3.1 Characteristics of weak precipitation
3.1.1 DSD characteristics of weak precipitation
According to the definition of precipitation events, the basic features of precipitation events in the study period are shown in Table 1. During March 15th to April 1st, there were nine precipitation events with large rainfall amounts in Pudong, Shanghai, and 1769 one-minute raindrop spectrum samples were collected.
Average DSD parameters for all precipitation events during the study period are shown in Table 2, including the mean and standard deviation of mass-weighted mean diameter (Dm), logarithm of the normalized intercept parameter (Log10Nw), rain rate (R), radar reflectivity factor (Z) and total number of raindrops concentration (NT). The mean values of Dm and Log10Nw for all DSD samples are 0.98 mm and 3.73 m-3 mm-1, respectively, which are similar to the weak precipitation in Xinjiang studied by Zeng et al. (2021), The mean values of R, Z and NT are 0.18 mm h-1, 19.65 dBZ and 338.33 m-3, respectively. It can be seen that there was weak precipitation during this period in Shanghai, which was consistent with the result measured by the rain gauge. Figure 3 shows the frequency distribution histogram of Dm and Log10Nw for all samples. For the whole data, the histograms of Dm and Log10Nw are positively skewed, but Dm is more skewed. All DSD samples were located in the range of Dm = 0.42-3.61 mm and Log10Nw = 1.99-5.18 m-3 mm-1.
Table 1.Precipitation event statistics
Event no.
|
Date
|
Time (BJT)
|
Sample numbers
|
E1
|
15-16 Mar
|
20:33-23:57 00:01-10:24
|
173 502
|
E2
|
17 Mar
|
13:16-15:11
|
115
|
E3
|
19 Mar
|
04:45-07:58
|
161
|
E4
|
19 Mar
|
11:06-13:30
|
131
|
E5
|
27 Mar
|
04:11-09:48
|
213
|
E6
|
31 Mar
|
04:37-08:44
|
248
|
E7
|
31 Mar
|
22:00-23:01
|
62
|
E8
|
1 Apr
|
01:51-03:21
|
89
|
E9
|
1 Apr
|
04:15-05:34
|
75
|
Table 2. Average DSD parameters for all samples of precipitation events
Parameters
|
Dm (mm)
|
Log10Nw (m-3 mm-1)
|
R (mm h-1)
|
Z (dBZ)
|
NT (m-3)
|
Mean
|
0.98
|
3.73
|
0.18
|
19.65
|
338.33
|
SD
|
0.33
|
0.56
|
0.24
|
8.08
|
401.03
|
DSD characteristics in different rain rate classes
In order to further understand the DSD characteristics of weak precipitation, the DSD data of nine precipitation events are divided into five classes: 0.01<R≤0.1 mm h-1 (C1), 0.1<R≤0.2 mm h-1 (C2), 0.2<R≤0.5 mm h-1 (C3), 0.5<R≤1 mm h-1 (C4), and R>1 mm h-1 (C5). The sample numbers were 875, 421, 340, 100 and 33, respectively. The variation of Dm and Log10Nw for different rain rate classes are presented in Fig. 4a-b. The mean value of Dm increases with the increase of rain intensity, while the mean value of Log10Nw increases first and then decreases with the increase of rain intensity, which is consistent with the results of Ma et al. (2019). The mean and standard deviation of Dm, Log10Nw and R as well as the number of samples for different rain rate classes are shown in Table 3. Dm and Log10Nw vary in the range of 0.85-1.65 mm and 3.54-4.23 m-3 mm-1, respectively.
The average raindrop spectrum of precipitation with different rain intensity levels is shown in Fig. 4c. It can be seen that the width of the raindrop spectrum becomes large with the increase of rain intensity and the maximum raindrop diameter reaches 5.5 mm. The raindrop spectrum of different rain intensity classes presents a single-peak distribution, and the number concentration of small raindrop is high. The peak diameter of rain rate is 0.437 mm when 0.01<R≤0.1 mm h-1, and 0.687 mm when R>1 mm h-1, and that of the rest of the rainfall intensity is 0.562 mm. That is to say, as the rain intensity increases, the peak position moves to a larger diameter, which is the same as the findings of Jash et al. (2019). When 0.312≤D≤0.437 mm, the number concentration of raindrops at each size bin presents a double-peak type with the increase of rain rate, and the peaks are around R ~ 0.1-0.2 mm h-1 and R ~ 0.5-1 mm h-1, When 0.437<D≤0.812 mm, the number concentration of raindrops at each size bin shows a single peak with the increase of rain intensity, and there is a peak around R ~ 0.5-1 mm h-1, When D>0.812 mm, the raindrop number concentration increases with the increase of rain intensity.
Table 3. Mean and standard deviation of Dm, Log10Nw and R for different rain rate classes.
|
Rain rate threshold
|
Sample numbers
|
Dm (mm)
|
Log10Nw (m-3 mm-1)
|
R (mm h-1)
|
Mean
|
SD
|
Mean
|
SD
|
Mean
|
SD
|
C1
|
0.01<R≤0.1
|
875
|
0.85
|
0.25
|
3.54
|
0.52
|
0.05
|
0.02
|
C2
|
0.1<R≤0.2
|
421
|
1.01
|
0.32
|
3.83
|
0.57
|
0.14
|
0.03
|
C3
|
0.2<R≤0.5
|
340
|
1.15
|
0.30
|
3.91
|
0.50
|
0.31
|
0.08
|
C4
|
0.5<R≤1
|
100
|
1.16
|
0.36
|
4.23
|
0.51
|
0.66
|
0.13
|
C5
|
R>1
|
33
|
1.62
|
0.57
|
3.98
|
0.47
|
1.38
|
0.41
|
3.2 Dust aerosol effect on weak precipitation properties
3.2.1 Two dust events
A large area of strong dust weather occurred in northern China from March 14th to 18th and from March 27th to 29th, 2021. Some cities in Inner Mongolia, northwest China, North China, northeast China and eastern coastal areas are successively affected by dust weather, which makes the air quality of the areas transported by dust deteriorate rapidly (Yang et al. 2021, Guan et al. 2021). It not only threatens people's life and property safety, but also affects ecology, environment, traffic safety and human health (Jaafari et al. 2021, Yin et al. 2022).
During the long-distance transportation, some of the dust is carried by cyclones over the Bohai and Yellow seas. The studies of Filonchyk (2022) and Gui et al. (2021) also showed that the dust was transported to the Shanghai area during the two dust events. As can be seen from Fig. 5a-d, the dominant wind direction over the Yellow Sea was northeasterly on March 16th-17th and March 30th. On March 17th and March 30th, under the influence of the northeast wind over the Yellow Sea, the sand and dust trapped on the sea flowed back and temporarily increased the concentration of PM10 in Shanghai. In addition, HYSPLIT backward trajectory model was used to analyze the source path of dust aerosol in Shanghai on March 16th and March 30th, and the results are shown in Fig. 5e-f. The backward trajectory analysis also shows that the dust affecting the air quality of Shanghai mainly comes from the backflow of the sea.
From the hourly changes of AQI and PM10 concentrations in Fig. 6, affected by the return of dust from the sea, the air quality in Shanghai was slightly polluted for a short time on March 16th, and it reached severe pollution for a short time on March 30th. The concentration of PM10 began to increase at 16:00 on March 16, and then rose rapidly to make the air quality slightly polluted. The PM10 concentration reached a maximum value of 162 μg m-3 at 3:00 on March 17th. On March 30th, the PM10 concentration began to increase sharply at 9:00, making the air quality seriously polluted at 11:00, and the PM10 concentration peaked at 711 μg m-3 at 12:00. Due to the removal of precipitation, the influence of the first dust event (D1) on air quality in Shanghai was less than that of the second dust event (D2). According to the above analysis, event 2 (E2) and event 6 (E6) in Table 1 were dust-carrying precipitation with effective sample numbers of 115 and 248, respectively.
3.2.2 Dust aerosol effect on raindrop size distribution
In order to understand the influence of dust aerosol on the DSD of weak precipitation, dust precipitation and dust-free precipitation with similar average rainfall intensity were selected from the precipitation events in Table 1. Then, the difference of DSD between them were compared and analyzed. The DSD feature parameters are shown in Table 4, D1 and D2 are dust-carrying precipitation, while DF1 and DF2 are dust-free precipitation (The precipitation period is shown in Table 1). The mean values of Dm and Log10Nw for D1 and D2 were 1.03 mm and 1.00 mm, 3.52 m-3 mm-1 and 3.59 m-3 mm-1, respectively. The mean values of Dm and Log10Nw for DF1 and DF2 were 1.03 mm and 1.06 mm, 3.42 m-3 mm-1 and 3.48 m-3 mm-1, respectively. It can be seen that when the rain intensity is similar, the average value of Dm for dust-carrying precipitation is smaller than that of dust-free precipitation, while the average value of Log10Nw is larger than that of dust-free precipitation. Independent sample t-test (Table 4) was used to examine the significance of Dm and Log10Nw differences between precipitation with and without dust (significance level: P<0.05). The results show that the Log10Nw of dust-carrying and dust-free precipitation are significantly different. However, for Dm, the difference between D1 and DF1 is not obvious, while the difference between D2 and DF2 is obvious. The standard deviations of Dm and Log10Nw for dust-carrying precipitation are smaller than those of dust-free precipitation, indicating that dust-carrying precipitation has a smaller variability in Dm and Log10Nw.
The frequency distribution histogram and raindrop spectrum characteristic parameters of Dm and Log10Nw for dust-carrying precipitation and dust-free precipitation are shown in Fig. 7a-b. Combining with Table 4, the histogram of Dm for dust-carrying precipitation and dust-free precipitation are all positively skewed, while Log10Nw are all negatively skewed. In addition, the distribution ranges of Dm and Log10Nw for dust-carrying precipitation (Dm = 0.61-1.88 mm, Log10Nw = 2.59-4.14 m-3mm-1) are both smaller than those for dust-free precipitation (Dm = 0.56-2.1 mm;Log10Nw = 2.04-4.36 m-3mm-1). This indicates that under the influence of dust aerosol, not only the weighted average diameter of raindrops decreases, but also the distribution range of raindrop diameters. That is to say, in the case of high dust aerosol concentration, the number concentration of small raindrops increases and the mass weighted average diameter of raindrops decreases.
In addition, the average raindrop size spectra for dust-carrying precipitation and dust-free precipitation in Shanghai were analyzed (Fig. 7c). As shown in Fig. 7c, the average raindrop spectra for dust-carrying precipitation and dust-free precipitation are both single-peak, with a peak diameter of 0.562 mm. When D<1.625 mm, the raindrop number concentration of DF2 (D2) is greater than that of DF1 (D1), when D≥1.625 mm, the raindrop number concentration of DF2 (D2) is smaller than that of DF1 (D1). Overall, the average spectrum of dust-carrying precipitation is higher than that of dust-free precipitation when the particle size is small, but the opposite feature was observed when the particle size is large. It is further explained that the increase of dust aerosol concentration may lead to an increase in the number of small raindrops and a significant decrease in the number of large raindrops, resulting in a more concentrated raindrop spectrum and a smaller distribution range. This is consistent with the study of Wen et al. (2016), whose results show that in the case of high aerosol concentration, a large number of condensation nuclei and sufficient water will lead to a high concentration of small raindrops.
Table 4. Mean and standard deviation of Dm, Log10Nw and R for dust-carrying precipitation and dust-free precipitation
|
Date
|
Dm (mm)
|
Log10Nw (m-3 mm-1)
|
R (mm h-1)
|
Mean
|
SD
|
Skew
|
P
|
Mean
|
SD
|
Skew
|
P
|
Mean
|
SD
|
DF1
|
15 Mar
|
1.03
|
0.28
|
1.34
|
0.990
|
3.42
|
0.36
|
-0.09
|
0.018*
|
0.13
|
0.21
|
D1
|
17 Mar
|
1.03
|
0.26
|
0.84
|
3.52
|
0.32
|
-0.54
|
0.13
|
0.13
|
DF2
|
27 Mar
|
1.06
|
0.29
|
0.76
|
0.005**
|
3.48
|
0.39
|
-0.45
|
0.000**
|
0.14
|
0.14
|
D2
|
31 Mar
|
1.00
|
0.20
|
1.06
|
3.59
|
0.26
|
-0.33
|
0.13
|
0.12
|
*Pass the 95% confidence level;** Pass the 99% confidence level.
Dust aerosol effect on weak precipitation
In Section 3.2.2, we analyzed the effect of dust aerosol on DSD of weak precipitation. In this section, we explore the impact of dust aerosols participating in cloud microphysical processes on cloud life cycles and weak precipitation. According to the MERRA-2 dust mixing ratio profile in Fig. 8a and PM10 in Fig. 6, we observed dust transport at 500-700 hPa on March 30th, 2021. With the deposition of dust, the mixing ratio of dust in the upper air decreased and the mixing ratio of dust at ground increased, which was consistent with the results of ground observation. In the early morning of March 31st (2:00-8:00), the dust mixing ratio at around 850 hPa gradually decreased, while the dust mixing ratio at the ground did not change, which may be related to the fact that the dust aerosol mainly participated in the cloud microphysics process and only a small part of the dust aerosol settled to the ground. Here, we record the two precipitation events on March 31st (E6 and E7 in Table1) as event 331-1 (E331-1) and event 331-2 (E331-2) respectively in chronological order. Our analysis showed that dust aerosols at ground and high altitude of E331-1 were larger than those of E331-2. Therefore, the precipitation event on March 31st was selected as an example to further reveal the influence of dust aerosol on weak precipitation. Figure 8b shows the time evolution of the raindrop spectrum, Dm and R of the two precipitation events on March 31st, 2021. The precipitation duration of E331-1 (247 min) was longer than that of E331-2 (62 min), and the two precipitation events were dominated by a large number of small raindrops (less than 1 mm). In this study, according to the raindrop diameter classification of Krishna et al. (2016), raindrops with a diameter of less than 1mm are regarded as small raindrops, 1-3 mm as medium raindrops, and those larger than 3 mm as large raindrops.
The evolution of raindrop spectrum of E331-1 presents a trend of fluctuation and decline. The variation of raindrop spectrum is mainly related to the collision-merging process and the fragmentation process of raindrops. When the raindrop fragmentation process is dominant, the number concentration of small raindrops increases with the large raindrop fragmentation, and the average Dm value is small. When the collision-merging process of raindrops is dominant, the generation of small raindrops decreases, the number concentration of medium and large raindrops increases, and the average Dm value is larger. We found that during the early precipitation period (5:07-5:19), the peak number concentration of small raindrops moved to a smaller diameter and stabilized around 0.5mm. This may be due to the increase of CCN by dust aerosols, and a large amount of CCN promotes the movement of raindrops to smaller diameters (Chen et al. 2021). High concentration of CCN usually increase the number concentration of cloud droplets and decrease the effective radius of cloud droplets, thus inhibiting the collision-merging process, reducing the precipitation efficiency and prolonging the cloud lifetime (Rosenfeld, 2000, Ackerman et al. 2003).
We further analyzed the mean DSD characteristics of the two precipitation events on March 31st, 2021 (Fig. 9). The results show that higher dust aerosol concentration may increase the number concentration of small raindrops and decrease the number concentration of large raindrops. In Fig. 9a, the average raindrop spectrum under the influence of dust aerosol is unimodal, but the peak diameter of E331-1 (0.562 mm) is smaller than that of E331-2 (0.687 mm). This suggests that higher dust aerosols make smaller raindrops. The raindrop spectrum of E331-1 in the vigorous precipitation stage is selected, as shown in Fig. 9b. E331-1.1, E331-1.2 and E331-1.3 correspond to the average raindrop spectra of 4:37-5:32, 4:37-5:32 and 6:02-6:30 respectively. It can be seen that, in terms of the number concentration of small raindrops, the average spectrum of E331-1 in each time period is similar, but that of E331-1.3 is higher. For the number concentration of large raindrops, the average spectrum of E331-1.1 and E331-1.2 in the early stage of precipitation is similar to that of E331-2, while the average spectrum of E331-1.3 in the later stage is lower and narrower. The above analysis shows that the high concentration of dust aerosol leads to the increase of CCN, and the high concentration of CCN makes the number concentration of small raindrops more and smaller, while the number concentration of large raindrops significantly decreases. In addition, high concentration of CCN can reduce the effective radius of cloud droplets, thus reducing the precipitation efficiency of the collision process, which may suppress the weak precipitation.