Introduction

Atmospheric aerosols in the accumulation-mode size range of approximately 0.05–1 μm greatly affect Earth’s radiation budget directly by scattering and absorbing sunlight and indirectly by modifying the properties of clouds1,2. The key factors that affect these climate impacts, including the number and mass concentrations, lifetimes, vertical distributions and long-range transport of accumulation-mode aerosols, are commonly regulated by the removal process associated with cloud and precipitation (wet removal)3,4. Therefore, a fundamental understanding and quantitative numerical modelling of the wet removal process are critical for evaluating aerosol’s climate forcing, which is the major scientific uncertainty in predicting human-induced climate change in this century5,6.

Incorporation of an aerosol particle into a water droplet (scavenging) in a cloud-and-precipitation system can occur via two distinct physical mechanisms: nucleation scavenging and impaction scavenging3 (see Fig. 1). The former mechanism refers to an aerosol (cloud condensation nuclei: CCN) inducing formation of a cloud droplet in supersaturated water vapour and the latter mechanism refers to collision-coalescence of an aerosol and a water droplet. Scavenged aerosols are irreversibly removed from the atmosphere if the water droplets gravitationally fall and reach the ground.

Figure 1
figure 1

A diagram of the wet removal process for aerosols and our observational method.

During the moist convection process, some number fractions of aerosols suspended in an ascending parcel are incorporated into water droplets through nucleation or impaction scavenging and subsequently removed from the atmosphere via precipitation. The mass equivalent diameter DBC of each BC-containing particle is invariant throughout the wet removal process. The removal efficiency RE(DBC) is defined as the ratio of the DBC-resolved number concentrations in rainwater to air. In our method, the observed DBC dependences of the CCN fraction FCCN(DBC,SS) and removal efficiency RE(DBC) are compared. See the text for details.

Full size image

According to theoretical studies using numerical models, the number concentration of accumulation-mode aerosols can be reduced by up to 94% by nucleation scavenging7,8. However, the relative importance of the nucleation and impaction scavenging mechanisms on wet removal in actual precipitating clouds remains unknown because of the lack of observational methods. Although a comparison of the chemical compositions of filter-sampled aerosols and collected rainwater indicated a higher wet removal efficiency of more hygroscopic aerosol compounds9,10, such a bulk chemical analysis is unable to confirm the detailed relationships between the particle size and wet removal of aerosols. Recent aircraft observations found that the size of black carbon aerosols were smaller in an air mass that had experienced a greater amount of precipitation, implying a substantial particle-size dependence of wet removal11,12. Despite these observational implications of causal links between the microphysical properties of aerosols and their wet removal efficiency, no quantitative data have been available for identifying the critical mechanism of wet removal.

In this study, we introduce a novel observational method to reveal the relative contribution of the nucleation scavenging mechanism compared with the impaction scavenging mechanism in precipitating clouds, using black carbon (BC) aerosol as an observable particle tracer during the removal process. The BC-tracer method is based on simultaneous ground-based observations of BC in the planetary boundary layer air (prior to removal) and in rainwater (after removal) during a precipitation event. We apply the method to ten distinct precipitation events observed in the urban atmosphere of Tokyo, during a 20-day period in the summer of 2014.

BC-tracer method

First, we succinctly explain the key ideas underlying the BC-tracer method. Here, we focus on an air parcel ascending from the atmospheric boundary layer to the free troposphere via the moist convection process (see Fig. 1). Before the ascending parcel is detrained from the precipitating cloud at higher altitudes, some number fractions of aerosols within the ascending parcel are scavenged by below-cloud impaction, nucleation, or in-cloud impaction mechanisms. A particular volume of rainwater collected on the ground can contain scavenged aerosols from many different ascending parcels. Among these aerosols, we focus on BC, a combustion-generated light-absorbing solid carbonaceous aerosol13. BC is chemically inert and insoluble in water and the mass equivalent diameter (DBC) of each BC particle is invariant in rainwater14 (see the Methods section). Therefore, we can observe the DBC dependence of the wet removal efficiency via simultaneous ground-based measurements of the DBC-resolved number concentration of BC in air16,17 and rainwater14,15. Namely, we can use an observable invariant, DBC, as an identifier for each family of particles containing a particular mass of BC and examine whether some families are more susceptible to wet removal than others. The physical properties of BC-containing aerosols (internal mixtures of BC and other aerosol compounds) that affect removal processes, such as particle size and void-free density, are not very dissimilar to those of BC-free aerosols (i.e., sulfates and organics). In addition, BC-containing aerosols serve as CCN in a similar manner to that of BC-free aerosols18. Thus, observational results obtained for BC-containing aerosols are applicable to other BC-free aerosols because the physical mechanisms of aerosol scavenging are independent of the aerosol’s internal structure.

Next, we explain the observation and data analysis methods underlying the BC-tracer method. A schematic diagram of the data analysis is shown in Supplementary Fig. S1 and all the symbols used in this study are summarized in Supplementary Table S1. We measure both the DBC-resolved number concentration of BC-containing aerosols in surface air (Nair(DBC), units of m–3) prior to removal and the DBC-resolved number concentration of BC particles in rainwater (Nrain(DBC), units of L–1) after removal. RE(DBC) ≡ Nrain(DBC)/Nair(DBC) is interpreted as the DBC-resolved wet removal efficiency of BC-containing aerosols. The DBC dependence of RE(DBC) is a direct reflection of the scavenging mechanism experienced by the removed BC-containing aerosols. On the other hand, the magnitude of RE(DBC) can depend on the macroscopic dynamical processes and the degree of evaporation of falling rain droplets. Because our current focus is the aerosol’s scavenging mechanism, only the DBC dependence of RE(DBC) is of interest herein. Along with the measurements of RE(DBC), the DBC-resolved CCN fraction before removal FCCN(DBC,SS) ≡ NCCN(DBC,SS)/Nair(DBC) is estimated using κ-Köhler theory19 from a measured amount of coating materials on the BC (R; the dry shell-to-core ((BC + non-BC coating)/BC) diameter ratio) and the hygroscopicity parameter (κ) of non-BC aerosol compounds that likely form a coating on the BC20. Here, NCCN(DBC,SS) is the DBC-resolved number concentration of BC-containing CCN under a particular supersaturation (SS). In this study, the FCCN(DBC,SS) is evaluated only for 185 nm ≤ DBC ≤ 370 nm, for which the R data are available. In addition to FCCN(DBC,SS), we theoretically estimate the DBC-resolved number fraction of BC-containing aerosols in an ascending parcel that will be scavenged by the impaction with water droplets in cloud (Fic-imp(DBC)) and below cloud (Fbc-imp(DBC)). The detailed procedures for estimating the FCCN(DBC,SS), Fic-imp(DBC) and Fbc-imp(DBC) are explained in the Methods section. Finally, we compare the calculated DBC dependence of FCCN(DBC,SS), Fic-imp(DBC) and Fbc-imp(DBC) with the observed DBC dependence of RE(DBC). In this study, the possible effects of the re-generation of scavenged aerosols by droplet evaporation and any ice processes are not considered in our data analysis.

Results

We applied the above method to ten distinct precipitation events (Table 1) observed at the Hongo campus of the University of Tokyo during a 20-day period in the summer of 2014. The details of the observation site and sampling methods are explained in the Methods section and in Supplementary Fig. S2. Approximately 30% of the precipitation during the observation period occurred when a cold front associated with a cyclone passed through the observation site. Approximately 30% of the precipitation occurred due to a stationary front sustained over central Japan. Precipitation also occurred due to the influence of a typhoon approaching the main island of Japan. Monitoring of the vertical aerosol distribution using a light detection and ranging (LIDAR) system installed near the observation site revealed that the urban aerosols in the Tokyo metropolitan area were highly concentrated below the cloud base altitude (< ~1 km). In each precipitation event, the horizontal scale of precipitating convective cell typically did not exceed ~10 km, as revealed by the rain radar maps of the Tokyo metropolitan area. For these reasons, we assume that the BC-containing aerosols in the surface air observed within the last 1-h period before the onset of precipitation represent the initial condition of the BC particles observed in rainwater. This 1-h assumption is not critical for the following discussion; the results are similar if we use the data observed within the last 2-h period before the onset of precipitation.

Table 1 Precipitation events observed during the field campaign BC-CARE Tokyo.
Full size table

Figure 2a shows the normalized number size distribution of BC in the air (i.e., the initial distribution) and in rainwater (i.e., the distribution of removed BC) and the resulting removal efficiency RE(DBC) for each of the 10 precipitation events. Each RE(DBC) was scaled to RE(200 nm) = 1 for illustrative convenience. The RE(DBC) exhibits a steep increase with DBC: on average, RE(100 nm) ~0.5 and RE(400 nm) ~2. The observed DBC dependence of RE(DBC) is qualitatively consistent with the observed DBC-resolved CCN activity of BC particles before removal, as shown in Fig. 2b. Figure 2b shows the DBC-dependent critical supersaturation SSc(DBC, Rmed) of BC-containing aerosols for 185 nm ≤ DBC ≤ 370 nm for the same events shown in panel (a); Rmed denotes the median R value observed for BC-containing aerosols with a particular DBC. The steep decrease in SSc(DBC, Rmed) with DBC (i.e., higher cloud-nucleating ability for particles with larger DBC) results from the relatively small amount of coating material on BC (Rmed ~ 1.1) in the size range of 185 nm ≤ DBC ≤ 370 nm. For this size range, the observed Rmed values were not strongly dependent on DBC, and, thus, the total diameter of BC-containing particles (BC + non-BC coating) and the coating mass increased with the increase in DBC. Therefore, particles with larger DBC nucleate cloud droplets more easily and this tendency is consistent with the DBC dependence of RE(DBC) shown in Fig. 2a. The SSc(200 nm, Rmed) value for each precipitation event is listed in Table 1 to show the event-dependent variation of the typical SSc value.

Figure 2
figure 2

DBC-resolved microphysical properties of BC-containing aerosols for all precipitation events.

(a) The number size distribution of BC in air (i.e., the initial distribution) and in rainwater (i.e., the removed distribution) and the resulting removal efficiency (RE) for each of 10 precipitation events (thin lines). The size distribution was normalized by the total number concentration. Bold lines show the average result for the 10 events. (b) Critical supersaturation (SSc) of BC-containing aerosols in the air as a function of DBC and the median shell-to-core diameter ratio (Rmed) for each of the 10 precipitation events (thin lines). The bold line shows the average result for the 10 events. The bars indicate a range of 1 standard deviation.

Full size image

In each precipitation event, we examined whether the observed DBC dependence of the removal efficiency RE(DBC) within the 185 nm ≤ DBC ≤ 370 nm range can be quantitatively explained solely by the CCN fraction FCCN(DBC,SSest), where SSest denotes the SS value that provides the best fit between RE(DBC) and FCCN(DBC). Figure 3 shows the FCCN(DBC,SSest) for precipitation event No. 4. Corresponding results for other 6 events are shown in Supplementary Fig. S3. As in this event, in each of other precipitation events, we found good agreement between the DBC dependence of RE(DBC) and FCCN(DBC,SSest) for a particular SSest value between 0.08% and 0.45% (Table 1). These SSest values seem plausible from the theory: the maximum supersaturation predicted by an adiabatic cloud parcel model does not exceed ~1.0% under realistic meteorological conditions (i.e., upward velocity and aerosol concentration)21. The average SSest for each precipitation type of cold front, stationary front and typhoon were 0.22%, 0.12% and 0.30%, respectively. It is important to note that the SSest is the effective supersaturation: a weighted average of the maximum supersaturations experienced by BC-containing aerosols in many different ascending air parcels, all of which contribute to the concentration of BC particles in the observed rainwater. The theoretically-estimated upper limits of the Fic-imp(DBC) and Fbc-imp(DBC) values for an ascending air parcel within the range 185 nm ≤ DBC ≤ 370 nm are also shown in Fig. 3. Our theoretical estimates reveal that under realistic assumptions, Fic-imp(DBC) always shows a monotonic decrease with DBC and Fbc-imp(DBC) shows no appreciable DBC dependence (see the Methods section). Therefore, neither in-cloud nor below-cloud impaction scavenging mechanisms can account for the steep increases of RE(DBC) observed in this particle size range. Furthermore, the maximum values of Fic-imp(DBC) and Fbc-imp(DBC) do not exceed 0.11 and 0.06, respectively, both of which are much smaller than FCCN(DBC) for this size range.

Figure 3
figure 3

DBC-resolved wet removal efficiency (RE) and the scavenged number fractions (FCCN, Fic-imp and Fbc-imp) of BC-containing aerosols for a selected precipitation event.

The results for the event No. 4 are shown. The solid and dashed black lines show the upper limits of the theoretically-estimated number fractions scavenged by the in-cloud impaction Fic-imp(DBC) and the below-cloud impaction Fbc-imp(DBC), respectively (see the Methods section for details). The red line shows the CCN number fraction FCCN(DBC,SSest), where SSest denotes the SS value that provides the best agreement between FCCN(DBC,SS) and RE(DBC). The SSest value was determined to be 0.25% in this event (No. 4). The blue lines show RE(DBC) scaled to RE(200 nm) = FCCN(200 nm,SSest). The shading shows the range of FCCN(DBC,SS) for SSmin < SS < SSmax. Here, the assumed values of SSmin and SSmax are 0.18% and 0.38%, respectively.

Full size image

From the good agreement of the observed DBC dependence between RE(DBC) and FCCN(DBC, SSest), along with our theoretical estimates that both Fic-imp(DBC) and Fbc-imp(DBC) are much less than FCCN(DBC, SSest), we conclude that nucleation scavenging is the dominant process via which the BC-containing particles are incorporated into water droplets and irreversibly removed, at least for 185 nm ≤ DBC ≤ 370 nm. The chemical composition of an aerosol affects the SSc value but does not affect the physics of the scavenging processes. Therefore, this conclusion can be generalized to other BC-free aerosols with a similar particle size.

Discussion

Our observations using the water-insoluble BC as a particle tracer present conclusive data to prove the hypothesis suggested by previous theoretical7,8 and observational studies11,12: nucleation scavenging is predominant over other scavenging mechanisms and therefore controls the wet removal efficiency of accumulation-mode aerosols. This work firmly indicates the importance of refining the parameters that affect the nucleation scavenging process (i.e., the size, composition and mixing state of aerosols in addition to the maximum supersaturation experienced by the ascending air parcels) for future development of regional/global aerosol models. Such sophistication in models will be required for accurate predictions of the abundance and resulting climate forcings of anthropogenic and natural accumulation mode aerosols. Furthermore, our BC-tracer method provides the supersaturation (SSest) that effectively controls the typical number fraction of aerosols removed from ascending air parcels. Because the accuracy of the estimate of SSest by the present method is higher for steeper slopes in FCCN-DBC correlations, observations in near-source regions where BC particles are relatively fresh and less-coated, such as Tokyo, are more suitable in applying this method. While a supersaturation level was previously estimated from in-situ measurements of aerosol and cloud microphysical properties for a non-precipitating single cloud22,23,24,25, our new observational quantity SSest is obtained for precipitating cloud clusters, which is the average over space and time. This observational quantity can be compared with the theoretical maximum supersaturation for precipitating cloud clusters, which can be computed using a cloud resolving model26. In the future, such observational constraints on the effective supersaturation in precipitating clouds will be invaluable for testing the numerical simulations of aerosol wet removal and vertical transport.

Methods

Experimental setup

Our observation site was the Hongo campus of the University of Tokyo (35.71°N, 139.76°E), located within the Tokyo metropolitan area. A major source of BC in Tokyo is reported to be diesel emissions27. During the field campaign BC-CARE Tokyo, air samples were aspirated from the sixth floor and rooftop of Science Building 1 (approximately 20 and 40 m above ground level, respectively). Supplementary Figure S2 shows a schematic of the experimental setup. The BC mass and mixing state of individual BC-containing aerosols were measured using a single particle soot photometer (SP2, Droplet Measurement Technology, Boulder, CO, USA) with a laser-induced incandescence technique16,17,20,28. The SP2 calibration for BC mass measurement was performed using fullerene soot particles (Stock #40971, Lot #FS12S011, Alfa Aesar, Lancashire, UK) mass-classified with an aerosol particle mass analyser (APM, Kanomax, Osaka, Japan). The mass equivalent diameter of the BC (DBC) was derived from the BC mass assuming a void-free BC density of 1.8 g cm–3. We deployed three different SP2s modified for specific purposes—standard SP2, humidified SP2 and wide-range SP2—as detailed below.

The standard SP2 is able to measure the amount of coating on a BC using a position-sensitive scattering detector in addition to the BC mass. The capability of this SP2 is similar to that of the commercial DMT-SP2 and was fully described in our previous publications16,17. The detectable BC size range is 70 nm < DBC < 850 nm. The amount of coating materials on the BC was estimated by inverting the measured scattering cross section of the BC-containing particles using the light-scattering theory for concentrically stratified spheres. In this inversion, the refractive indices for BC and non-BC coatings were assumed to be 2.26 + 1.26i and 1.52 + 0i, respectively. This coating analysis was performed for a BC size range of 185 nm < DBC < 370 nm.

The humidified SP2 (h-SP2) is able to measure the scattering cross section of the individual BC-containing and BC-free particles at a controlled relative humidity (RH)20,28. The hygroscopicity parameter κ of non-BC materials for individual BC-containing and BC-free aerosols were estimated based on h-SP2 measurements for mass-sorted aerosol flows. In this system, dried ambient aerosols (RH < 10%) were first mass selected by the APM and then introduced into the h-SP2 at a controlled RH of 85%. The hygroscopicity data obtained by this system during the field campaign were included and fully discussed in our recent publication20.

The wide-range SP2 (w-SP2) is able to measure the BC mass in an extended size range of 70 nm < DBC < 2000 nm by utilizing a low-sensitivity incandescence detector instead of a position-sensitive scattering detector14. In our observations, the w-SP2 was used to measure the DBC-resolved number concentrations of BC-containing particles in air and rainwater.

Next, we describe our rain sampling system installed on the 12th floor (Supplementary Fig. S2). The size-resolved number concentration of the rain droplets was measured with a laser precipitation monitor (Adolf Thies GmbH & Co. KG, Göttingen, Germany) on the rooftop. During non-precipitation periods (rain rate < ~1 mm h–1), the w-SP2 directly sampled the ambient aerosols downstream of a Nafion dryer. During precipitation periods, the rainwater droplets collected by a 30-cm-diameter plastic funnel installed on the rooftop were continually transferred to a small buffer tank (glass bottle) with a volume of ~5 mL. The rainwater in the buffer tank was transferred to a pneumatic nebulizer (Marin-5, Cetac Technologies Inc., Omaha, NE, USA) driven by a continuously operated peristaltic pump. When the liquid level sensor attached to the buffer tank sensed a sufficient amount of water, a compressed purified airflow of 16.7 cm3 s–1 into the nebulizer drove continuous aerosolization of the rainwater. The BC particles aerosolized from rainwater were then measured using the w-SP2. Our recent work14 showed that this Marin-5 + w-SP2 system could measure the size-resolved number concentration of BC particles suspended in rainwater, in the size range of 70 nm < DBC < 2000 nm, without size distribution artefacts. The work also showed that the coagulation of BC particles suspended in collected rainwater is negligible during the sample storage for at least approximately one month. Because the water/BC volume ratio in the liquid water should not be very dissimilar between falling rain droplets and collected rainwater, the coagulation of BC particles within a rain droplet should also be negligible, considering the much shorter time scale of cloud-precipitation processes. In addition, the laboratory experiment in this work showed that the incorporation of BC particles into liquid water and the subsequent re-aerosolization of these particles do not change the original size distributions of the BC particles. Based on these experimental facts, we assume that neither cloud-precipitation nor the measurement process changes the size of individual BC particles. Our experiment demonstrated that the loss of BC particles in collected rainwater through the funnel and tubing was no more than 7%, without significant changes in the BC size distribution. The funnel on the rooftop was cleaned with purified water once per day. The number of dry-deposited BC particles on the funnel was estimated to be negligible and did not affect the BC concentration in the rainwater14.

Estimates of FCCN

The technical assumptions made to estimate the CCN number fraction FCCN(DBC, SS) in our method were as follows. The distribution of coating amount R for the specific DBC was derived by inverting the scattering signals measured by the standard SP2 and then converted to the distribution of critical supersaturation SSc according to κ-Köhler theory. In this study, we assumed that the aerosol compositions of the coating materials on the BC were identical to those of BC-free aerosols. This assumption is supported by the discussion in our recent publication20. Thus, the median κ value observed for BC-free aerosols with a mass of 7.4 fg was applied in the κ-Köhler calculations of SSc(DBC, R) for BC-containing aerosols with 185 nm < DBC < 370 nm. Because of the predominance of freshly emitted BC particles at the observation site20, substantial number fractions of BC-containing particles were identified to consist of nearly bare BC particles (R < 1.05). Because of the nonspherical shape of BC-containing particles, our inversion of coating amount from scattering signal should suffer from a substantial uncertainty20. To mitigate the unknown artefacts of the R measurement for thinly coated BC particles, we applied the following assumptions for the number concentration of BC-containing aerosols with R < 1.05. We assumed that the R value could not be less than 1.001 and the number concentration of BC-containing aerosols linearly decreased with R in the domain of 1.001 < R < 1.05. For BC-containing aerosols with R > 1.05, we found that the R distribution was approximated well by a power function. These empirically determined R distribution functions were applied to the calculations of SSc(DBC,R) and thus FCCN(DBC,SS) for each precipitation event.

According to κ-Köhler theory, the SSc value of a thinly coated BC particle is highly sensitive to the amount of coating. For instance, the values of SSc(200 nm, 1.001) and SSc (200 nm, 1.01) under a coating κ of 0.19 are 0.82% and 0.45%, respectively. We investigated the sensitivity of SSest to the minimum R. For every precipitation event, the change in SSest between the two assumptions of minimum R values of 1.001 and 1.01 was limited to within 0.1%. In this study, SSest was estimated assuming a minimum R of 1.001 exclusively.

Estimates of Fic-imp and Fbc-imp

Supplementary Figure S4 shows a schematic of the in-cloud and below-cloud impaction scavenging processes of BC-containing aerosols considered for our theoretical calculations of the Fic-imp and Fbc-imp. The number fractions of BC-containing aerosols in an ascending air parcel scavenged by impactions with water droplets in cloud Fic-imp(DBC) and below cloud Fbc-imp(DBC) were theoretically estimated based on the following formulation of the scavenging rate3:

where Nair(DBC,t) is the DBC-resolved number concentration of interstitial BC-containing aerosols in an air parcel at time t, K(DBC,D) is the collection kernel for the various attachment processes between a BC-containing particle with a diameter of DBC and a water droplet (cloud or rain droplet) with a diameter of D and N(D,t) is the D-resolved number concentration of water droplets in an air parcel at time t. The number fraction of scavenged BC-containing aerosols via in-cloud (below-cloud) impaction at a specific t, Fic(bc)-imp(DBC), is then defined as

If N(D,t) does not vary with time, Fic(bc)-imp(DBC) is determined by

Using Eq. (3), we calculate the upper limits of the Fic-imp(DBC) and Fbc-imp(DBC) under realistic assumptions to estimate the maximum contribution of these impaction scavenging mechanisms to wet removal of BC-containing aerosols in the size range of 185 nm ≤ DBC ≤ 370 nm.

In the calculation of Fic-imp(DBC), we have taken into account the stochastic collision-coalescence between interstitial BC-containing aerosols and cloud droplets in the ascending air parcel, as well as the collision-coalescence between interstitial BC-containing aerosols and falling rain droplets passing through the air parcel (see Supplementary Fig. S4). The Brownian diffusion, inertial impaction and interception are considered in the collection kernel K(DBC,D)3,29. To estimate the upper limit of impaction scavenging, we assume that the N(D,t) for cloud and rain droplets are independent of time (altitude) in our calculations. The N(D) for cloud droplets was assumed to be as follows. The total number concentration of droplets (N) was varied from 150 to 1,000 cm–3, the size distribution is a lognormal function with a geometric standard deviation of 1.42 and the count median diameter (CMD) was varied from 10 to 20 μm. The N(D) for falling rain droplets was assumed to be the average size-resolved number concentration of rain droplets observed on the ground. The temperature of the air and water droplets was assumed to be 275 K and the relative humidity was assumed to be 100%. In our computations, the residence time of the air parcel in the cloud was assumed to be 60 min, which is typical or somewhat longer than the actual residence time. Supplementary Figure S5 shows the calculated Fic-imp(DBC) for 185 nm ≤ DBC ≤ 370 nm for precipitation event No. 4, which is the same event shown in Fig. 3. The Fic-imp(DBC) increases with N and CMD. For any N and CMD values, the Fic-imp(DBC) shows a monotonic decrease with DBC, as opposed to the observed DBC dependence of RE(DBC) (Fig. 3). Even under the extreme assumption resulting in the plausible upper limit of the Fic-imp(DBC) (i.e., N = 1,000 cm–3 and CMD = 20 μm), the Fic-imp(DBC) does not exceed 0.11 in the range of 185 nm ≤ DBC ≤ 370 nm.

We estimate the Fbc-imp(DBC) by assuming that the falling rain droplets were constantly passing through the air parcel for 60 min. A plausible upper limit of Fbc-imp(DBC) was evaluated by doubling the assumed number concentration of rain droplets. In our Fbc-imp(DBC) calculation, Brownian diffusion, inertial impaction, interception and thermophoresis were considered for the collection kernel K(DBC,D)3,29. The temperatures of the air and droplets were assumed to be 299 K and 297.5 K, respectively. The relative humidity was set to 90%. The calculated Fbc-imp(DBC) for precipitation event No. 4 is shown in Supplementary Fig. S5. The Fbc-imp(DBC) does not exceed 0.06 and there is little DBC dependence in the 185 nm ≤ DBC ≤ 370 nm range, as opposed to the observed systematic increase of RE(DBC) with DBC (Fig. 3).

From these results, we conclude that neither below-cloud nor in-cloud impaction scavenging mechanisms can explain the observed size-dependence of RE(DBC). The same conclusion was obtained for all the precipitation events in this study.

Additional Information

How to cite this article: Ohata, S. et al. A key process controlling the wet removal of aerosols: new observational evidence. Sci. Rep. 6, 34113; doi: 10.1038/srep34113 (2016).