Impact of Precipitation with Different Intensity on PM2.5 over Typical Regions of China

Atmospheric aerosol pollution has significant impacts on human health and economic society. One of the most efficient way to remove the pollutants from the atmosphere is wet deposition. This study selected three typical atmospheric pollution regions in China, the Beijing-Tianjin-Hebei (BTH), the Yangtze River Delta (YRD) and the Pearl River Delta (PRD) regions, as research areas, and used the hourly precipitation and PM2.5 mass concentration data from 2015 to 2017 to investigate the removal impacts of precipitation on PM2.5. The PM2.5 mass concentration difference before and after the hourly precipitation events was used to denote as the impacts of precipitation. Hourly precipitation event was selected so that the time difference between two PM2.5 observations was short enough to limit the PM2.5 change caused by other factors. This study focused on the differences in the removal effect of precipitation on PM2.5 under different precipitation intensities and pollution levels. The results show that both precipitation intensity and aerosol amount affected the removal effect. A negative removal effect existed for both light precipitation and low PM2.5 mass concentration conditions. In contrast, a positive removal effect occurred for both high precipitation and high PM2.5 mass concentration conditions. The removal effect increased with increasing precipitation intensity and PM2.5 mass concentration before precipitation and was consistent with the change trend of wind speed at a height of 100 m. The findings of this study can help understand the mechanism of wet scavenging on air pollution, providing support for air pollution control in future.


Introduction
With the rapid development of the economy and urbanization progress in China, more industrial waste gases and particulate matter (PM) have been released to the atmosphere, which threaten the human health and life to a certain extent [1] and have also brought serious ecological and environmental problems [2]. Among various air pollution components, PM with aerodynamic diameters less than 2.5 µm (PM 2.5 ) is often the major contributing factor to air pollution [3,4], which has drawn broad attention worldwide.
Aerosol particles can affect the weather and climate by serving as cloud condensation nuclei (CCN) or changing the radiation balance [5][6][7][8][9][10][11]. By serving as CCN, aerosol particles can increase cloud droplet number concentration and thus enhance the cloud reflection to solar radiation, which is referred to as the cloud albedo effect [8]. For longwave radiation, by serving as CCN, aerosol particles can increase thin cloud thermal emissivity and trap more longwave radiation within the atmosphere, which is referred to as the cloud thermal emissivity effect [6,11]. These two effects are the shortwave by affecting the PM2.5 diffusion direction. Chen et al. [16] figured out that the PM2.5 mass concentration is affected by various meteorological factors including air pressure, temperature, winds, relative humidity, and so on. However, we should note that most existing studies have taken a single city or site as the research area, which generally has problems in representing a large domain. Moreover, uncertainty remains regarding the impacts of precipitation on PM2.5 in China, particularly over the well-known pollution regions.
This study took three typical regions in China as the research areas to investigate the precipitation impacts on concentration of PM2.5, and the effects from winds during precipitation period were also briefly discussed. By limiting the particular precipitation events, we intended to identify quantitatively the variation of precipitation impacts on PM2.5 with both precipitation intensity and PM2.5 amount. The paper is organized as follows. Section 2 provides the study area, data, and method. The analysis and results are presented in Section 3. Section 4 summarizes the findings and gives a discussion.

Region of Interest
This study focused on three typical pollution regions in China, which are the Beijing-Tianjin-Hebei (BTH), the Yangtze River Delta (YRD), and the Pearl River Delta (PRD) regions, as shown in Figure 1. The BTH region lies in Northern China, and belongs to the semihumid area, which is located on the 'leeward slope' on the east side of Taihang Mountains and the south side of Yanshan Mountains, forming a semi-closed terrain. The PRD region, closed to the South China Sea, lies in the southeast of China and has abundant precipitation throughout the year, making the climate warm and humid in this region. Similar to the PRD region, the YRD region also lies in southern China but in northern part of PRD region, generally with less water supply and precipitation.

Data
Due to the site difference, we selected the observations of precipitation and PM2.5 at the same site or closest sites at the same time to form a pair of time series to make further investigation. To avoid the impacts from temporal variation of PM2.5 due to either emission or planetary boundary layer variations, we only investigated the hourly precipitation events. For winds near the site, this

Data
Due to the site difference, we selected the observations of precipitation and PM 2.5 at the same site or closest sites at the same time to form a pair of time series to make further investigation. To avoid the impacts from temporal variation of PM 2.5 due to either emission or planetary boundary layer variations, we only investigated the hourly precipitation events. For winds near the site, this study analyzed the changes of wind speed over a region around the precipitation site. Note that hourly observation data were adopted in this study. The precipitation data used in this study was the integration product of ground observation and Climate Prediction Morphing (CMORPH) hourly precipitation products in China. The CMORPH data is a fusion precipitation product with a resolution of 0.1 • × 0.1 • generated by the National Oceanic and Atmospheric Administration (NOAA) and the Climate Prediction Center (CPC) of the United States Atmospheric and Oceanic Administration. The integration product is generated using a PDF (probability density function matching method) + OI (optimum interpolation method) two-step fusion method based on the observation of precipitation at national automatic sites and the precipitation data retrieved by the CMORPH satellite [38].
The hourly PM 2.5 mass concentration data used in this study is provided by China National Environmental Monitoring Station, a national air quality real-time publishing platform [39]. PM 2.5 is measured by the β-ray absorption method and the micro-oscillation balance method, and the reference method is used for testing. The PM 2.5 data have hourly time resolution and have been qualified officially, and only the data with quality assurance were used in this study.
The hourly reanalysis wind data at 100 m above the ground level from the European Center for Medium-Term Weather Forecast (ECMWF) was used in this study [40]. This data is obtained by applying the laws of physics to combine model data with observational data from all over the world into a global complete and consistent data set, with a spatial resolution of 0.25 • × 0.25 • . The hourly wind data at the grid that the precipitation site lies in is used to represent the large-scale wind condition at the ground site.

Method
As indicated earlier, in order to reduce the impact of daily changes in PM 2.5 , this study selected one-hour precipitation events within a relatively stable period (15:00-17:00 Beijing time) of daily PM 2.5 mass concentration as the research objects. We screened out 1083 cases from 78 sites in the BTH, 5341 cases from 56 sites in the YRD, 2491 cases, from 191 sites in the PRD region for analysis. By doing these data selections, our method was more reliable and meaningful to investigate the precipitation impacts on aerosols, because this processing method limited the natural change of PM 2.5 due to both planetary boundary layer (PBL) variation or emission variation. Under the premise of clarifying the annual (daily) changes in regional precipitation and PM 2.5 , this study focused on the differences in the removal effect of precipitation on PM 2.5 under different precipitation intensities and different pollution levels.
According to the studies of Olszowski [41] and Wang and Feng [24], the removal effect (∆C) of precipitation on PM 2.5 can be used to represent the change of PM 2.5 mass concentration after precipitation, which is defined as where C b is the PM 2.5 mass concentration value before precipitation, and C p is the PM 2.5 mass concentration value after precipitation. When ∆C is larger than 0, the PM 2.5 mass concentration in the atmosphere decreases with precipitation, which is a positive removal process. In contrast, when it is smaller than 0, the PM 2.5 mass concentration in the atmosphere increases with precipitation, which is a negative removal process. When it is equal to 0, the mass concentration of PM 2.5 in the atmosphere has no change, which implies that precipitation plays no role to the PM 2.5 removal process.
This study focused on the relationship between the removal effect and precipitation intensities, along with the relationship between the removal effect and pollution levels. Based on the number of samples, the precipitation events were divided into three categories according to the precipitation intensity and the PM 2.5 mass concentration before precipitation to explore their potential differences in the removal effect of the precipitation on the PM 2.5 . In addition, the effect of winds on the PM 2.5 mass concentration were also investigated and discussed based on hourly winds from ECMWF by selecting typical cases of precipitation events.  Figure 2 shows the monthly variation of precipitation (bar plots) and PM 2.5 (lines) at three study regions, the BTH, YRD, and PRD regions. For all three study regions, there were clear seasonal variations of precipitation, with maximum values in summer (June, July, and August) and minimum values in winter (December and the following January and February). This should be associated with the Asian monsoon system. In summer, the southeast monsoon prevails and carries abundant water vapor from the Pacific Ocean, which enriches summer precipitation. In contrast, wintertime northwest monsoon from the Eurasian continent causes low water vapor content and high-pressure systems, resulting in much less precipitation.  January. It was found that the PM2.5 mass concentration in the BTH region was between 37 to 147 µg·m −3 in the past three years. According to the PM2.5 mass concentration level, there were 25 months with good air quality, 8 months light pollution, and 3 months moderate pollution. The PM2.5 mass concentration in the YRD region was between 24 µg·m −3 to 87 µg·m −3 , which was excellent in air quality for 12 months, good for 22 months, and lightly polluted for 2 months. The PM2.5 mass concentration in the PRD region was between 15 µg·m −3 to 57 µg·m −3 , which was excellent in air quality for 23 months and good for 13 months. Therefore, the air quality in the PRD region was the best in the three years from 2015 to 2017, followed by the YRD, and the worst in BTH region.

Temporal Variation of Precipitation and PM 2.5
The air quality status and temporal variation of PM2.5 is the combination resultant of the PM2.5 emissions, topography, and meteorological factors. The BTH region is located on the 'leeward slope' Statistically, the annual total precipitation amounts in the BTH, YRD, and PRD regions are about 550 mm, 1300 mm, and 1800 mm, respectively. The total annual precipitation in the PRD region is the largest, and in the BTH region is the least. The regional difference is highly related to their relative location in China. As known, the PRD is located in the southeast, closed to the South China Sea, and has sufficient water vapor to provide the necessary support for precipitation. Furthermore, it is located in the subtropical region, with strong solar radiation, which can promote the development of convection and support the formation of precipitation. With lower solar radiation than that in the PRD region, the precipitation in YRD region is slightly less than in the PRD region. It is worth mentioning that due to the frequent occurrence of typhoons in recent years, the YRD and the PRD regions have been seriously affected. Particularly, large amounts of water vapor and then precipitation have been brought to these regions. Differently, the BTH region lies in northern China with insufficient water supply and less precipitation amount.
For the study period, the annual precipitation amount varied with time, with the largest value in 2016 and the lowest value in 2017, and the monthly averaged precipitation amount showed clear temporal variation, with the largest values in July 2016 over BTH region, in June 2015 over the YRD region, and in May 2015 over PRD region. In addition to the Asian monsoon system, the temporal variation of precipitation could be also associated with the El Niño event from 2014 to 2016. For example, due to the influence of El Niño, there was a trend of flooding in the southern China and drought in the northern China in 2015. As a result, April was the month with the highest precipitation in 2015 in BTH region, and the precipitation in June, July, and August significantly decreased compared with the same period in previous years. Figure 2 also shows the monthly variations of PM 2.5 mass concentration in BTH, YRD, and PRD regions. Different from precipitation, the monthly variations of PM 2.5 mass concentration demonstrate the lowest values from June to September and maximum values in December or January. It was found that the PM 2.5 mass concentration in the BTH region was between 37 to 147 µg·m −3 in the past three years. According to the PM 2.5 mass concentration level, there were 25 months with good air quality, 8 months light pollution, and 3 months moderate pollution. The PM 2.5 mass concentration in the YRD region was between 24 µg·m −3 to 87 µg·m −3 , which was excellent in air quality for 12 months, good for 22 months, and lightly polluted for 2 months. The PM 2.5 mass concentration in the PRD region was between 15 µg·m −3 to 57 µg·m −3 , which was excellent in air quality for 23 months and good for 13 months. Therefore, the air quality in the PRD region was the best in the three years from 2015 to 2017, followed by the YRD, and the worst in BTH region.
The air quality status and temporal variation of PM 2.5 is the combination resultant of the PM 2.5 emissions, topography, and meteorological factors. The BTH region is located on the 'leeward slope' on the east side of Taihang Mountains and the south side of Yanshan Mountains, forming a semiclosed terrain. Therefore, the role of the northwest monsoon in autumn and winter seasons is greatly weakened. PM 2.5 is poorly diffused due to unfavorable meteorological conditions and is easily accumulated. In winter, various human activities, such as coal burning and heating, emit a large amount of PM 2.5 into the atmosphere and deteriorate the air quality. In contrast, the planetary boundary layer is high and precipitation is strong in summer (as shown in Figure 2), making the PM 2.5 in BTH region low. Note that the air quality in the BTH area has gradually improved from 2015 to 2017. The PM 2.5 in the YRD and PRD regions has a poor diffusivity under the influence of long-term 'calm weather' and 'radiation inversion' in winter. Moreover, the relatively high humidity helps the hygroscopic growth of PM 2.5 particulates, further causing pollution. Similarly, the planetary boundary layer is high and precipitation is strong in summer (as shown in Figure 2), making the PM 2.5 in YRD and PRD region low. Actually, a good negative relationship between precipitation and PM 2.5 in seasonal variation has been shown in Figure 2. In next section, we investigate the scavenging effect of precipitation on PM 2.5 .

Removal Effect of Precipitation on PM 2.5
In order to clearly show the removal effect of precipitation on PM 2.5 , we first examined the diurnal variation of PM 2.5 mass concentration for cases with and without precipitation over the BTH, YRD, and PRD regions from 2015 to 2017, which is shown in Figure 3. Compared with the averages for cases with precipitation, the averaged mass concentration of PM 2.5 for cases without precipitation was much larger, along with much more significant diurnal variations. This result implies that significant scavenging effect of precipitation on PM 2.5 could exist while other influential factors, such as winds and planetary boundary layer, could also play roles. We next investigate the removal effect of precipitation on PM 2.5 for different precipitation intensities and pollution levels.
YRD, and PRD regions from 2015 to 2017, which is shown in Figure 3. Compared with the averages for cases with precipitation, the averaged mass concentration of PM2.5 for cases without precipitation was much larger, along with much more significant diurnal variations. This result implies that significant scavenging effect of precipitation on PM2.5 could exist while other influential factors, such as winds and planetary boundary layer, could also play roles. We next investigate the removal effect of precipitation on PM2.5 for different precipitation intensities and pollution levels. (a) Removal effect of precipitation on PM2.5 under different precipitation intensities As shown in Figure 3, there are clear diurnal variation of PM2.5 mass concentration, mainly due to the diurnal variation of planetary boundary layer (PBL). Several studies have shown that the PBL varies significantly diurnally and seasonally [42][43][44]. For diurnal variations in the three study regions, it is generally 200-300 m at early morning, increases to about 1000 m around noon and varies little between noon and 17:00 Beijing time, and decreases after 17:00 Beijing time [42]. To isolate the impacts from PBL variation, we investigated the removal effect of precipitation on PM2.5 for a period when PBL varies little, which is 15:00-17:00 Beijing time. Considering this time period is (a) Removal effect of precipitation on PM 2.5 under different precipitation intensities As shown in Figure 3, there are clear diurnal variation of PM 2.5 mass concentration, mainly due to the diurnal variation of planetary boundary layer (PBL). Several studies have shown that the PBL varies significantly diurnally and seasonally [42][43][44]. For diurnal variations in the three study regions, it is generally 200-300 m at early morning, increases to about 1000 m around noon and varies little between noon and 17:00 Beijing time, and decreases after 17:00 Beijing time [42]. To isolate the impacts from PBL variation, we investigated the removal effect of precipitation on PM 2.5 for a period when PBL varies little, which is 15:00-17:00 Beijing time. Considering this time period is short, we only focused on short-term convective precipitation events with duration time no more than one hour. By binning the hourly precipitation with an interval of 0.02 mm/h, Figure 4 shows the variation of precipitation removal effect with precipitation intensity over the BTH, YRD, and PRD regions. Note that three ranges of precipitation intensities have been considered. For all three regions, the removal effects of precipitation on PM 2.5 had clear variation with precipitation intensity, which changed from negative values to positive values with the increase of precipitation intensity. That is to say, with the increase of precipitation intensity, the removal effect changed from the negative removal process to the positive removal process. This finding is consistent with that found by Sun et al. [21] When the precipitation intensity is low, the negative value of removal effects indicates that PM 2.5 mass concentration increases while the increased amount is generally less than 20%. A likely explanation is as follows. When the precipitation is weak, the scavenging effect due to the collision of precipitation droplets and PM 2.5 particles is generally low, but the hygroscopic growth of aerosol particles associated with the high humidity becomes strong, which results in an increasing combination effect on PM 2.5 mass concentration. With the increase of precipitation intensity, the collision efficiency becomes larger and so as the precipitation scavenging effects, which will result in a more positive removal effect of precipitation on PM 2.5 . At the same time, heavy precipitation is often accompanied by strong winds, which further improves the PM 2.5 diffusion capacity, reducing more the PM 2.5 mass concentration. To confirm the findings in this section, we conducted further analysis by dividing the precipitation into three groups with equal sample size according to the precipitation intensity. Correspondingly, the threshold values of precipitation intensities for the classifications were 0.16 mm/h and 0.40 mm/h over BTH region, 0.16 mm/h and 0.36 mm/h over YRD region, and 0.20 mm/h and 0.56 mm/h over PRD region. With this classification, both positive and negative removal effects were found at different precipitation intensity ranges. We then counted the percentages of samples with positive or negative removal effects, which are shown in Table 1 for BTH, YRD, and PRD regions. As can be seen from Table 1, when the precipitation intensity increased, the sample percentage with positive removal effect gradually increased, and the sample percentage with negative removal effect gradually decreased. However, the decreasing (increasing) trend of positive (negative) removal effect with precipitation amount was as clear as that shown in Figure 4. A potential reason is that the threshold values for equal sample size analysis were too close to each other due to the large amount of weak precipitation, which made the differences among the three ranges small. Actually, there were even abnormal trends in the YRD (the sample percentage with positive removal efficiency increased first and then decreased with the intensity of precipitation) and PRD regions (the sample percentage with negative removal efficiency decreased first and then increased with precipitation intensity). More potential reasons are further discussed with case analysis in next section.  The removal effects over the three areas all showed the same change law with precipitation intensity, but there were also differences. In BTH region, the negative removal process was dominant when the precipitation intensity was less than 0.28 mm/h, and the positive removal process was dominant when the precipitation intensity was more than 2 mm/h. For precipitation intensity between 0.28 and 2 mm/h, the removal efficiency slightly increased with intensity (from negative to positive) in tendency, but values fluctuated around the 0 value. In other words, there were three stages of removal effects (negative, roughly neutral, and positive) with precipitation intensities. The removal effect also showed the same three stages in the YRD and PRD regions, but with different threshold values of precipitation intensities. The three stages were for precipitation intensities less than 0.6 mm/h, 0.6 to 4.5 mm/h, and more than 4.5 mm/h in the YRD region, and for precipitation intensities less than 0.25 mm/h, 0.25 to 5 mm/h, and more than 5 mm/h in the PRD region.
Shortly, for precipitation intensities less than 0.28 mm/h, 0.6 mm/h, and 0.25 mm/h, the PM 2.5 mass concentration increased with precipitation amount in the BTH, YRD and PRD regions, respectively. The threshold values for the negative removal effects were small and close to each other in values among the three regions. In contrast, for precipitation intensities larger than 2 mm/h, 4.5 mm/h, and 5 mm/h, the PM 2.5 mass concentration decreased with precipitation amount in the BTH, YRD, and PRD regions, respectively. This result is similar to the findings of Kluska et al. [45]. They took Rzeszów (SE Poland) as the research area and found that aerosol concentrations (pollen) decreased only for precipitation intensities less than approximately 5 mm/h, which indicates that precipitation only decreases the aerosol concentration when the precipitation intensity is greater than a certain value. Combined with the findings from Kluska et al. [45], the critical value of precipitation intensity below which the aerosol concentration increases with precipitation is different among different regions. Since the removal efficiency is dependent on both the hygroscopic growth of aerosols and the collision-coalescence efficiency of precipitation droplets, the threshold values of precipitation intensity for both negative and positive removal efficiencies should be related to both PM 2.5 and precipitation properties, along with the potential impacts from other meteorological components such as winds. In next section, the change of removal efficiency to PM 2.5 mass concentration is further investigated.
To confirm the findings in this section, we conducted further analysis by dividing the precipitation into three groups with equal sample size according to the precipitation intensity. Correspondingly, the threshold values of precipitation intensities for the classifications were 0.16 mm/h and 0.40 mm/h over BTH region, 0.16 mm/h and 0.36 mm/h over YRD region, and 0.20 mm/h and 0.56 mm/h over PRD region. With this classification, both positive and negative removal effects were found at different precipitation intensity ranges. We then counted the percentages of samples with positive or negative removal effects, which are shown in Table 1 for BTH, YRD, and PRD regions. As can be seen from Table 1, when the precipitation intensity increased, the sample percentage with positive removal effect gradually increased, and the sample percentage with negative removal effect gradually decreased. However, the decreasing (increasing) trend of positive (negative) removal effect with precipitation amount was as clear as that shown in Figure 4. A potential reason is that the threshold values for equal sample size analysis were too close to each other due to the large amount of weak precipitation, which made the differences among the three ranges small. Actually, there were even abnormal trends in the YRD (the sample percentage with positive removal efficiency increased first and then decreased with the intensity of precipitation) and PRD regions (the sample percentage with negative removal efficiency decreased first and then increased with precipitation intensity). More potential reasons are further discussed with case analysis in next section. (b) Removal effect of precipitation on PM 2.5 under different pollution levels We classified the PM 2.5 into three groups with the same sample volume based on PM 2.5 mass concentration before precipitation, and then further analyzed the change of precipitation removal effect on PM 2.5 . The threshold values of PM 2.5 for the three-group classification were 25 µg·m −3 and 55 µg·m −3 over TBH, 19 µg·m −3 and 37 µg·m −3 over YRD, and 14 µg·m −3 and 26 µg·m −3 over PRD. Table 2 shows the sample frequencies with positive, negative, and neutral removal effects over the three study regions. For all three study regions, it is clear that the negative removal effect of precipitation dominated at low PM 2.5 mass concentration condition, and the sample frequency with negative removal effect of precipitation decreased with PM 2.5 mass concentration. In contrast, the sample frequency with positive removal effect of precipitation increased with PM 2.5 mass concentration, and the positive removal effect played a dominant role under heavy pollution condition. A likely explanation is that the collision-coalescence effect increased with PM 2.5 mass concentration due to the increased particle density, which made the precipitation scavenging effect more prominent at heavy pollution condition. Differently, when the PM 2.5 mass concentration was low, the collision-coalescence effect was weak, and the hygroscopic growth of aerosol particles could be prominent instead, making the negative removal cases more frequent. Considering that the removal effect of precipitation on PM 2.5 depends on both precipitation intensity and PM 2.5 amount, Figure 5 shows the change of removal effect of precipitation for three equal-sample bins of precipitation intensities under three different PM 2.5 mass concentration conditions over the BTH, YRD, and PRD regions. Consistent with the findings from Tables 1 and 2 for similar precipitation intensity, the removal effect increased with the increase of PM 2.5 mass concentration, that is, the precipitation scavenging was more effective when PM 2.5 mass concentration before precipitation was higher. For a similar PM 2.5 mass concentration before precipitation, the removal effect roughly increased with the increase of precipitation intensity. Note that the negative removal effect values under low PM 2.5 mass concentration conditions were clearly larger than the positive removal effect values under high PM 2.5 mass concentration conditions. A likely reason is that the PM 2.5 mass concentration before precipitation was much smaller at low PM 2.5 mass concentration conditions, making the removal efficiency (relative change of PM 2.5 mass concentration) much higher than that for heavy pollution conditions.
(c) Case analysis Figure 5 shows that under low PM 2.5 mass concentration conditions, the negative removal effect decreased first and then increased with precipitation intensity over BTH, and increased first and then decreased with precipitation intensity over YRD. Under moderate PM 2.5 mass concentration conditions, the removal effect increased first and then decreased with precipitation intensity over PRD. These results imply that other influential factor than precipitation, such as winds, could also affect the PM 2.5 mass concentration simultaneously. We here give a brief discussion about the winds impact on the precipitation removal effect to PM 2.5 using case analysis.
The average wind speed, at a height of 100 m within a domain of 0.25 • × 0.25 • where the study site is located, was explored for three selected cases in the three study regions. The three selected precipitation event cases are those which occurred at the site of Tangshan  Note that for these three cases, the precipitation intensity and PM 2.5 mass concentrations were all different. For the BTH region case, the precipitation intensity was 3.25 mm/h and the PM 2.5 mass concentration before precipitation was 90 µg/m 3 . For the YRD region case, the precipitation intensity was 6.76 mm/h and the PM 2.5 mass concentration before precipitation was 51 µg/m 3 . For the PRD region case, the precipitation intensity was 5.70 mm/h and the PM 2.5 mass concentration before precipitation was 18 µg/m 3 . (c) Case analysis Figure 5 shows that under low PM2.5 mass concentration conditions, the negative removal effect decreased first and then increased with precipitation intensity over BTH, and increased first and then decreased with precipitation intensity over YRD. Under moderate PM2.5 mass concentration conditions, the removal effect increased first and then decreased with precipitation intensity over PRD. These results imply that other influential factor than precipitation, such as winds, could also affect the PM2.5 mass concentration simultaneously. We here give a brief discussion about the winds impact on the precipitation removal effect to PM2.5 using case analysis.
The average wind speed, at a height of 100 m within a domain of 0.25° × 0.25° where the study site is located, was explored for three selected cases in the three study regions. The three selected precipitation event cases are those which occurred at the site of Tangshan Ceramic Company Note that for these three cases, the precipitation intensity and PM2.5 mass concentrations were all different. For the BTH region case, the precipitation intensity was 3.25 mm/h and the PM2.5 mass concentration before precipitation was 90 µg/m 3 . For the YRD region case, the precipitation intensity was 6.76 mm/h and the PM2.5 mass concentration before precipitation was 51 µg/m 3 . For the PRD region case, the precipitation intensity was 5.70 mm/h and the PM2.5 mass concentration before precipitation was 18 µg/m 3 .  Figure 6 shows the hourly variation of wind speed (at a height of 100 m above surface) and removal effect of precipitation on PM 2.5 from 15:00 to 23:00 on the study day over the BTH, YRD, and PRD regions. Note that the precipitation event was still an event with a duration time of no more than one hour, while the PM 2.5 mass concentration at different hours after precipitation was used for the calculation of removal effect at that time in Figure 6. Roughly, the precipitation removal effect and the wind speed had a similar hourly variation trend. This could be easily understood since the winds can enhance the diffusivity of PM 2.5 and reduce its mass concentration. Figure 6 shows that the wind speeds at the hour when precipitation event occurred in BTH (17:00), YRD (16:00) and PRD (16:00) regions were about 4 m/s, 2 m/s, and 3 m/s, and the removal efficiencies at those hours were all negative. This suggests that even with considerable wind speeds, the PM 2.5 still likely increases with precipitation for particular cases due to the aerosol's hygroscopic growth. However, when the winds are large enough, such as more than 4.7 m/s, 3.1 m/s, and 6.6 m/s over BTH, YRD and PRD, the impacts of winds could cause the removal effect change from negative values to positive values. Of course, when the precipitation is large, the positive removal effect of precipitation on PM 2.5 could be further enhanced by the accompanied winds. Gong et al. [14] also found the significant removing effects of winds on PM 2.5 particles, which indicated that nearly 60% of PM 2.5 mass concentrations could be reduced when the wind speed was up to 6 m/s. Thus, the combination of precipitation and winds could enhance the removal effects of precipitation on PM 2.5 , which is definitely worthy of further investigation in future.
Of course, when the precipitation is large, the positive removal effect of precipitation on PM2.5 could be further enhanced by the accompanied winds. Gong et al. [14] also found the significant removing effects of winds on PM2.5 particles, which indicated that nearly 60% of PM2.5 mass concentrations could be reduced when the wind speed was up to 6 m/s. Thus, the combination of precipitation and winds could enhance the removal effects of precipitation on PM2.5, which is definitely worthy of further investigation in future.

Conclusions
Seasonal variations of precipitation and PM 2.5 mass concentration were investigated first. The precipitation intensity showed clear seasonal variation, with maximum values in summer and minimum values in winter for all three study regions of the Beijing-Tianjin-Hebei (BTH), the Yangtze River Delta (YRD), and the Pearl River Delta (PRD) regions. This is highly associated with the Asian monsoon system, including both the southeast and southwest summer Monsoon and northwest winter monsoon. Due to these monsoon system and water supply, the precipitation amount was the largest in the PRD region and the smallest in BTH region. Abnormal changes in the seasonal variations were also affected by the El Niño events. Similar seasonal variations have been found for PM 2.5 mass concentration, implying the potential removal effects of precipitation on PM 2.5 , while other meteorological components, such as winds and planetary boundary layer heights, can also affect the PM 2.5 mass concentration. In all three study regions, the PM 2.5 mass concentration under clear sky was much larger than that under precipitation conditions, also implying the potential removal effect of precipitation on PM 2.5 .
The removal effect of precipitation on PM 2.5 was then investigated. The study shows that the removal effect of precipitation on PM 2.5 is dependent on both the precipitation intensity and PM 2.5 mass concentration, along with other coexisting meteorological components. We found that the negative removal effect dominated when the intensity of precipitation was weak. With the increase of precipitation intensity, the removal effect of precipitation on PM 2.5 changed from negative to positive values gradually. When the precipitation was heavy, the positive removal effect dominated and reduced the PM 2.5 mass concentration clearly. A likely mechanism has been proposed: The hygroscopic growth of aerosols plays a more (less) important effect than the collision-coalescence between precipitation droplets and aerosols when the precipitation is weak (strong), making PM 2.5 mass concentration increase (decrease). Further analysis using the equal-sample bins of precipitation data confirmed these results. By binning the PM 2.5 mass concentration into three groups with the same sample size, this study found that the negative removal effect dominated for low PM 2.5 mass concentration condition, and positive removal effect dominated for high PM 2.5 mass concentration condition. This was likely due to the increasing collision-coalescence efficiency of precipitation with aerosol particles when aerosol density (so PM 2.5 ) increased. Shortly, the removal effect was higher when the pollution was heavier and precipitation was stronger.
The impacts of wind speed on the removal effect of precipitation were finally discussed using case studies based on wind speed at 100 m above the surface. As expected, a positive relationship was found between removal effect and wind speed. However, negative removal effects were still found for three cases when the wind speed was high, which indicates that the hygroscopic growth effect was prominent for particular cases. We can also expect that the positive removal effect by precipitation when precipitation or pollution is heavy could be enhanced by the accompanied winds.