other time of the day.Table 2 lists the statistics of rain parameters for the two rain types at diferent time periods.From the table, it can be found that the mass-weighted average diameter D m for convective rain is larger than that for stratiform rain.In contrast, the mean log 10 N w , μ, and Λ values of stratiform rain are higher than those of convective rain.LST has the highest mean value of log 10 N w and the lowest mean value of D m .Chen et al. [34] analyzed three-year disdrometer data over the central Tibetan Plateau and found that convective daytime rain could be identifed as continental-like rain, while convective rain at night could be identifed as maritime-like rain.Figure 10 demonstrates that the convective rain in the Ili River Valley over 14-19 LST can be characterized as the continental convective cluster, while in the rest time of the day, it is neither in the maritime cluster nor in the continental cluster.

Class (mm•h
3.6.Z-R Relationship.Te power-law relationship between radar refectivity factor Z and rainfall rate R (Z � A•R b ) is the most widely used method for single-polarization radar QPE.Establishing a suitable Z-R relationship can minimize the uncertainties in radar QPE for a certain area.
Figure 11 shows the scatter plots of Z versus R, as well as the ftting power-law relationships for stratiform and convective rain types.Te Z-R relationships for stratiform and convective rain are Z � 123.53R 1.73 and Z � 97.08R 1.85 , respectively.Te empirical relationship of Z � 300R 1.4 is commonly used in the Next-Generation Weather Radar (NEXRAD) for convective precipitation [59], and the empirical relationship of Z � 200R 1.6 is recommended in midlatitude areas for stratiform precipitation [60].Te ftting curve of stratiform precipitation in Zhaosu is basically coincident with the ftting curve of stratiform precipitation in midlatitude areas, and there is not much diference in the estimation of rainfall between the two relationships (Figure 11(a)).Te empirical relationship (Z � 300R 1.4 ) underestimates the convective rain at a rainfall rate below 12 mm•h − 1 , while it overestimates the convective rain at a rainfall rate above 12 mm•h − 1 .In other words, the default NEXRAD relationship may increase the uncertainty of QPE.

Discussion
Many studies have indicated that the diurnal variation of RSD characteristics is not a random behaviour.It is closely related to microphysical processes and environment factors [5,61].To explore the mechanisms for the observed diurnal variation of RSD in the Ili River Valley, temperature ( °C), relative humidity (%), wind speed (m•s − 1 ), and CAPE (J•kg − 1 ) obtained from the automatic weather station (AWS) data and ERA5 reanalysis data for the rainy seasons of 2020-2021 in Zhaosu are employed in the present study.As shown in Figure 12, temperature, humidity, wind speed, and CAPE have obvious diurnal variation characteristics.Temperature and wind speed remain low from the nighttime to early morning, then gradually increase and reach their peaks in the late afternoon.Relative humidity, however, shows an opposite trend.Lower temperature and wind speed are not conducive to the evaporation and collision of small raindrops, which may be one of the reasons for the abundance of small raindrops and the lack of large and medium raindrops in the midnight and early morning.From 08 LST, temperature gradually increases while the heating efect of solar radiation increases the atmospheric instability, and CAPE begins to accumulate.In addition, due to the topography blocking efect, lowlevel airfows converge at the foot of the mountain and ascend along the windward slope, triggering convective activities.Te trumpet-shaped topography in the Ili River Valley allows the warm moist airfow from the Atlantic Ocean to converge there, and the abundant water vapour is conducive to the maintenance and enhancement of convection.In the afternoon, shortwave solar radiation intensifes sharply, ascending motion further develops, pressure decreases, and wind speed increases signifcantly.When CAPE gets released at 17 : 00, precipitation reaches an intense stage.Te precipitation gradually decays afterwards, resulting in more stratiform rainfall from the anvil clouds.Strong ascending motion and uneven heating of the land surface can prolong the interaction time of particles and provide a favourable condition for collision-coalescence processes of particles, causing them to grow bigger in size.Tis process is sometimes referred to as drop sorting [5,62].Tis helps to explain the higher concentration of 10 Advances in Meteorology medium and large raindrops from the afternoon to early evening.At the same time, the warmer and drier environment accelerates the evaporation of small raindrops, and the processes of collision-coalescence will also capture a large number of small particles, leading to a signifcant decrease in the concentration of small raindrops.Note that the fndings of this work are based on observations from a single measurement instrument.Diferent types of advanced observation instruments should also be used to further study the microphysical characteristics of clouds and precipitation in the Ili River Valley.

Summary and Conclusions
In this study, the diurnal variation of RSD for diferent rainfall rate categories and diferent rain types in the Ili River Valley during rainy seasons from 2020 to 2021 is investigated using observations of a Parsivel 2 disdrometer.Integral rainfall and gamma parameters are also investigated to interpret possible microphysical processes associated with the RSD.Te main conclusions are as follows: (1) Tere are signifcant diurnal diferences in precipitation and RSD in Zhaosu during rainy seasons.Precipitation mainly occurs in the late afternoon and early evening.Te variations in rainfall rate, radar refectivity, and liquid water content are controlled by the size of raindrops.Te increase in rainfall intensity from the afternoon to early evening is mainly owing to the contribution of medium and large drops.Advances in Meteorology (2) Te RSD exhibits diurnal diferences between different rainfall rate classes; the diurnal diference of RSD is more pronounced in the case of high rainfall rates.In all rainfall rate classes, the RSDs in 14-19 LST have higher mass-weighted mean diameter and lower normalized intercept parameter than in other time of the day.(3) Stratiform precipitation can occur at any time of the day, yet convective precipitation mainly occurs during the late afternoon and early evening.Te convective spectrum has a higher number concentration of medium and large drops than that of stratiform rainfall in all the four time periods.
Convective rain in the Ili River Valley over 14-19 LST can be characterized as the continental convective cluster, while in the rest time of the day it is neither in the maritime cluster nor in the continental cluster.(4) Te ftting curve of stratiform precipitation in Zhaosu is basically coincident with the empirical relationship of Z � 200R 1.6 .Te empirical relationship (Z � 300R 1.4 ) underestimates the convective rain when R < 12 mm•h − 1 , while it overestimates the convective rain when R > 12 mm•h − 1 .(5) Te possible mechanisms responsible for the diurnal variation of RSD are discussed.Distinct diferences in temperature, relative humidity, wind speed, and CAPE may be responsible for the diurnal variation characteristics of RSD.

Figure 1 :
Figure 1: Location of Zhaosu observation site.Te blue rectangle represents the Ili River Valley.

Figure 2 :Figure 3 :
Figure 2: Accumulated number of raindrops corresponding to diferent diameter size and fall velocity bins (a) before and (b) after QC for the entire observation period.Te solid lines represent the empirical fall velocity-diameter relationship proposed by Atlas et al. [53] and the air-density adjustment is considered.Te dashed lines indicate the ±60% fall velocity-diameter relationship.

Figure 8 :
Figure 8: Diurnal variations of accumulated rain amount (blue) and rain duration (red) for (a) stratiform and (b) convective rain types.

3. 5 .
Distributions of D m and N w .Figure 10 depicts the distributions of log 10 N w versus D m for stratiform and convective rain in the four time periods.Te two black rectangles correspond to the maritime and continental convective clusters, and the black dashed line is a separation line between stratiform and convective rain proposed by Bringi et al. [1].In the case of maritime convective cluster, the D m values are scattered from 1.5 to 1.75 mm, and the log 10 N w values are between 4 and 4.5.And the D m values of continental convective cluster are distributed between 2 and 2.75 mm, and the log 10 N w values are concentrated between 3 and 3.5.Te mean D m (log 10 N w ) values for convective rain are about 1.65 (3.67), 1.7 (3.71), 2.06 (3.28), and 1.67(3.68)  in the periods of 02-07, 08-13, 14-19, and 20-01 LST, respectively, which are larger (lower) than their counterparts for stratiform rain during the same time period.Te mean D m -log 10 N w pairs difer signifcantly in the diurnal cycle for stratiform and convective rain types.For stratiform rain, most points are concentrated in the region with lower D m and higher log 10 N w .It is found that the stratiform rain over 02-07

Figure 9 :Figure 10 :
Figure 9: Composite raindrop spectra for stratiform and convective rain types at diferent time periods of the day.

Figure 11 :Figure 12 :
Figure 11: Scatter plots of Z (mm 6 •m − 3 ) versus R (mm•h − 1 ) for stratiform and convective rain types.Te solid blue lines indicate the ftting power-law relationships in Zhaosu.Te black solid line and black dashed line, respectively, represent the empirical relationships (Z � 200R 1.60 for stratiform rain type and Z � 300R 1.40 for convective rain type).

12
is the number concentration of raindrops per unit volume with diameters between D i and D i + ∆D i ; D i (mm) is the median value of the i th size bin; ∆D i (mm) is the diameter interval of the i th size bin; ∆t (s) is the 2Advances in Meteorology sampling time interval; n ij is the number of drops within size bin i and velocity bin j; V j (m•s − 1 ) is the fall velocity of class j; and S eff (D i ) (m 2 ), the efective sampling area, is expressed as 180 mm × (30 mm-0.5 D i ).

Table 1 :
Mean values of R, D m , log 10 N w , μ, and Λ for the fve rainfall rate classes at diferent time periods of the day.

Table 2 :
Mean values of R, D m , log 10 N w , μ, and Λ for stratiform and convective rain types at diferent time periods of the day.