Fine particulate matter (aerodynamic diameter <2.5 μm; PM2.5) poses risks to human health. While precipitation is the main process for decreasing ambient pollutant concentrations, scavenging of PM2.5 by precipitation remains to be investigated. Here we formulated the processes of PM2.5 scavenging by precipitation from observed PM2.5 concentrations ([PM2.5]) and precipitation intensities. Then we analyzed how changes in precipitation patterns would affect health risks related to PM2.5 on the basis of a Monte Carlo simulation. Tokyo, the capital of Japan, was selected as the target for this study because of its social significance. We found that [PM2.5] decreased significantly through scavenging of PM2.5 from the atmosphere by precipitation. In contrast, we found no significant correlation between reduction of [PM2.5] and precipitation intensity. Our model for estimating the reduction of PM2.5 and the Monte Carlo simulation showed good agreement with observations. Among various changes in potential precipitation patterns, changes in the arithmetic mean of the number of events and/or in precipitation duration were more influential on reduction of [PM2.5] than changes in their standard deviations. Health risks due to PM2.5 will increase with decreases in precipitation duration and occurrence.

cancer risk, climate change, particulate matter 2.5, precipitation

Among various risk factors, exposure to ambient particulate matter (PM) is of global concern for human health. Lim et al. (2012) estimated that 3.2 million people died in 2010 due to exposure to particles with an aerodynamic diameter smaller than 2.5 μm (PM2.5). Because of the serious health risks, sources and environmental processes of PM2.5 and the effectiveness of countermeasures are of interest (e.g., Harrison et al. 1997; Sanderson et al. 2013).

Precipitation decreases ambient pollutant concentrations through scavenging. Schumann et al. (1988) found that the scavenging coefficient of the ambient pollutant concentration depended on rainfall rate. Poissant & Béron (1994) confirmed an inverse relationship between ambient pollutant concentration and rainfall intensity, and found that concentration decreased notably during the initial stage of a rainfall event. Yoo et al. (2014) found that particles with an aerodynamic diameter smaller than 10 μm (PM10) were scavenged more effectively than other ambient pollutants (SO2, NO2, CO, O3). However, to the best of our knowledge, the scavenging of PM2.5 by precipitation has not been investigated, with the exception of a few studies (Feng & Wang 2012; Ouyang et al. 2015) and therefore the scavenging process has not been well formulated. Although model-based studies for the projection of PM2.5 transport and health risks have been carried out (e.g., Goto et al. 2014), the wet deposition of PM2.5 is considered as sub-cloud scavenging, which lacks the optimization of parameters in the relationship between precipitation patterns and decreases in PM2.5 concentration ([PM2.5]). Therefore, the processes of PM2.5 scavenging by precipitation should be investigated to quantify the effects of precipitation on health risks through the removal of PM2.5 from the atmosphere. Furthermore, since climate change may cause changes in precipitation patterns (IPCC 2013), the effects of such changes on health risks related to PM2.5 need to be understood.

The aim of this paper is twofold: to analyze the relationship between [PM2.5] and precipitation patterns and formulate the process of PM2.5 scavenging by precipitation; and to calculate the effect of changes in precipitation patterns on health risks due to PM2.5.

Data and models

Tokyo, the capital of Japan, was selected as the target, because it has a large population (∼13 million people) and the impact of PM2.5 on human health is a social concern (Yorifuji et al. 2005). We obtained and used hourly-monitored PM2.5 data (Bureau of Environment, Tokyo Metropolitan Government 2014; collected at 35.69°N, 139.77°E) and precipitation data (Japan Meteorological Agency 2014; collected at 35.69°N, 139.76°E) collected in Tokyo in 2012. Hourly data were used because of data availability. Characteristics of precipitation were described in Supplementary Material S1 (available online at http://www.iwaponline.com/wst/072/346.pdf). The physical distance between these two stations is about 1 km. Note that the PM2.5 observations have missing values (184 of 8784, i.e., 2.1% of total), and these were excluded in the proceeding calculation. While this procedure may result in underestimation in calculating reduction of PM2.5 (stated below), its effect is negligible.

To evaluate whether precipitation effectively removes PM2.5 from the atmosphere, we estimate decreases in PM2.5 per hour as –ΔC/Δt, where ΔC is the change in [PM2.5] and Δt = 1 hour. We then performed a one-sample two-tailed t-test to evaluate whether the rates of decrease were significantly higher than 0 during both dry and wet weather. Dry and wet hours are 8135 and 465 hours, respectively (the period when PM2.5 data were missing was excluded).

While the precipitation scavenging process has been formulated by a theoretical approach (e.g., Beverland & Crowther (1992) modeled scavenging with the integrated cross sectional area of rain droplets), models with easily available data are preferable in terms of practical use. Schumann et al. (1988) proposed that the rate of decrease in ambient pollutant concentrations depends on rainfall intensity, particularly for large particles. On the other hand, Dickhut & Gustafson (1995) found that this relationship was weaker for particulate organic pollutants than for gaseous organic pollutants, and that precipitation intensity might be less important for removal of particles. Hence, we compared two models:
formula
formula
where = decrease in [PM2.5] between time t − 1 and t, C(t − 1) = [PM2.5] (μg/m3) at time t − 1, P(t) = precipitation intensity (mm/h) at time t, and k is a parameter with values determined as explained below. Note that C(t − 1) and P(t) are antecedent hourly monitored data at time t − 1 and t, respectively. The reason for the difference in the time steps of [PM2.5] and precipitation intensity is that decrease in [PM2.5] at time t is influenced by [PM2.5] at the previous time step and precipitation during the period of t − 1 and t.

Model 1 assumes that the decrease in [PM2.5] depends on both the concentration and the precipitation intensity at each time step. Model 2 assumes that it depends only on the concentration, which indicates that precipitation duration, rather than precipitation intensity, is influential in decreasing [PM2.5]. Parameter k is optimized per precipitation event by fitting the above models and applying the least squares method (we also optimized k for all precipitation events. This result is available in the Supplementary Material S2, online at http://www.iwaponline.com/wst/072/346.pdf). To remove weak precipitation events, we focused on events lasting >2 hours. This allowed us to use plots with at least four points to obtain the parameter k to improve reliability of optimization. Note that the number of removed events with 1 and 2 hour precipitation is 118 (the number of all precipitation events is 191), their total amount of precipitation is 195 mm (12.5% of all precipitation events, 1561 mm), and their total time is 159 hours (22.5% of all precipitation events, 707 hours). Hence, 1 and 2 hour precipitation events are minor in terms of total precipitation and precipitation duration. Correlation coefficients between observed and estimated [PM2.5] were calculated to evaluate the model performance, and we chose the better of the two models for further analyses.

Exposure to PM2.5 was defined as the integration of concentration over time:
formula
where tend is the end of a target year.
Reduction of PM2.5 due to precipitation was defined as the total decrease only during the event (this conservative approach avoids overestimating the effects of scavenging):
formula
where m = a particular precipitation event, n = the number of events in a target year, tm = the duration of event m, and Cm = the initial [PM2.5] just before precipitation event m starts.

Reduction of PM2.5 was divided by the total number of hours in the year to calculate the yearly-average reduction. This reduction was considered as a positive effect of scavenging by precipitation on human health.

Estimation of [PM2.5] under changes in various precipitation patterns

Two precipitation parameters were used to estimate the amount of PM2.5 scavenging: the duration of each event and the number of events in each year. We formulated three types of simulations: observed values in 2012 (‘2012sim’), observed values in 1976–2007 (‘historical’), and scenarios of future precipitation pattern change. Hourly precipitation data in Tokyo in 1976–2007 were obtained from CD-ROM (the Japan Meteorological Agency).

To estimate exposure to PM2.5 in scenarios of precipitation pattern change, we assumed changes of ±10% in both the arithmetic mean and the standard deviation (SD) of the distributions of the duration of each event and the number of events in a year. Climate change is likely to cause changes in the long-term mean (monthly or daily) precipitation and intensification of extreme events (IPCC 2013). However, how they will change is still unknown. Furthermore, in an urban setting such as Tokyo, other factors such as urbanization contribute to changes in precipitation pattern, so it is difficult to set reasonable assumptions. We therefore assumed various scenarios for the duration of each event and the number of events in a year (Table 1) and compared the results of all scenarios.

Table 1

Scenarios for precipitation change. Scenario 0 is the historical simulation (1976–2007) and the others are simulations. SD = standard deviation

ScenariosNumber of events
Precipitation duration
Arithmetic mean (%)SD (%)Arithmetic mean (%)SD (%)
0 ± 0 ± 0 ± 0 ± 0 
1-1 +10 +10 +10 +10 
1-2 −10 −10 −10 −10 
2-1 +10 +10 ± 0 ± 0 
2-2 −10 −10 ± 0 ± 0 
2-3 ± 0 ± 0 +10 +10 
2-4 ± 0 ± 0 −10 −10 
3-1 +10 ± 0 +10 ± 0 
3-2 −10 ± 0 −10 ± 0 
3-3 ± 0 +10 ± 0 +10 
3-4 ± 0 −10 ± 0 −10 
4-1 +10 ± 0 ± 0 ± 0 
4-2 −10 ± 0 ± 0 ± 0 
4-3 ± 0 +10 ± 0 ± 0 
4-4 ± 0 −10 ± 0 ± 0 
4-5 ± 0 ± 0 +10 ± 0 
4-6 ± 0 ± 0 −10 ± 0 
4-7 ± 0 ± 0 ± 0 +10 
4-8 ± 0 ± 0 ± 0 −10 
ScenariosNumber of events
Precipitation duration
Arithmetic mean (%)SD (%)Arithmetic mean (%)SD (%)
0 ± 0 ± 0 ± 0 ± 0 
1-1 +10 +10 +10 +10 
1-2 −10 −10 −10 −10 
2-1 +10 +10 ± 0 ± 0 
2-2 −10 −10 ± 0 ± 0 
2-3 ± 0 ± 0 +10 +10 
2-4 ± 0 ± 0 −10 −10 
3-1 +10 ± 0 +10 ± 0 
3-2 −10 ± 0 −10 ± 0 
3-3 ± 0 +10 ± 0 +10 
3-4 ± 0 −10 ± 0 −10 
4-1 +10 ± 0 ± 0 ± 0 
4-2 −10 ± 0 ± 0 ± 0 
4-3 ± 0 +10 ± 0 ± 0 
4-4 ± 0 −10 ± 0 ± 0 
4-5 ± 0 ± 0 +10 ± 0 
4-6 ± 0 ± 0 −10 ± 0 
4-7 ± 0 ± 0 ± 0 +10 
4-8 ± 0 ± 0 ± 0 −10 
View Large

We conducted a Monte Carlo simulation by randomly choosing values of parameters in order to simulate the reduction of PM2.5 and overall exposure to PM2.5. Since we selected model 2 in the assessment (see ‘Modeling the change in [PM2.5] due to precipitation’ in the Results and discussion section), three values were needed as inputs to the model: initial [PM2.5], k, and precipitation duration. The initial [PM2.5] was randomly selected from observation collected during dry weather in 2012. The duration was also randomly selected from the data assumed in each scenario. Since k and initial [PM2.5] showed a weak but significant correlation (Spearman's rank test, ρ = –0.24, P < 0.05), we divided the distribution of k into four groups based on the median of initial [PM2.5] and precipitation duration as thresholds, to enhance the independence of the values of the three parameters. Next, k was randomly selected according to the corresponding four groups, which was determined from the initial [PM2.5] and duration. We then simulated PM2.5 reduction due to one precipitation event 10,000 times and obtained the distribution of the simulated PM2.5 reduction due to one precipitation event.

Next we estimated reductions of [PM2.5] under the different scenarios. For each scenario we used the number of precipitation events in 1 year (N) and the 10,000-member dataset of PM2.5 reduction for one event. The PM2.5 reduction for one event was extracted N times from the 10,000-member dataset, and the reductions in a year were summed. This computation was also performed 10,000 times. The reduction of PM2.5 in a year in each scenario was averaged over the year (divided by the total number of hours in the year). The annual mean [PM2.5] in each scenario was then calculated by subtracting the median of the reduction of [PM2.5] in a year in each scenario from the sum of the observed annual mean of [PM2.5] and the observed reduction of [PM2.5] in 2012.

Calculation of health risks

We estimated the health risks due to PM2.5 using the method of Yorifuji et al. (2005). We did not consider the effect of location because [PM2.5] was similar among fixed sites and indoor or outdoor residences (Michikawa et al. 2014). First, Pe was calculated as:
formula
where Pe = the expected health outcome frequency at the reference exposure level B (6, 8, or 10 μg/m3), Po = the observed health outcome frequency (8.3 per 1000 people in Tokyo in 2012; Ministry of Health, Labour and Welfare 2012), RR = the relative risk of all-cause mortality associated with a 10 μg/m3 increase in PM2.5 (1.14; Lepeule et al. 2012: 1974–2009), and O = the observed [PM2.5] (μg/m3). Although RR depends on location, we adopted 1.14 because similar values were found in several cohort surveys (Lepeule et al. 2012). One in Japan reported RR = 1.24 for lung cancer mortality (Katanoda et al. 2011), but we did not use this value because the survey period was only 10 years. Since Lepeule et al. (2012) showed that the concentration–response relationship was linear down to [PM2.5] = 8 μg/m3, we used values of B similar to those reported previously (Lim et al. 2012: 5.8­–8.8 μg/m3).
The annual mortality risk due to PM2.5 under each scenario was calculated as:
formula
where Ps = the annual mortality risk, and Es = the estimated [PM2.5] (μg/m3).

Modeling the change in [PM2.5] due to precipitation

[PM2.5] decreased after a precipitation event (Figure 1). Here we exemplify the result in July, when Japan has its rainy season, so that a decrease in [PM2.5] can be clearly detected. It also fluctuated in the absence of precipitation because of variations in the outputs of PM2.5 from biomass burning, industry, and motor vehicles (e.g., Song et al. 2006; Zhang et al. 2013).

Figure 1
Figure 1. Precipitation and [PM2.5] in Tokyo in July 2012. Gaps represent missing data.
View largeDownload slide

Precipitation and [PM2.5] in Tokyo in July 2012. Gaps represent missing data.

Figure 1
Figure 1. Precipitation and [PM2.5] in Tokyo in July 2012. Gaps represent missing data.
View largeDownload slide

Precipitation and [PM2.5] in Tokyo in July 2012. Gaps represent missing data.

Close modal

During dry weather (0 mm/h), the rate of decrease of [PM2.5] did not deviate from zero (one-sample, two-tailed t-test, P > 0.05; Figure 2). However, during wet weather, [PM2.5] decreased significantly (P < 0.001), similar to other studies (Feng & Wang 2012; Ouyang et al. 2015). This shows that precipitation scavenges PM2.5 from the atmosphere, with potentially positive effects on human health. Interestingly, there were no significant correlations between reduction of [PM2.5] and precipitation intensity during wet weather (Spearman's rank test, ρ = 0.003, P > 0.05). The extent of decrease did not differ greatly among precipitation intensities (Figure 2). Note that the monitoring time step of 1 hour may have an unidentifiable effect on the analysis.

Figure 2
Figure 2. Change in [PM2.5] per hour by different precipitation intensities per hour. The black lines located in the middle of boxes indicate median values. Bottom of box, 25th percentile (Q1); horizontal line in middle of box, 50th percentile; top of box, 75th percentile (Q3). Minimum and maximum values represent Q1 – 1.5 (Q3 – Q1) and Q3 + 1.5 (Q3 – Q1), respectively. The numbers of precipitation for each intensity are: 8008 (0 mm/h), 141 (1 mm), 167 (2 mm/h), 33 (3 mm/h), 53 (4 mm/h), 12 (5 mm/h), 19 (6 mm/h), 7 (7 mm/h), 7 (8 mm/h), 10 (9, 10 mm/h), 10 (11–15 mm/h), 9 (16–26 mm/h). Note that the sum of these numbers (8476) is not equal to the number of available [PM2.5] data (8784–184 = 8600). This is because one missing [PM2.5] datum at time t makes it impossible to calculate decrease in [PM2.5] due to precipitation at time t and t + 1, which cause the above discrepancy in the precipitation numbers.
View largeDownload slide

Change in [PM2.5] per hour by different precipitation intensities per hour. The black lines located in the middle of boxes indicate median values. Bottom of box, 25th percentile (Q1); horizontal line in middle of box, 50th percentile; top of box, 75th percentile (Q3). Minimum and maximum values represent Q1 – 1.5 (Q3 – Q1) and Q3 + 1.5 (Q3 – Q1), respectively. The numbers of precipitation for each intensity are: 8008 (0 mm/h), 141 (1 mm), 167 (2 mm/h), 33 (3 mm/h), 53 (4 mm/h), 12 (5 mm/h), 19 (6 mm/h), 7 (7 mm/h), 7 (8 mm/h), 10 (9, 10 mm/h), 10 (11–15 mm/h), 9 (16–26 mm/h). Note that the sum of these numbers (8476) is not equal to the number of available [PM2.5] data (8784–184 = 8600). This is because one missing [PM2.5] datum at time t makes it impossible to calculate decrease in [PM2.5] due to precipitation at time t and t + 1, which cause the above discrepancy in the precipitation numbers.

Figure 2
Figure 2. Change in [PM2.5] per hour by different precipitation intensities per hour. The black lines located in the middle of boxes indicate median values. Bottom of box, 25th percentile (Q1); horizontal line in middle of box, 50th percentile; top of box, 75th percentile (Q3). Minimum and maximum values represent Q1 – 1.5 (Q3 – Q1) and Q3 + 1.5 (Q3 – Q1), respectively. The numbers of precipitation for each intensity are: 8008 (0 mm/h), 141 (1 mm), 167 (2 mm/h), 33 (3 mm/h), 53 (4 mm/h), 12 (5 mm/h), 19 (6 mm/h), 7 (7 mm/h), 7 (8 mm/h), 10 (9, 10 mm/h), 10 (11–15 mm/h), 9 (16–26 mm/h). Note that the sum of these numbers (8476) is not equal to the number of available [PM2.5] data (8784–184 = 8600). This is because one missing [PM2.5] datum at time t makes it impossible to calculate decrease in [PM2.5] due to precipitation at time t and t + 1, which cause the above discrepancy in the precipitation numbers.
View largeDownload slide

Change in [PM2.5] per hour by different precipitation intensities per hour. The black lines located in the middle of boxes indicate median values. Bottom of box, 25th percentile (Q1); horizontal line in middle of box, 50th percentile; top of box, 75th percentile (Q3). Minimum and maximum values represent Q1 – 1.5 (Q3 – Q1) and Q3 + 1.5 (Q3 – Q1), respectively. The numbers of precipitation for each intensity are: 8008 (0 mm/h), 141 (1 mm), 167 (2 mm/h), 33 (3 mm/h), 53 (4 mm/h), 12 (5 mm/h), 19 (6 mm/h), 7 (7 mm/h), 7 (8 mm/h), 10 (9, 10 mm/h), 10 (11–15 mm/h), 9 (16–26 mm/h). Note that the sum of these numbers (8476) is not equal to the number of available [PM2.5] data (8784–184 = 8600). This is because one missing [PM2.5] datum at time t makes it impossible to calculate decrease in [PM2.5] due to precipitation at time t and t + 1, which cause the above discrepancy in the precipitation numbers.

Close modal

Figure 3 illustrates the results of model calibration using a precipitation event on 14 July 2012 as an example (this event was taken from 73 precipitation events. The reason for the choice of this event is that its precipitation duration (6 hours) is the median of all precipitation events with >2 hours precipitation). Both models successfully (R ≥ 0.98) represented the decrease of [PM2.5] by precipitation. Figure 4 compares the reproducibility of estimated [PM2.5] per hour between models. Statistical measures of parameter k, which was optimized for each event, gave a relative SD of 1.84 for model 1 and 1.07 for model 2 (Figure 4(a), (b)). The relative SD in model 2 was smaller than that in model 1. R was 0.911 for model 1 and 0.932 for model 2 (Figure 4(c), (d)), indicating that the inclusion of precipitation intensity did not improve the representation of PM2.5 scavenging by precipitation. This result indicates that the duration of precipitation contributes more than intensity to decreasing [PM2.5]. This is supported by the fact that the longer the precipitation duration, the larger the decrease in [PM2.5] (see Supplementary Material S1 and Figure S3, available online at http://www.iwaponline.com/wst/072/346.pdf). Furthermore, model 2 is simpler than model 1 with regard to their parameters. Thus, we selected model 2 for the calculation of health risk. Even if we optimize k for all precipitation events, the correlation coefficient of model 2 was slightly larger than model 1 in accordance with results shown above (see Figure S4 in Supplementary Material S2, online at http://www.iwaponline.com/wst/072/346.pdf).

Figure 3
Figure 3. Examples of the model calibration for changes in [PM2.5] in the case of the precipitation event on 14 July 2012: (a) model 1, (b) model 2. Bars show precipitation.
View largeDownload slide

Examples of the model calibration for changes in [PM2.5] in the case of the precipitation event on 14 July 2012: (a) model 1, (b) model 2. Bars show precipitation.

Figure 3
Figure 3. Examples of the model calibration for changes in [PM2.5] in the case of the precipitation event on 14 July 2012: (a) model 1, (b) model 2. Bars show precipitation.
View largeDownload slide

Examples of the model calibration for changes in [PM2.5] in the case of the precipitation event on 14 July 2012: (a) model 1, (b) model 2. Bars show precipitation.

Close modal
Figure 4
Figure 4. Comparison between models of reproducibility of all precipitation events with >2 h precipitation in 2012. (a), (b) Distribution of k and (c), (d) correlation between observed and estimated [PM2.5] by (a), (c) model 1 and (b), (d) model 2. Each plot indicates observed and estimated [PM2.5] data per hour. R in each figure indicates correlation coefficient.
View largeDownload slide

Comparison between models of reproducibility of all precipitation events with >2 h precipitation in 2012. (a), (b) Distribution of k and (c), (d) correlation between observed and estimated [PM2.5] by (a), (c) model 1 and (b), (d) model 2. Each plot indicates observed and estimated [PM2.5] data per hour. R in each figure indicates correlation coefficient.

Figure 4
Figure 4. Comparison between models of reproducibility of all precipitation events with >2 h precipitation in 2012. (a), (b) Distribution of k and (c), (d) correlation between observed and estimated [PM2.5] by (a), (c) model 1 and (b), (d) model 2. Each plot indicates observed and estimated [PM2.5] data per hour. R in each figure indicates correlation coefficient.
View largeDownload slide

Comparison between models of reproducibility of all precipitation events with >2 h precipitation in 2012. (a), (b) Distribution of k and (c), (d) correlation between observed and estimated [PM2.5] by (a), (c) model 1 and (b), (d) model 2. Each plot indicates observed and estimated [PM2.5] data per hour. R in each figure indicates correlation coefficient.

Close modal

Estimation of [PM2.5] under precipitation pattern changes and calculation of health risks

Observations in 2012 gave an annual mean [PM2.5] of 14.70 μg/m3 and a yearly-average reduction of 0.39 μg/m3. The reduction due to precipitation was 2.6% (=0.39/(14.70 + 0.39)) of total possible exposure in the absence of rain.

Table 2 summarizes the distribution of the yearly-average reduction of [PM2.5], the change in reduction relative to the baseline scenario (Scenario 0—historical), and annual mean [PM2.5] calculated from the Monte Carlo simulation. Scenario 2012sim gave a yearly average reduction of [PM2.5] with a median of 0.36 μg/m3 and an arithmetic mean of 0.35 μg/m3, close to the observed value of 0.39 μg/m3, indicating that model 2 and the Monte Carlo simulation successfully estimated the reduction of [PM2.5].

Table 2

Estimated reduction of [PM2.5], change relative to the baseline scenario, annual mean [PM2.5], annual mortality risk due to PM2.5, and change in risk relative to baseline scenario. See Table 1 for details in each scenario

ScenariosYearly-average reduction of PM2.5 concentration (μg/m3)
Reduction of total PM2.5 (%)aChange in reduction relative to baseline (%)bAnnual mean concentration of PM2.5 (μg/m3)cAnnual mortality risks due to PM2.5 (10−4)dIncrease of annual mortality risk to baseline (10−6)b
5%25%50%75%95%Arithmetic mean
2012sim 0.17 0.29 0.36 0.42 0.52 0.35 2.3 – 14.73 7.15 (5.16–9.05) – 
0 0.14 0.22 0.27 0.33 0.42 0.24 1.6 – 14.82 7.24 (5.25–9.13) – 
1-1 0.18 0.30 0.37 0.44 0.56 0.33 2.2 34.7 14.72 7.14 (5.15–9.03) −10 
1-2 0.12 0.17 0.21 0.26 0.33 0.21 1.4 −22.0 14.88 7.31 (5.32–9.19) 6.4 
2-1 0.14 0.24 0.30 0.36 0.47 0.28 1.9 10.3 14.79 7.21 (5.22–9.10) −3.0 
2-2 0.13 0.20 0.25 0.30 0.39 0.24 1.6 −9.0 14.84 7.27 (5.28–9.16) 2.6 
2-3 0.18 0.27 0.34 0.40 0.52 0.33 2.2 22.5 14.76 7.18 (5.18–9.07) −6.5 
2-4 0.12 0.19 0.24 0.29 0.37 0.23 1.5 −13.7 14.85 7.28 (5.29–9.17) 4.0 
3-1 0.17 0.29 0.36 0.42 0.53 0.31 2.1 29.9 14.73 7.15 (5.16–9.05) −8.7 
3-2 0.10 0.17 0.21 0.26 0.35 0.19 1.3 −22.3 14.88 7.31 (5.32–9.20) 6.5 
3-3 0.14 0.24 0.30 0.36 0.47 0.29 1.9 8.7 14.79 7.22 (5.23–9.11) −2.5 
3-4 0.14 0.22 0.27 0.32 0.41 0.26 1.7 −1.7 14.82 7.25 (5.26–9.14) 0.5 
4-1 0.16 0.25 0.30 0.36 0.46 0.29 1.9 10.8 14.79 7.21 (5.22–9.10) −3.1 
4-2 0.12 0.20 0.25 0.30 0.39 0.21 1.4 −10.2 14.84 7.27 (5.28–9.16) 3.0 
4-3 0.13 0.22 0.28 0.34 0.44 0.27 1.8 1.1 14.81 7.24 (5.25–9.13) −0.3 
4-4 0.13 0.23 0.28 0.34 0.43 0.27 1.8 2.7 14.81 7.23 (5.24–9.13) −0.8 
4-5 0.17 0.27 0.33 0.39 0.50 0.32 2.1 19.9 14.76 7.18 (5.19–9.08) −5.8 
4-6 0.11 0.19 0.24 0.29 0.38 0.22 1.5 −12.5 14.85 7.28 (5.29–9.17) 3.6 
4-7 0.13 0.24 0.30 0.36 0.47 0.28 1.8 8.4 14.79 7.22 (5.23–9.11) −2.4 
4-8 0.14 0.22 0.27 0.32 0.41 0.26 1.8 −2.2 14.82 7.25 (5.26–9.14) 0.6 
ScenariosYearly-average reduction of PM2.5 concentration (μg/m3)
Reduction of total PM2.5 (%)aChange in reduction relative to baseline (%)bAnnual mean concentration of PM2.5 (μg/m3)cAnnual mortality risks due to PM2.5 (10−4)dIncrease of annual mortality risk to baseline (10−6)b
5%25%50%75%95%Arithmetic mean
2012sim 0.17 0.29 0.36 0.42 0.52 0.35 2.3 – 14.73 7.15 (5.16–9.05) – 
0 0.14 0.22 0.27 0.33 0.42 0.24 1.6 – 14.82 7.24 (5.25–9.13) – 
1-1 0.18 0.30 0.37 0.44 0.56 0.33 2.2 34.7 14.72 7.14 (5.15–9.03) −10 
1-2 0.12 0.17 0.21 0.26 0.33 0.21 1.4 −22.0 14.88 7.31 (5.32–9.19) 6.4 
2-1 0.14 0.24 0.30 0.36 0.47 0.28 1.9 10.3 14.79 7.21 (5.22–9.10) −3.0 
2-2 0.13 0.20 0.25 0.30 0.39 0.24 1.6 −9.0 14.84 7.27 (5.28–9.16) 2.6 
2-3 0.18 0.27 0.34 0.40 0.52 0.33 2.2 22.5 14.76 7.18 (5.18–9.07) −6.5 
2-4 0.12 0.19 0.24 0.29 0.37 0.23 1.5 −13.7 14.85 7.28 (5.29–9.17) 4.0 
3-1 0.17 0.29 0.36 0.42 0.53 0.31 2.1 29.9 14.73 7.15 (5.16–9.05) −8.7 
3-2 0.10 0.17 0.21 0.26 0.35 0.19 1.3 −22.3 14.88 7.31 (5.32–9.20) 6.5 
3-3 0.14 0.24 0.30 0.36 0.47 0.29 1.9 8.7 14.79 7.22 (5.23–9.11) −2.5 
3-4 0.14 0.22 0.27 0.32 0.41 0.26 1.7 −1.7 14.82 7.25 (5.26–9.14) 0.5 
4-1 0.16 0.25 0.30 0.36 0.46 0.29 1.9 10.8 14.79 7.21 (5.22–9.10) −3.1 
4-2 0.12 0.20 0.25 0.30 0.39 0.21 1.4 −10.2 14.84 7.27 (5.28–9.16) 3.0 
4-3 0.13 0.22 0.28 0.34 0.44 0.27 1.8 1.1 14.81 7.24 (5.25–9.13) −0.3 
4-4 0.13 0.23 0.28 0.34 0.43 0.27 1.8 2.7 14.81 7.23 (5.24–9.13) −0.8 
4-5 0.17 0.27 0.33 0.39 0.50 0.32 2.1 19.9 14.76 7.18 (5.19–9.08) −5.8 
4-6 0.11 0.19 0.24 0.29 0.38 0.22 1.5 −12.5 14.85 7.28 (5.29–9.17) 3.6 
4-7 0.13 0.24 0.30 0.36 0.47 0.28 1.8 8.4 14.79 7.22 (5.23–9.11) −2.4 
4-8 0.14 0.22 0.27 0.32 0.41 0.26 1.8 −2.2 14.82 7.25 (5.26–9.14) 0.6 

aReduction of total possible exposure in the absence of precipitation. Arithmetic mean of PM2.5 reduction for each scenario (14.70 + 0.39).

bBaseline is scenario 0 (historical).

cMedian values were used to estimate annual mean concentrations.

dValues represent estimates from reference exposure level = 8 μg/m3 (10–6 μg/m3).

View Large

Using precipitation data recorded in 1976–2007, we simulated historical reductions and prospective scenarios of precipitation patterns. While annual means of [PM2.5] are not much different among scenarios (ranging between 14.7–14.9), changes in annual means of [PM2.5] were not negligible among scenarios. Scenario 1-1 (mean and SD of the number of events and duration increase by 10%) increased scavenging by 34.7%. In contrast, Scenario 3-2 (mean of the number of events and duration decrease by 10%) reduced scavenging by 22.0%. Changes in the means of the number of events and/or duration influenced the reduction of [PM2.5] more than changes in their SDs (Scenarios 3-1 vs 3-3, 3-2 vs 3-4, 4-1 vs 4-3, 4-2 vs 4-4, 4-5 vs 4-7, and 4-6 vs 4-8). Decreases in the number of events and in duration made similar contributions to the reduction of [PM2.5] (Scenarios 3-2 vs 4-2 vs 4-6), whereas the increase in duration reduced [PM2.5] more than the increase in the number of events (Scenarios 3-1 vs 4-1 vs 4-5). An increase in the SD of the number of events did not greatly affect the reduction of [PM2.5] (Scenario 4-3), but increases in the SD of duration reduced [PM2.5] substantially (Scenarios 3-3 and 4-7).

Table 2 also shows the estimated annual mortality risks due to PM2.5 and changes in risk relative to the baseline scenario. The annual mortality risk in 2012sim was estimated as 7.15 × 10–4, comparable to preventable risks due to PM2.5 in another study (Yorifuji et al. 2005). As pointed out before (e.g., Yorifuji et al. 2005; Katanoda et al. 2011; Lepeule et al. 2012; Lim et al. 2012; Sanderson et al. 2013), the health risks due to exposure to PM2.5 are serious. Scenario 3-2 (10% decrease in means) and 1-2 (10% decrease in means and SDs) gave the greatest increase in annual mortality risk (6.5 × 10–6 and 6.4 × 10–6) in comparison with the baseline scenario. Scenario 4-2 and 4-6 (10% decrease in means of either the number of events or precipitation duration) gave relatively large increase (3.0 × 10–6 and 3.6 × 10–6). Thus, decreases in the arithmetic means of the number of events and/or duration may increase the health risks due to PM2.5.

We confirmed statistically that precipitation reduces [PM2.5] in the atmosphere. Our formulation of [PM2.5] decrease gave results consistent with observations. The decrease depended more on the precipitation duration than on intensity. Among various scenarios of precipitation pattern changes, 10% decreases in the arithmetic mean of the number of events and in duration increased the annual mortality risk by 6.5 × 10–6.

Beverland
I. J.
Crowther
J. M.
1992
On the interpretation of event and sub-event rainfall chemistry
.
Environmental Pollution
75
,
163
–
174
.
Google Scholar
Crossref
Search ADS
PubMed
 
Bureau of Environment, Tokyo Metropolitan Government
.
2014
measurement results of air pollutants in Tokyo. http://www.kankyo.metro.tokyo.jp/en/index.html (accessed 20 May 2014) (in Japanese).
Dickhut
R. M.
Gustafson
K. E.
1995
Atmospheric washout of polycyclic aromatic hydrocarbons in the southern Chesapeake Bay region
.
Environmental Science & Technology
29
,
1518
–
1525
.
Google Scholar
Crossref
Search ADS
PubMed
 
Goto
D.
Dai
T.
Satoh
M.
Tomita
H.
Uchida
J.
Misawa
S.
Inoue
T.
Tsuruta
H.
Ueda
K.
Ng
C. F. S.
Takami
A.
Sugimoto
N.
Shimizu
A.
Ohara
T.
Nakajima
T.
2014
Application of a global nonhydrostatic model with a stretched-grid system to regional aerosol simulations around Japan
.
Geoscientific Model Development Discussions
7
,
131
–
179
.
Google Scholar
Crossref
Search ADS
 
Harrison
R. M.
Deacon
A. R.
Jones
M. R.
Appleby
R. S.
1997
Sources and processes affecting concentrations of PM10 and PM2.5 particulate matter in Birmingham (UK)
.
Atmospheric Environment
31
,
4103
–
4117
.
Google Scholar
Crossref
Search ADS
 
IPCC
2013
Summary for Policymakers
. In:
Climate Change 2013 The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
(
Stocker
T.F.
Qin
D.
Plattner
G.-K.
Tignor
M.
Allen
S. K.
Boschung
J.
Nauels
A.
Xia
Y.
Bex
V.
Midgley
P. M.
, eds).
Cambridge University Press
,
Cambridge, NY
,
USA
.
Google Scholar
 
Japan Meteorological Agency
.
2014
Search for past climate data. http://www.data.jma.go.jp/obd/stats/etrn/index.php (accessed 20 May 2014) (in Japanese).
Katanoda
K.
Sobue
T.
Satoh
H.
Tajima
K.
Suzuki
T.
Nakatsuka
H.
Takezaki
T.
Nakayama
T.
Nitta
H.
Tanabe
K.
Tominaga
S.
2011
An association between long-term exposure to ambient air pollution and mortality from lung cancer and respiratory diseases in Japan
.
Journal of Epidemiology
21
(
2
),
132
–
143
.
Google Scholar
Crossref
Search ADS
PubMed
 
Lim
S. S.
Vos
T.
Flaxman
A. D.
Danaei
G.
Shibuya
K.
Adair-Rohani
H.
Amann
M.
Anderson
H. R.
Andrews
K. G.
Aryee
M.
Atkinson
C.
Bacchus
L. J.
Bahalim
A. N.
Balakrishnan
K.
Balmes
J.
Barker-Collo
S.
Baxter
A.
Bell
M. L.
Blore
J. D.
Blyth
F.
Bonner
C.
Borges
G.
Bourne
R.
Boussinesq
M.
Brauer
M.
Brooks
P.
Bruce
N. G.
Brunekreef
B.
Bryan-Hancock
C.
Bucello
C.
Buchbinder
R.
Bull
F.
Burnett
R. T.
Byers
T. E.
Calabria
B.
Carapetis
J.
Carnahan
E.
Chafe
Z.
Charlson
F.
Chen
H.
Chen
J. S.
Cheng
A. T.-A.
Child
J. C.
Cohen
A.
Colson
K. E.
Cowie
B. C.
Darby
S.
Darling
S.
Davis
A.
Degenhardt
L.
Dentener
F.
Des Jarlais
D. C.
Devries
K.
Dherani
M.
Ding
E. L.
Dorsey
E. R.
Driscoll
T.
Edmond
K.
Ali
S. E.
Engell
R. E.
Erwin
P. J.
Fahimi
S.
Falder
G.
Farzadfar
F.
Ferrari
A.
Finucane
M. M.
Flaxman
S.
Fowkes
F. G. R.
Freedman
G.
Freeman
M. K.
Gakidou
E.
Ghosh
S.
Giovannucci
E.
Gmel
G.
Graham
K.
Grainger
R.
Grant
B.
Gunnell
D.
Gutierrez
H. R.
Hall
W.
Hoek
H. W.
Hogan
A.
Hosgood
H. D.
III
Hoy
D.
Hu
H.
Hubbell
B. J.
Hutchings
S. J.
Ibeanusi
S. E.
Jacklyn
G. L.
Jasrasaria
R.
Jonas
J. B.
Kan
H.
Kanis
J. A.
Kassebaum
N.
Kawakami
N.
Khang
Y.-H.
Khatibzadeh
S.
Khoo
J.-P.
Kok
C.
Laden
F.
Lalloo
R.
Lan
Q.
Lathlean
T.
Leasher
J. L.
Leigh
J.
Li
Y.
Lin
J. K.
Lipshultz
S. E.
London
S.
Lozano
R.
Lu
Y.
Mak
J.
Malekzadeh
R.
Mallinger
L.
Marcenes
W.
March
L.
Marks
R.
Martin
R.
McGale
P.
McGrath
J.
Mehta
S.
Mensah
G. A.
Merriman
T. R.
Micha
R.
Michaud
C.
Mishra
V.
Hanafiah
K. M.
Mokdad
A. A.
Morawska
L.
Mozaffarian
D.
Murphy
T.
Naghavi
M.
Neal
B.
Nelson
P. K.
Miquel Nolla
J.
Norman
R.
Olives
C.
Omer
S. B.
Orchard
J.
Osborne
R.
Ostro
B.
Page
A.
Pandey
K. D.
Parry
C. D. H.
Passmore
E.
Patra
J.
Pearce
N.
Pelizzari
P. M.
Petzold
M.
Phillips
M. R.
Pope
D.
Pope
C. A.
III
Powles
J.
Rao
M.
Razavi
H.
Rehfuess
E. A.
Rehm
J. T.
Ritz
B.
Rivara
F. P.
Roberts
T.
Robinson
C.
Rodriguez-Portales
J. A.
Romieu
I.
Room
R.
Rosenfeld
L. C.
Roy
A.
Rushton
L.
Salomon
J. A.
Sampson
U.
Sanchez-Riera
L.
Sanman
E.
Sapkota
A.
Seedat
S.
Shi
P.
Shield
K.
Shivakoti
R.
Singh
G. M.
Sleet
D. A.
Smith
E.
Smith
K. R.
Stapelberg
N. J. C.
Steenland
K.
Stoeckl
H.
Stovner
L. J.
Straif
K.
Straney
L.
Thurston
G. D.
Tran
J. H.
Van Dingenen
R.
van Donkelaar
A.
Veerman
J. L.
Vijayakumar
L.
Weintraub
R.
Weissman
M. M.
White
R. A.
Whiteford
H.
Wiersma
S. T.
Wilkinson
J. D.
Williams
H. C.
Williams
W.
Wilson
N.
Woolf
A. D.
Yip
P.
Zielinski
J. M.
Lopez
A. D.
Murray
C. J. L.
Ezzati
M.
2012
A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010
.
Lancet
380
,
2224
–
2260
.
Google Scholar
Crossref
Search ADS
PubMed
 
Ministry of Health, Labour and Welfare
.
2012
Demographic statistics in 2012. http://www.mhlw.go.jp/toukei/saikin/hw/jinkou/kakutei12/index.html (accessed 4 September 2014).
Ouyang
W.
Guo
B.
Cai
G.
Li
Q.
Han
S.
Liu
B.
Liu
X.
2015
The washing effect of precipitation on particulate matter and the pollution dynamics of rainwater in downtown Beijing
.
Science of The Total Environment
505
,
306
–
314
.
Google Scholar
Crossref
Search ADS
PubMed
 
Poissant
L.
Béron
P.
1994
Parameterized rainwater quality model in urban environment
.
Atmospheric Environment
28
,
305
–
310
.
Google Scholar
Crossref
Search ADS
 
Sanderson
W.
Striessnig
E.
Schopp
W.
Amann
M.
2013
Effects on well-being of investing in cleaner air in India
.
Environmental Science & Technology
47
,
13222
–
13229
.
Google Scholar
Crossref
Search ADS
PubMed
 
Song
Y.
Zhang
Y.
Xie
S.
Zeng
L.
Zheng
M.
Salmon
L. G.
Shao
M.
Slanina
S.
2006
Source apportionment of PM2.5 in Beijing by positive matrix factorization
.
Atmospheric Environment
40
,
1526
–
1537
.
Google Scholar
Crossref
Search ADS
 
Yoo
J.-M.
Lee
Y.-R.
Kim
D.
Jeong
M.-J.
Stockwell
W. R.
Kundu
P. K.
Oh
S.-M.
Shin
D.-B.
Lee
S.-J.
2014
New indices for wet scavenging of air pollutants (O3, CO, NO2, SO2, and PM10) by summertime rain
.
Atmospheric Environment
82
,
226
–
237
.
Google Scholar
Crossref
Search ADS
 
Yorifuji
T.
Yamamoto
E.
Tsuda
T.
Kawakami
N.
2005
Health impact assessment of particulate matter in Tokyo, Japan
.
Archives of Environmental & Occupational Health
60
(
4
),
179
–
185
.
Google Scholar
Crossref
Search ADS
PubMed
 
Zhang
R.
Jing
J.
Tao
J.
Hsu
S.-C.
Wang
G.
Cao
J.
Lee
C. S. L.
Zhu
L.
Chen
Z.
Zhao
Y.
Shen
Z.
2013
Chemical characterization and source apportionment of PM2.5 in Beijing: seasonal perspective
.
Atmospheric Chemistry and Physics
13
,
7053
–
7074
.
Google Scholar
Crossref
Search ADS
 
© IWA Publishing 2015

Supplementary data

SUPPLEMENTARY MATERIAL- pdf file