Distinct Impacts of Light and Heavy Precipitation on PM 2.5 Mass Concentration in Beijing

Using hourly observation data of precipitation and PM 2.5 at 12 sites in Beijing from 2015 to 2017, this study investigates the impacts of different types of precipitation on PM 2.5 mass concentration, along with the characteristics of precipitation and PM 2.5 . There were totally 91 – 123 precipitation events annually, 69.7 – 79.4% of which has precipitation amount less than 5 mm. By investigating the differences of PM 2.5 mass concentration between 1 hr after and before the precipitation events, this study ﬁ nds distinct impacts of different types of precipitation on PM 2.5 mass concentration. For precipitation events with amount of 0.1 – 0.5 mm, PM 2.5 mass concentration increased with precipitation amount with a rate of 0.85 μ g/m 3 per 0.1 mm. For precipitation events with amount of 0.5 – 10 mm, there was no clear relationship between precipitation amount and PM 2.5 mass concentration. For precipitation events with amount larger than 10 mm, PM 2.5 mass concentration decreased with precipitation amount with a rate of 0.17 μ g/m 3 per 1 mm. Further analysis shows that weak precipitation less than 10 mm increased PM 10 , and heavy precipitation larger than 10 mm decreased PM 10 . The aerosol amount also affects the response of PM 2.5 to precipitation, with weak pollution prone to increase with precipitation and heavy pollution prone to decrease with precipitation. Likely mechanisms are discussed, which include the aerosol hygroscopic growth and gas ‐ particle conversion that increase aerosol amount and precipitation scavenging that decreases aerosol amount. Shortly, the mechanisms that increase (decrease) aerosol amount more probably dominate when precipitation is light (heavy).


Introduction
With the acceleration of urbanization, air pollution has become one of the most important ecological and environmental problems in China. Among the various air pollution components, aerosol particles, particularly those with dynamic diameter less than 2.5 μm (PM 2.5 ), whose main pollutant sources are industrial emissions, coal burning, dust, and motor vehicle emissions, have been found as the primary contributor and studied broadly (Huang et al., 2017;Lu et al., 2018;Lv et al., 2017;Qiu et al., 2014;Song et al., 2012;Sun et al., 2018;Wang et al., 2017;Wu et al., 2017;Zhao et al., 2019).
On the other hand, the meteorology, including the planetary boundary layer (PBL), temperature, pressure, relative humidity, wind speed, wind direction, and precipitation, also exerts strong impacts on the PM 2.5 mass concentration (Chen et Li, Feng, et al., 2017;Li, Guo, et al., 2017;Lou et al., 2019;Pearce et al., 2011). The variation of PBL height governs the concentrations of atmospheric pollutants near the surface since the aerosol particles generally concentrate within the PBL height (Garratt, 1994;Guo et al., 2019;Seidel et al., 2010). PM 2.5 mass concentration has negative relationship with temperature and relative humidity while has positive relationship with pressure, by affecting the growth and transport of pollutants (Li, Feng, et al., 2017;Li, Guo, et al., 2017;Yue et al., 2016;Zhao et al., 2014). Wind is a very important meteorological factor influencing pollutants, which influence the horizontal and vertical transport of air pollutants. Moreover, wind speed has different impacts on fine and coarse particles. It was shown that the fine PM concentrations decreased gradually with the increase of wind speed, while coarse PM concentrations would increase due to dust resuspension under strong winds (X. Li, Feng, et al., 2017;B. Zhang, Jiao, et al., 2018;Zhou et al., 2003).
There are also many studies regarding the impacts of precipitation on PM 2.5 . In the natural removal of pollutants, precipitation scavenging efficiency is typically much larger than dry deposition (Garrett et al., 2010), which highly correlates with a variety of atmospheric chemical components and associates with the physical and chemical processes. Precipitation scavenging includes in-cloud and below-cloud scavenging mechanisms. The three most important removal mechanisms in below-cloud-scavenging process are Brownian diffusion, interception, and inertial impaction, which wash out the aerosol particles from the atmosphere (Mircea et al., 2000;Rodhe & Grandell, 1972;Yao et al., 1999). In-cloud-scavenging process is that aerosol particles enter the cloud as cloud condensation nuclei and grow into raindrops that land on the ground. Previous study showed that the scavenging efficiency of urban aerosol is higher than that of rural, remote continental, and marine aerosol, especially remote continental and marine aerosol (Mircea et al., 2000). Dong et al. (2016) andB. Zhang, Jiao, et al. (2018) showed that the efficiency of precipitation scavenging to PM 2.5-10 was more than twice as that to PM 2.5 . The precipitation amount also shows a great influence on scavenging efficiency (Han et al., 2019;Luan et al., 2019;Mircea et al., 2000). Relatively intense precipitation events were more efficient in removing the polluted aerosols in the atmosphere (Andronache, 2003;Chate, 2005;Loosmore & Cederwall, 2004;Luan et al., 2019;Mircea et al., 2000;Yao et al., 1999). In general, the precipitation scavenging efficiency depends on various influential factors including both aerosol and cloud properties. Many studies regarding precipitation scavenging efficiency have focused on the relationship of the particle diameter, size distribution, aerosol types, precipitation intensity, and precipitation scavenging efficiency (Chate, 2005;Hales & Dana, 1978;Han et al., 2019;Luan et al., 2019;Mircea et al., 2000). However, previous studies did not consider the diurnal variation of PM 2.5 mass concentration, which is a very important factor that cannot be ignored. Particularly, there are generally obvious diurnal variation in the PBL, making PM 2.5 likely vary a lot with time in a day, which will make it challenging to identify the precipitation scavenging effect from the natural diurnal variation of PM 2.5 when precipitation occurs. Moreover, few studies have investigated the potential changes about the precipitation effects on PM 2.5 with magnitudes of precipitation and pollution. This study tries to answer this question by analyzing the changes of PM 2.5 between before and after precipitation in Beijing. Beijing (39°26′-41°03′ N, 115°25′-117°30′ E) was chosen here simply because it is the political, economic, and cultural center of China, and its pollution status has been paid close attentions during recent years.
The paper is organized as follows. Section 2 describes the data and method used in this study. Section 3 shows the analysis and results. The summary and discussion are provided in section 4.

Data
This study takes use of the hourly precipitation and PM 2.5 observation data from 2015 to 2017 obtained at 12 sites in Beijing, including Wanshouxigong (WSXG), Dingling (DL), Dongsi (DS), Tiantan (TT), Nongzhanguan (NZG), Guanyuan (GY), Haidian Wanliu (HDWL), Shunyi Xincheng (SYXC), Huairou Town (HRT), Changping Town (CPT), Olympic Sports Center (OSC), and Gucheng (GC), as shown in Figure 1. The hourly China Merged Precipitation Analysis Version 1.0 product is used in this study, which is obtained from the national automatic weather stations and Climate Precipitation Center Morphing product (Tian et al., 2019). This product has a spatial resolution of 0.1°and a temporal resolution of 1 hr in China. The hourly precipitation product is downloaded online (ftp://nwpc.nmc.cn), and the hourly PM 2.5 and PM 10 data at 12 sites are downloaded from http://beijingair.sinaapp. com. Note that the PM 2.5 and PM 10 data website is provided by the China Environmental Monitoring Station of the national air quality realtime release platform with data quality assurance.

Analysis Method
The impact of precipitation on PM 2.5 is studied here by calculating the changes of PM 2.5 between 1 hr after and before the precipitation event, which is where R is the change of PM 2.5 mass concentration between 1 hr after and before precipitation, PM 2.5endtime is the PM 2.5 at time 1 hr after the precipitation, and the PM 2.5starttime is the PM 2.5 at time 1 hr before the precipitation. A precipitation event is defined as the event that precipitation continuously exists in all adjacent hours. Once there is no precipitation for 1 hr or more, new precipitation event will be searched. Similar methods have been used by previous studies. For example, Luan et al. (2019) defined R as the change of PM 2.5 mass concentration between time during and before the rain; Feng and Wang (2012) and Olszowski (2016) defined R as the change of PM 2.5 mass concentration between time after and before the rain. To exclude the impact of natural daily changes of PM 2.5 mass concentration, we only choose precipitation events with duration time no more than 1 hr. It should be more reasonable than the method defined by Luan et al. (2019) due to constraining the precipitation events within 1 hr. The impact of precipitation on PM 10 is also studied with the similar method by calculating the changes of PM 10 between 1 hr after and before the precipitation event, which is where R 10 is the change of PM 10 mass concentration between 1 hr after and before precipitation, PM 10endtime is the PM 10 at time 1 hr after the precipitation, and the PM 10starttime is the PM 10 at time 1 hr before the precipitation.
For the selected precipitation cases, we classify the precipitation into four types according to precipitation amount, which are 0-1, 1-10, 10-20, and 20-50 mm with sample numbers of 1,021, 2,637, 357, and 71, respectively. Noting that there are different sample volumes for different types of precipitation, we further classify each type of precipitation into different number of bins for statistical analysis. Table 1 shows the details of the classification. The changes of PM 2.5 and PM 10 by precipitation are analyzed for every type of precipitation using equations (1) and (2). Figure 1 shows that the annual average precipitation amount in unit of millimeters for the period of 2015-2017 over all sites in Beijing, which was generally 577.6-761.4 mm. The average precipitation amount for all sites during the study period was 639.6 mm. Among all sites, the precipitation at HRT site was the maximum with the annual average precipitation above 700 mm. Compared to other sites, the precipitation amounts at DL and TT sites were relatively low with the annual average precipitation less than 600 mm. In general, the precipitation amount for a precipitation event in the study region was 0.02-331.2 mm. Figure 2 shows the maximum and minimum precipitation amounts for all precipitation events at every site during the study period. There are significant differences in the precipitation amount for the maximum precipitation events among the sites. The minimum precipitation amount of each site was less 0.12 mm, which was generally similar in absolute values among the sites.  Note. For every type of precipitation, it is further classified into 10, 10, 5, and 6 bins, respectively, with different bin width.

Precipitation Characteristics in Beijing
The precipitation amounts for the most serious precipitation events at every station were 121.2-331.2 mm, with the largest value at HRT and the smallest value at SYXC. Figure 3 shows the annual average number of precipitation events at 12 sites, including precipitation events with duration time less than 1 hr and longer than 1 hr. There are 91-123 precipitation events annually in Beijing, with the most frequent precipitation occurrence at WSXG and the least frequent occurrence at GC. For most sites, there were more than 60% short-term precipitation events with duration time less than 1 hr. The all-site average ratio of precipitation events with duration time less than 1 hr to total was 61%, with the largest value of 67% at WSXG.
Figure 4 further shows the normalized probability distribution function of precipitation with different precipitation amount during the period 2015-2017. Around 50% precipitation events were light precipitation with amount less than 1 mm. Specifically, the probability of precipitation with amount less than 1 mm was more than 40% for all 12 sites, and the probability of precipitation with amount less than 1 mm was even more than 50% at NZG, HDWL, and OSC. Accumulatively, the annual average probabilities of precipitation events with amount less than 5 and 10 mm were 69.7-79.4% and 78.7-89.4%, respectively. In contrast, the probabilities of precipitation events with amount larger than 10 and 50 mm were generally less than 20% and 5%, respectively. These results imply that the precipitation in Beijing was dominated by light precipitation events.
We also examine the relationship between precipitation amount and duration time of precipitation events at 12 sites in Beijing for the time period 2015-2017, which is shown in Figure 5. To limit the potential errors from binning method, the data have been analyzed in Figure 5 only when sample volume of precipitation events for each 1-hr duration time bin is more than 50. In general, there was a good positive correlation (r = 0.68, p < 0.05) between the amount of precipitation and the duration time of precipitation. Quantitatively, the precipitation amount increased about 1.5 mm for an increase of 1-hr precipitation duration time when precipitation amount was less than 50 mm.
Figures 2-5 show the characteristics of precipitation in Beijing during the study period, with light precipitation and short-term precipitation dominating: 70% of all precipitation events are with precipitation amount less than 5 mm, and 61% of all precipitation events are with duration time no more than 1 hr. The different types of precipitation events could have different impacts on the PM 2.5 mass concentration, which will be investigated in section 3.3. Figure 6 shows the temporal variations of annual and seasonal average PM 2.5 mass concentrations at 12 sites in Beijing for the time period 2015-2017. For all observation sites, it was clear that the annual average PM 2.5 mass concentration decreased in the recent 3 years, implying the gradually improving air quality in Beijing. Combined with Figure 1, we can find that the PM 2.5 mass concentration in the north of Beijing was obviously lower than that in other areas, which is associated with terrain, meteorological conditions, and ecosystem. For example, mountains lie in the north and west of Beijing while there are plains in the middle and east, which facilitates the transport of pollutants from north to south by valley wind, especially during nighttime (Li et al., 2015;Xin et al., 2010;Zhao et al., 2014;Zhou et al., 2003). Xu et al. (2004) showed that pollutants from cities around  southern Beijing cannot be transported efficiently long distance due to the terrain. Due to the dry air mass from Siberia, the main wind direction in winter is north and northwest in Beijing, which causes the pollutants to be transported to the south of Beijing with the flow . Figure 6 also shows that the seasonally average PM 2.5 mass concentrations decreased with time for all four season. Note that the decreasing trends of PM 2.5 is the strongest in winter and the weakest in summer. The reduction in pollutant emission was the primary cause for the decrease in PM 2.5 mass concentration and other pollutants such as SO 2 and NO 2 (Z. Zhang, Ma, et al., 2018).

Characteristics of PM 2.5 in Beijing
We also investigated the diurnal variation of hourly average PM 2.5 in spring, summer, autumn, and winter during the study period, which is shown in Figure 7. The PM 2.5 mass concentration showed significant different diurnal variations in all seasons, which is associated with anthropogenic activities, atmospheric mixing and dilution by vertical convection, and so on (Chen et al., 2015;Li et al., 2015;Liu et al., 2015Liu et al., , 2006Zhao et al., 2009). For example, the diurnal variation of solar radiation makes corresponding variations of mixed boundary layer height, further causing PM 2.5 mass concentration high at night and low at noon. The combination of anthropogenic activities and mixed boundary layer height diurnal variation could make PM 2.5 diurnal variation different in each season. In winter, the diurnal variation of PM 2.5 mass concentration was the strongest with high values at night and low values in day time. The maximum PM 2.5 mass concentration occurred at around 22:00 local time with a value of 108 μg/m 3 , and the minimum PM 2.5 mass concentration occurred near noon time with a value of 73 μg/m 3 . In contrast, the bimode diurnal variation patterns were found in spring, summer, and autumn, with maximum values at both night and near noon time and minimum values in the morning and evening time. Specifically, the maximum PM 2.5 mass concentration occurred around noon time with a value of 58 μg/m 3 in summer, around 10:00 with a value of 73 μg/m 3 in spring and around 20:00 with a value of 79 μg/m 3 in autumn. The minimum PM 2.5 mass concentration occurred around 18:00 with a value of 49 μg/m 3 in summer, around 18:00 with a value 59 μg/m 3 in spring, and around 7:00 with a value of 60 μg/m 3 in autumn. The bimodal diurnal pattern of PM 2.5 in summer is associated with the solar radiation, which determines the diurnal variation of PBL height, anthropogenic emissions, and hygroscopic growth (Liu et al., 2015;Zhao et al., 2009). The peak value at night should be more associated with the low PBL height at night, while the peak value at noon is more likely caused by the hygroscopic growth of fine aerosols.
We can also find from Figure 7 that the magnitude of diurnal variation of PM 2.5 was the largest in winter and smallest in summer, with that in spring and autumn as between. The magnitude of diurnal variation was slightly larger in autumn than in spring. From spring to winter, the magnitude of diurnal variation was 13.2, 8.6, 18.5, and 35.2 μg/m 3 in turn. In the next section, when we analyze the change of PM 2.5 between before and after precipitation, the natural diurnal variation of PM 2.5 may also contribute to the change of PM 2.5 . In order to reduce the uncertainty caused by diurnal variation of PM 2.5 , we only select precipitation with duration time no more than 1 hr as the research object.

Impacts of Precipitation on PM 2.5
We here investigate the impacts of precipitation on PM 2.5 mass concentration by analyzing the differences of PM 2.5 between 1 hr before and 1 hr after the precipitation. After several tests of the threshold values, we find that 0.5 and 10 mm were the turning points for precipitation impacts on PM 2.5 . The threshold values generally depend on the aerosol type strongly, making them different in various locations (Mircea et al., 2000;Svenningsson et al., 2006). To identify the impacts on PM 2.5 from different types of precipitation, we then The deep green, light green, red, blue, yellow, and purple represent precipitation amount 0~-1mm, 1-~5mm, 5-~10mm, 10-~50mm, 50-~100mm, and 100-~332 mm, respectively.

10.1029/2019EA000717
Earth and Space Science classify the precipitation events into three types based on these two threshold values, which are 0-0.5, 0.5-10, and 10-50 mm in precipitation amount. Figure 8 shows the change of PM 2.5 between 1 hr after and 1 hr before precipitation with precipitation amount. The impact of precipitation on PM 2.5 mass concentration was completely different for three types of precipitation. As shown in Figure 8a, for precipitation with amount 0-0.5 mm, the PM 2.5 mass concentration increased with increasing precipitation amount. This positive linear relationship was significant with r = +0.91 and p value less than 0.05. This observational phenomenon is likely associated with the hygroscopic growth of aerosol particles and the gas-particle conversion (secondary formation) of aerosols. While the precipitation scavenging might reduce the aerosol amount, the weak precipitation limits its significance. Instead, the hygroscopic growth and secondary formation of aerosols dominate the change of PM 2.5 , making it increase with precipitation amount (Su et al., 2016;Wu et al., 2018). Quantitatively, for a 0.1-mm increase of precipitation amount, the PM 2.5 mass concentration increased by 0.85 μg/m 3 .
As shown in Figure 8b, for precipitation with amount 0.5-10 mm, the PM 2.5 mass concentration shows no clear relationship with the precipitation amount. While it seemed that PM 2.5 mass concentration increased with the precipitation amount, the increasing trend was weak, and the correlation was quite poor. It is likely that the precipitation scavenging effect increases with the increasing precipitation amount, which competes with the effect of hygroscopic growth and secondary formation of aerosols (Kulmala et al., 2000(Kulmala et al., , 2004(Kulmala et al., , 2005Zhang et al., 2014;, making the relationship between PM 2.5 and precipitation amount unclear for this type of precipitation. Figure 8c shows that the PM 2.5 mass concentration decreased clearly with increasing precipitation for the precipitation events with amount 10-50 mm. The negative relationship was significant at a confidence level of 95% with a strong correlation coefficient (r = −0.68, p < 0.05) between the PM 2.5 mass concentration and the precipitation amount. For every 10-mm increase of precipitation amount, the PM 2.5 mass concentration was reduced by 1.7 μg/m 3 . For these relatively heavy precipitation events, the precipitation scavenging effect plays the dominant role, making the PM 2.5 mass

10.1029/2019EA000717
Earth and Space Science concentration decrease with precipitation amount (Loosmore & Cederwall, 2004;Luan et al., 2019;Mircea et al., 2000). We use the same method to investigate the impacts of precipitation on PM 10 mass concentration.
After several tests of the threshold values, we find that 10 mm was the turning point for precipitation impacts on PM 10 . We then classify the precipitation events into two types based on this threshold value, which were 0-10 and 10-50 mm in precipitation amount. Figure 9 shows the change of PM 10 between 1 hr after and 1 hr before precipitation with precipitation amount. As shown in Figure 9a, for precipitation with amount 0-10 mm, PM 10 mass concentration roughly increased with the precipitation amount (r = 0.56, p < 0.05). The phenomenon may also be due to the fact that precipitation scavenging effect competes with the effect of hygroscopic growth and secondary formation of aerosols, making the relationship between PM 10 and precipitation amount complicated. Figure 9b shows that the PM 10 mass concentration decreases clearly with increasing precipitation for the precipitation events with amount 10-50 mm. The negative relationship was significant with a strong correlation

10.1029/2019EA000717
Earth and Space Science coefficient (r = −0.95, p < 0.01) between the PM 10 mass concentration and the precipitation amount. For every 10-mm increase of precipitation amount, the PM 10 mass concentration was reduced by 4.0 μg/m 3 , 11% of which was the PM 2.5 mass concentration. This indicates that precipitation is more efficient for the scavenging of large particles, such as PM 2.5-10 (Dong et al., 2016;Loosmore & Cederwall, 2004).

The Effect of PM 2.5 Mass Concentration on Precipitation Scavenging Efficiency
In order to study the relationship between the different effects of precipitation on PM 2.5 and the PM 2.5 mass concentration before precipitation, we selected the PM 2.5 mass concentration data at time 1 hr before precipitation and divided them into four groups: 0 < PM 2.5 ≤ 35, 35 < PM 2.5 ≤ 75, 75 < PM 2.5 ≤ 115, and PM 2.5 > 115 μg/m 3 , representing clean, semiclean, semipollution, and pollution condition, respectively. Under different pollution conditions, the precipitations are also divided into three groups using the two threshold values classified earlier, which are 0-0.5, 0.5-10, and 10-50 mm in precipitation amount. Figure 10 shows the relationship between the PM 2.5 mass concentration before precipitation and the impacts of different types of precipitation on PM 2.5 under different pollution conditions. It shows that the different impacts of precipitation on PM 2.5 were related to the PM 2.5 mass concentration before precipitation. When the PM 2.5 mass concentration before precipitation was ≤35 μg/m 3 , all types of precipitation, even including heavy precipitation, can enhance PM 2.5 mass concentration, with an average increase of PM 2.5 mass concentration by 1.76-5.00 μg/m 3 . For light pollution cases, precipitation is less likely to scavenge aerosol particles by collision due to few aerosol particles under clean conditions (Chate, 2005;Dong et al., 2016). In contrast, the hygroscopic growth of aerosol particles and the gas-particle conversion or secondary formation of aerosols promote PM 2.5 mass concentration (Su et al., 2016;Wu et al., 2018;Zhao, Qiu, et al., 2018). For PM 2.5 mass concentration in the range of 35-75 μg/m 3 , the light precipitation with amount 0-10 mm can enhance PM 2.5 mass concentration with an average increase of PM 2.5 mass concentration by 2.28-3.68 μg/m 3 ; however, the heavy precipitation with amount 10-50 mm can scavenge aerosol particles and make PM 2.5 mass concentration decrease while the scavenging efficiency was relatively low. For the PM 2.5 mass concentration in the range of 75-115 μg/m 3 , the scavenging efficiency also increased with precipitation amount. The scavenging efficiency was negligible under cases with 0-0.5-mm precipitation amount, while significant when precipitation amount was larger than 0.5 mm. When the PM 2.5 mass concentration was >115 μg/m 3 , the scavenging efficiency also increased with precipitation amount. The above results show that the more the aerosols and the heavier the precipitation (for large precipitation), the higher the scavenging efficiency under normal circumstances (Loosmore & Cederwall, 2004).
We should note that precipitation scavenging is indeed related to many factors, such as precipitation amount, precipitation intensity, aerosol particle numbers, and winds. Precipitation is often accompanied by winds, which can also help reduce the PM 2.5 . As indicated earlier, by limiting our study to precipitation with duration time no more than 1 hr, we could reduce uncertainties caused by factors other than precipitation. Of course, uncertainties could still exist, which is beyond the scope of current study. Shortly, this study did not discuss in depth the effects of various meteorological factors on the precipitation scavenging. Instead, we try to find out the statistical relationship between the precipitation amount and PM 2.5 mass concentration from the observational view.

Summary and Discussion
This study first examines the characteristics of precipitation and PM 2.5 and then focuses on the impacts of different types of precipitation on PM 2.5 mass concentration using the hourly observation of precipitation and PM 2.5 at 12 sites in Beijing from 2015 to 2017. It shows that there was 577.6-761.4 mm/year precipitation amount during the study period. Precipitation events were totally 91-123 annually with a big difference in Figure 10. The R value of three types of precipitation on PM2.5 mass concentration during the same PM2.5 mass concentration interval. The R is the change of PM2.5 mass concentration between 1 hour after and before precipitation.

10.1029/2019EA000717
Earth and Space Science precipitation amount per event in Beijing. Almost 50% precipitation events were precipitation with amount less than 1 mm, with the probability of less than 5-mm precipitation between 69.7% and 79.4%. The precipitation in Beijing was dominated by light precipitation events, and more than 60% precipitation events were short-term precipitation with duration time less than 1 hr. The annual average PM 2.5 mass concentration was 57.3-76.6 μg/m 3 during the study period and decreased yearly in Beijing with better air quality in suburban North Beijing area than that in south Beijing region. The seasonal variation of PM 2.5 mass concentration was clear, high in winter and low in summer. Clear diurnal variation of PM 2.5 mass concentration also existed, which was the most obvious in winter.
By investigating the changes of PM 2.5 mass concentration between 1 hr after and before the precipitation events, the impacts of precipitation on PM 2.5 have been investigated. It is found that there are clear differences regarding the impacts of precipitation on PM 2.5 mass concentration for different types of precipitation. For precipitation events with amount of 0.1-0.5 mm, PM 2.5 mass concentration increased with precipitation amount with a rate of 0.85 μg/m 3 per 0.1 mm (r = +0.91, p < 0.05). For precipitation events with amount of 0.5-10 mm, there were no clear relationships between precipitation amount and PM 2.5 mass concentration. For precipitation events with amount larger than 10 mm, PM 2.5 mass concentration significantly decreased with precipitation amount with a rate of 0.17 μg/m 3 per 1 mm. When the precipitation is light, aerosol hygroscopic growth and gas-particle conversion increase aerosol amount. When the precipitation is heavy, precipitation can scavenge aerosol particles more efficiently to decrease PM 2.5 mass concentration.