Number-size distributions of free tropospheric aerosol particles of 9–300-nm diameter were observed at Mt. Norikura (36.1°N, 137.5°E, 2770 m a.s.l.) in central Japan in August and September 2002. We observed two distinct air masses transported from over the Pacific Ocean and continental Asia. Size distributions of aerosol particles were analyzed in terms of their relation to the air mass transportation pathway, cumulative precipitation amount in the air mass for the last 24 h, and local precipitation rate. Particle concentrations of the accumulation mode range (100–300 nm) decreased by an order of magnitude when cumulative precipitation amount increased from 0 to 20 mm. The cumulative precipitation amount is suggested as an important factor to modify the size distribution by decreasing particle concentration. Considering data for air masses without experience of precipitation scavenging before arrival, concentrations of accumulation mode particles were almost identical (approximately 400 cm−3) in air masses transported from over the Pacific Ocean and continental Asia. Concentration of Aitken mode particles was high (890 cm−3 as a median value) in the air mass subsiding from over the subtropical Pacific Ocean (> 4 km a.s.l.). Based on the relationship with the controlling factors described above, this study provides number-size distribution parameters of free tropospheric aerosol particles in the west Pacific area.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Atmospheric aerosol particles affect the Earth’s radiation balance directly by absorbing and scattering incoming solar radiation and indirectly by acting as cloud condensation nuclei, thereby, modifying cloud properties [e.g., Twomey, 1977; Charlson and Heintzenberg, 1995; Lohmann and Feichter, 2005; Yu et al., 2006]. Along with chemical composition and mixing states, the number-size distribution of aerosol particles is a crucial parameter that determines optical properties of aerosols and the number of cloud condensation nuclei. The number-size distribution of atmospheric aerosol particles is characterized by a continuous and multimodal distribution, which can be approximated by the sum of lognormal functions [Whitby, 1978]. The shape (such as modal appearance and smoothness) of the size distribution of aerosol particles is modified by microphysical processes of particle formation, growth, and removal from the atmosphere and by mixing with different air masses.
 Typical modes appearing in number-size distribution of submicrometer aerosol particles are the so-called accumulation, Aitken, and nucleation modes [Raes et al., 2000; Kulmala et al., 2004]. Aerosol particles of the accumulation mode with maximum concentration at around 0.1 μm particle diameter (Dp) are most effective for interacting with incoming visible solar radiation [Seinfeld and Pandis, 1998, chapter 22]. Water-soluble aerosol particles in the accumulation mode can be activated as cloud condensation nuclei under realistic water vapor supersaturation in the atmosphere [Pruppacher and Klett, 1997, chapter 6]. Aerosol particles with Dp < 0.1 μm are less effective for climatic forcing; however, knowledge of the size distribution for sizes of Dp < 0.1 μm is important to elucidate mechanisms to sustain accumulation mode particles.
 Submicrometer atmospheric particles are produced mainly by gas-to-particle conversion processes in the atmosphere, whereas supermicrometer particles are formed by disintegration processes of bulk material (for example, sea salt and dust [Whitby, 1978]). Gas-to-particle conversion processes include homogeneous nucleation to form nuclei of a few nanometers. When a strong particle formation event by homogeneous nucleation occurs during atmospheric transport immediately before arrival at a site, a mode for Dp < 10 nm, so-called nucleation mode, is often observed in the number-size distribution. The particle concentration of the nucleation mode is often increased rapidly to the order of 104 cm−3 [e.g., Mäkelä et al., 1997; Weber et al., 1997; Birmili and Wiedensohler, 2000]. The newly formed aerosol particles grow into Aitken mode particles with Dp of several 10 to 100 nm and subsequently into accumulation mode particles with Dp > 100 nm by coagulation and condensation of nonvolatile gases such as gas-phase sulfuric acid. To form accumulation mode particles, cloud processes might also be important, especially in a clean atmosphere such as a remote marine boundary layer [Hoppel et al., 1986, 1994a, 1994b].
 Accumulation mode particles are eventually removed from the atmosphere, mainly by precipitation scavenging, because gravitational settling and coagulational scavenging processes are too slow to remove accumulation mode particles [Seinfeld and Pandis, 1998, chapter 19]. Precipitation scavenging includes in-cloud and below-cloud processes. For aerosol particles of 0.01 < Dp < 1 μm, below-cloud scavenging is not effective [e.g., Greenfield, 1957; Andronache, 2003]. In-cloud scavenging, including activation processes as cloud condensation nuclei (nucleation scavenging), may be a dominant removal process for accumulation mode particles.
 Associated with the seasonal change of synoptic wind patterns around Japan, the free troposphere over central Japan is covered by the following two distinct air masses: subtropical marine air mass transported from over the Pacific Ocean and continental air mass transported from over continental Asia. In summer, the dominant air mass is the marine air mass over Japan, but it yields to the continental air mass during other seasons. Associated with the seasonal changes of air mass transportation pathway, clear seasonal variation has been observed in aerosol optical properties and particle shape [Sakai et al., 2000], ionic constituents of aerosol particles [Osada et al., 2002], and aerosol particle concentrations at Dp > 0.3 μm [Osada et al., 2003] in the free troposphere over central Japan.
 For aerosol size distribution of particles of Dp < 0.3 μm, however, long-term continuous free tropospheric observations have not been made for the western Pacific area. A few campaign studies have been reported for aerosol number-size distribution for Dp < 0.3 μm in the free troposphere near Japan [Zaizen et al., 2004; McNaughton et al., 2004], but these have been short-term case studies that fail to provide reliable parameters of size distribution that are representative for this region. As a mountain-based long-term observation of free tropospheric aerosols, Raes et al.  observed size distributions of free tropospheric aerosol particles at the Canary Islands in the northeast Atlantic and showed average size distributions in a continental air mass transported from northern Africa and a marine air mass subsiding from over the northern Atlantic Ocean. Weingartner et al.  showed average size distributions of aerosol particles at Jungfraujoch in the Swiss Alps and asserted a difference of the average size distribution among seasons. However, modification of the average size distribution by precipitation scavenging during air transportation was not considered in those studies.
 We observed number-size distributions for 9–300-nm diameter in free tropospheric particles over Mt. Norikura (2770 m a.s.l.) in central Japan during August and September 2002. That observation period, which includes a transition period between summer and autumn, was selected to observe the two contrasting major air masses for Japan transported from the Pacific Ocean and continental Asia. The objective of this paper is to illustrate the important effects of precipitation scavenging processes on average size distribution of free tropospheric aerosol particles and to provide size distribution parameters that are representative for different air mass types as defined by air mass transportation pathway and the effects of precipitation scavenging.
2.1. Observation Site and Periods
 Aerosol observations were performed at the Norikura Observatory (36.1°N, 137.5°E, 2770 m a.s.l.) of the Institute for Cosmic Ray Research (ICRR), University of Tokyo near the top of Mt. Norikura in central Japan (Figure 1a). The observatory is above the forest limit in a preserved area, a national park, and is isolated from industrial urban areas. The nearest cities are Takayama and Matsumoto, which are cities of 100,000–200,000 people at a distance of 30–40 km from the observatory. The observatory is situated between mountains to the south and north (Figure 1b). Therefore the local wind directions are limited mostly to east and west. Possible local air pollution sources include exhaust of diesel power generators located east of the observation laboratory. No discernible local air pollution source is located west of the laboratory. Therefore the influence of the local air contaminants was eliminated from aerosol data according to local wind direction. Observations were performed from 29 July to 23 August 2002 and from 10 to 27 September 2002.
 Dry particle number-size distributions for 9–300-nm diameter particles were measured using a Scanning Mobility Particle Sizer (TSI Inc.) including a DMA (Electrostatic Classifier 3071A, TSI Inc.) and a condensation particle counter (UCPC3025A, TSI Inc.). The sample air was drawn into the instrument from an inlet at 2 m above the ground through a 2-m electrically conductive resinous tube. Particle losses onto the sampling tube were corrected using laminar flow diffusion theory [Hinds, 1999]. Relative humidity (RH) of the sheath airflow of the DMA was kept under 20% to measure dry particle size distributions. Commercial software provided by the manufacturer was used for data acquisition and conversion from the particle mobility distribution into the particle size distribution. Particle size distribution data were obtained every 3 min (150 s for data scan followed by a 30-s down time). The particle number concentrations presented in this paper are corrected to 1013 hPa condition from 725 hPa condition, where the observatory is located.
 Local meteorological parameters (wind direction, wind speed, ambient air temperature, and precipitation rate) were measured using a Mini-Met weather station (Grant Instruments). Dew point temperatures were measured with an impedance hygrometer (Cermet II, Michell Instruments Ltd.). Ozone (O3) concentrations were measured using a Dasibi-type UV absorption O3 monitor (Model 1150, Tokyo Dylec Corp.). Sulfur dioxide (SO2) concentrations were measured using a pulsed UV fluorescence SO2 analyzer (Model 43C-Trace Level, Thermo Electron Corp.).
2.3. Data Screening
 Diurnal evolution of upslope valley wind and downslope mountain wind has been recognized at high-elevation sites [Mendonca, 1969; Lugauer et al., 1998; Nyeki et al., 1998]. Upslope valley winds result from radiative heating of the mountain surface during daytime; downslope mountain winds are caused by radiative cooling of the mountain surface at night. Boundary layer air transported by upslope valley winds prevents observation of free tropospheric air at high-elevation sites during daytime. As a tracer of lower boundary layer air, the SO2 concentration might be a suitable parameter. Reportedly, SO2 concentration levels decrease with increasing altitude and are different between boundary layer and free troposphere [Warneck, 1999]. Figure 2 shows average diurnal variation of normalized SO2 concentrations of clear summer days (6 days during the period from 29 July to 8 August 2002). The SO2 concentrations were normalized; daily average values were set to unity. The SO2 concentration showed clear diurnal variation, an increase in daytime and a decrease during nighttime. Lowest SO2 concentrations were observed during 23:00–7:00 local time (LT; UT + 9 h), associated with downslope mountain winds during nighttime. In this study, aerosol data during 0:00–6:00 LT are discussed as containing free tropospheric aerosols.
 The 3-min size distribution data were screened in reference to local wind direction to remove local air contamination from the data. About 80% of the 3-min data were retained after the contamination screening. The hourly size distribution was calculated if more than five 3-min raw data were retained after the contamination screening for that hour. In all, 228 hourly measurements of free tropospheric air condition were retained.
2.4. Lognormal Fitting
 Hourly number-size distributions of aerosol particles were approximated by a sum of up to three lognormal functions individually. The sum of the lognormal function is given as:
 Therein, Dp is the particle diameter, n is the number of modes, Dg,p,i is the modal geometric mean diameter, σg,i is the modal geometric standard deviation, and Ni is the particle number concentration in mode i. Details of the lognormal distribution are described by Seinfeld and Pandis [1998, chapter 7]. These modal parameters were obtained using a nonlinear peak-fitting tool that was included with commercial software (Origin, OriginLab Corp.). We applied the restrictions as (1) Ni > 1% of total particle concentration and Ni > 1 cm−3, (2) 1.3 Dp,g,1 < Dp,g,2, (3) log σg,i < 0.5 to avoid unrealistic mode fitting. Generally, modes appearing in sub-micrometer sizes are called nucleation, Aitken, and accumulation modes according to their geometric mean diameter [Raes et al., 2000; Kulmala et al., 2004]. In this study, we follow the names of modes in submicrometer sizes as follows: nucleation mode, Dp,g < 30 nm; Aitken mode, 30 nm < Dp,g < 85 nm; and accumulation mode, Dg.p > 85 nm. The borders of size ranges among nucleation, Aitken, and accumulation modes were set to reduce the number of data of multimodes in the size range to a minimum. In addition, Ni of each mode is called nucleation, Aitken, and accumulation mode concentration in this study.
2.5. Air Mass Transportation Pathway and Meteorological Conditions
 Air mass backward trajectories for 5 days were analyzed to investigate their relation to observed size distributions of aerosol particles. Backward trajectories were computed using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model developed by National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory (ARL) [Draxler and Rolph, 2003; Rolph, 2003]. Calculations of backward trajectories were begun from 3000 m a.s.l., 230 m above the site at 1:00, 3:00, and 5:00 LT. The National Weather Service’s National Centers for Environmental Prediction (NCEP) FNL archive was used for meteorological input data. Vertical wind velocities were used to calculate vertical motions of air parcels.
 According to the backward trajectories, transportation pathways of observed air masses were classified into the following two groups: type A, transported from the Pacific Ocean; and type B, transported from other areas. Air masses were classified as type A if the endpoint of the 5-day backward trajectory was located at above the Pacific Ocean without passing over continental Asia. Air masses that did not meet conditions of type A were classified as type B. Figures 3a and 3b show backward trajectories classified, respectively, into types A and B. Figure 3b shows that almost all air masses of type B were transported from continental Asia and that several air masses stagnated near Japan.
Figure 4 shows the respective variations of ambient air and dew point temperatures (Figure 4a), O3 concentrations (Figure 4b), and local precipitation rate (Figure 4c). The parameters in Figure 4 are average values in free tropospheric conditions (0:00–6:00 LT) except for local precipitation rates (mm·h−1) of 0:00–6:00 LT. The air mass transportation patterns, types A and B, are indicated at the top of Figure 4a by black and gray bars. For the period when aerosol data were not obtained, classification of the transportation pattern is not shown at the top of Figure 4a.
 Air and dew point temperatures were constantly high in type A, but variable in type B (Figure 4a). Low O3 concentrations (10–30 ppb) were observed in type A (Figure 4b). In type B, consisting mostly of the air masses from continental Asia, O3 concentration levels were higher (20–60 ppb) than those in type A. The difference in the O3 concentrations between the air mass types might result from origin and air mass transportation pathway. Low O3 concentrations are thought to be characteristic of a clean marine troposphere. At the summit of Mt. Fuji (35.4°N, 138.7°E, 3776 m a.s.l.) in Japan [Tsutsumi et al., 1998] and over the western Pacific Ocean [Tsutsumi et al., 1996], lower O3 concentrations (approximately 20 ppb) have been reported for marine air masses transported from over the Pacific Ocean, although high O3 concentrations (40–50 ppb) have been reported for air masses transported from continental Asia. Therefore low O3 concentrations in type A (10–30 ppb) and high O3 concentrations in type B (20–60 ppb) are consistent with the low concentrations reported in clean marine air masses in the western Pacific area and high concentrations for air mass from continental Asia.
Figure 5 shows the 700 hPa geopotential height field on 6 August 2002 as a representative synoptic-scale meteorological condition of type A, which is frequently observed in summer months in Japan. Figure 5 shows that when an air mass was transported from the Pacific Ocean to Mt. Norikura, central Japan was covered by a Pacific high-pressure system centered over the northwestern Pacific Ocean. In contrast, various patterns of synoptic-scale meteorological conditions existed for the type B air mass. As examples of meteorological patterns of type B, Figure 6 shows 700 hPa geopotential height fields on 1 August (Figure 6a), 24 September (Figure 6b), and 17 September 2002 (Figure 6c). As described above, air and dew point temperatures were variable in type B (Figure 4a). The variations of air and dew point temperatures are apparently related to the variation of synoptic-scale meteorological condition of type B. For example, high air and dew point temperatures were often observed when a Pacific high-pressure system was located over western and central Japan, as shown in Figure 6a. Figure 6b shows that low air and dew point temperatures were often observed when central Japan was under the influence of midlatitudinal westerly winds. Local precipitation was observed frequently from 11–18 September (Figure 4c). Figure 6c shows that these local precipitation events were mainly brought by synoptic frontal systems, so-called Shurin, located near central Japan.
 In all cases, SO2 concentrations (not shown in Figure 4) were below or close to the detection limit of the SO2 analyzer (approximately 0.1 ppbv) in the free tropospheric air condition. Several active volcanoes emit large amounts of volcanic gases in Japan (for example, Miyakejima Volcano [Kajino et al., 2004] and Mt. Asama [Andres and Kasgnoc, 1998]). However, the low SO2 concentrations suggest that the observed aerosols were not influenced by the active volcanoes nearby.
3. Results and Discussion
3.1. Temporal Variation of Number-Size Distribution of Aerosol Particles
Figure 7 shows temporal variations of number-size distribution (dN/d log Dp, cm−3) of aerosol particles observed in free tropospheric conditions (0:00–6:00 LT; Figure 7a), modal geometric mean diameters (Dp,g,i) obtained by lognormal fitting (Figure 7b), particle number concentrations in each mode and total particle concentrations of 9 nm < Dp < 300 nm (N9–300) (Figure 7c), and the cumulative precipitation amount experienced by air mass for the preceding 24 h (Figure 7d). The air mass transportation patterns, types A and B, are indicated at the top of Figure 7a. Aerosol data were not obtained on the days indicated as “no data” in Figures 7a and 7c either because of failure in data acquisition or local air contamination.
 Parameters of precipitation scavenging on aerosol particle concentration include the magnitude and duration of precipitation experienced by the observed air mass before arrival. Thus cumulative precipitation amounts in the air mass might be a useful parameter to evaluate effects of precipitation scavenging on the concentration and size distribution of aerosol particles. The cumulative precipitation amount shown in Figure 7d was estimated for the 24 h before arrival based on hourly precipitation rates along backward trajectories obtained using the HYSPLIT model. Although the meteorological parameters obtained using the HYSPLIT model are not observed data but objective analytical data, these values can be useful for diagnosing meteorological conditions along the trajectory.
 Throughout the observations, accumulation and Aitken modes were dominant over nucleation modes in both frequency of occurrence (FO) and modal concentration (Figure 7c). About 50% of the observed size distributions were bimodal distributions with accumulation and Aitken modes, and about 40% were monomodal distributions of either accumulation or Aitken modes.
 During 6–20 August (predominantly type A), large variations of N9–300 and size distribution were observed (Figure 7c). Note that low N9–300 were observed in the air mass with higher cumulative precipitation amount (for example, 9–10 and 14–15 August). In contrast, lower cumulative precipitation amounts might engender higher N9–300 concentrations with large variability. In addition, accumulation mode concentrations were extremely low during a severe local precipitation event (12–18 September in Figure 4c). Thus precipitation scavenging is an important controlling factor of aerosol concentration, especially for accumulation mode particles. In other words, precipitation scavenging might modify the aerosol particles’ size distribution.
3.2. Relationship Between Precipitation Amount and Particle Number Concentrations
Figure 8 depicts the relationship between particle concentration and cumulative precipitation amounts for the last 24 h in air mass as follows: total particle concentration in sizes of 9 nm < Dp < 100 nm (N9–100) (Figure 8a) and total particle concentration in sizes of 100 nm < Dp < 300 nm (N100–300) (Figure 8b) to the cumulative precipitation amount. Negative correlations of the logarithm of N9–100 and N100–300 with cumulative precipitation amounts are shown, respectively, in Figures 8a and 8b. Least squares fitting lines using the exponential function are also shown in Figure 8. The fitting lines indicate that, when the cumulative precipitation amount was increased from 0 to 20 mm, N9–100 was decreased by about half (Figure 8a), and N100–300 was decreased by an order of magnitude (Figure 8b). Although the coefficients of determination (R2) of the fitting line are especially low for the fitting line in Figure 8a (0.08), the correlations shown respectively in Figure 8a and 8b are statistically significant at the 0.01 level.
 For comparison of our results with other studies, time constants of precipitation scavenging are estimated from the fitting lines. Assuming that the precipitation rate during transportation is 1 mm·h−1, value of cumulative precipitation amount (mm) is inferred to correspond with the duration of precipitation (h). Consequently, the e-folding time of N9–100 and N100–300 can be estimated from the decay constant (pre–x coefficient) obtained using the exponential fitting lines in Figure 8. The fitting lines in Figure 8 suggest e-folding times at around 20 h for N9–100 and 5 h for N100–300. For below-cloud scavenging process, the e-folding time of particle concentration in sizes of 0.01μm < Dp < 1 μm was estimated at the order of hundreds of hours for a small precipitation rate (1–10 mm·h−1) [Sparmacher et al., 1993; Andronache, 2003]. Consequently, these results indicate that below-cloud scavenging is too slow to explain the observed decrease in particle concentrations with the increasing cumulative precipitation amount. In the case of in-cloud scavenging, the e-folding time of accumulation mode concentration was estimated at the order of hours for small precipitation rate (1–10 nm·h−1) [Scott, 1982; Andronache, 2004]. This value agrees well with the e-folding time estimated for N100–300 in this study. Therefore the decrease in concentration of accumulation mode particles with the increasing cumulative precipitation amount is explainable mostly by in-cloud particle scavenging processes.
Figure 9 shows scatterplots of N9–100 (Figure 9a) and N100–300 (Figure 9b) versus the local precipitation rate. Different from Figure 8 a relation of particle concentration to the local precipitation rate is not apparent in Figure 9. The local precipitation rate in Figure 9 is dominated by small values less than 1 mm·h−1, which possibly causes the lack of correlation between particle concentration and local precipitation rate in Figure 9. However, high N100–300 (> 100 cm−3) was rarely observed under a higher local precipitation rate (> 0.1 mm·h−1; Figure 9b). Fog is usually associated with local precipitation at the site. Therefore nucleation scavenging might engender lower N100–300 rather than the scavenging process by raindrops of local precipitation. On the other hand, high N9–100 (> 100 cm−3) values were observed for higher local precipitation rate (> 0.1 mm·h−1; Figure 9a). Because unrealistically high supersaturation is required for activation of Aitken mode particles, nucleation scavenging under fog condition might not reduce N9–100.
3.3. Relationship Between Particle Number-Size Distribution and Air Mass Transportation With and Without Precipitation
 Because of the effect on precipitation scavenging, the size distribution data set used in this study will be summarized by dividing the degree of precipitation scavenging and transportation pathways. We will attempt to show statistical information of size distributions representative for different air mass transportation pathways with and without the effects of precipitation scavenging.
 The type A air mass transported from over the Pacific Ocean was divided into two subgroups according to its cumulative precipitation amount. The first subgroup (type A-1) was defined as having cumulative precipitation of less than 5 mm; the second subgroup (type A-2) was defined as having cumulative precipitation amounts greater then 5 mm. The criterion of 5 mm of cumulative precipitation was set on the basis of the decay constant of N100–300 estimated in section 3.2. Local precipitation was rarely observed in type A. Therefore local precipitation will not be considered for type A. In contrast, type B, which was transported mainly from continental Asia, was divided into two subgroups according to local precipitation rate. The first subgroup (type B-1) was defined as having a local precipitation rate lower than the detection limit (0.1 mm·h−1); the second subgroup (type B-2) was defined as having a local precipitation rate higher than 0.1 mm·h−1. Cumulative precipitation amounts in type B were rarely greater than 5 mm. Therefore the effect of the cumulative precipitation amount will not be considered for type B. Table 1 shows classifications based on the air mass transportation pathway, cumulative precipitation amount, and the local precipitation rate. The number of hourly data of each subtype is also presented in Table 1.
Table 1. Backward Trajectory, Median Cumulative Precipitation Amount, and Local Precipitation Rate for Different Air Mass Typesa
Cumulative Precipitation Amount, mm
Local Precipitation Rate, mm·h−1
Number of Hourly Data
Cumulative precipitation amounts are defined as cumulative precipitation amount for the preceding 24 h before arrival estimated from precipitation rate (objective analysis data) along backward trajectories. Values in square brackets indicate 25–75% ranges.
Table 2 portrays data for modal parameters obtained using lognormal fitting on individual hourly size distributions and frequency of occurrence (FO) of each mode. Regarding the nucleation mode, because lognormal parameters of nucleation mode did not change with the air mass types and because the nucleation mode FO was much lower (14% of entire data) than that of either the Aitken or accumulation mode, lognormal parameters of the nucleation mode are provided for all data in Table 2.
Table 2. Median Values of Lognormal Parameters Obtained by Lognormal Fitting on Hourly Size Distributionsa
Values in square brackets indicate 25–75% ranges. FO denotes frequency of occurrence of the mode.
Figure 10 shows median particle number-size distributions of air mass types A-1 (Figure 10a), B-1 (Figure 10b), A-2 (Figure 10c), and B-2 (Figure 10d). The number-size distributions in Figure 10 were calculated using the median of particle number density (dN/d logDp) of each size bin with resolution of 16 bins per decade. Note that the median number-size distributions shown in Figure 10 were not reproduced from size distributions of median lognormal parameters in Table 2.
 Types A-1 and B-1 are presumed to be less affected by precipitation and nucleation scavenging. For the accumulation mode, particle concentrations of types A-1 and B-1 were mutually similar (type A-1, 350 cm−3; type B-1, 400 cm−3; see Table 2). However, Aitken mode concentrations were quite different between types A-1 (890 cm−3) and B-1 (230 cm−3). The FO of the Aitken mode was also different for types A-1 (100%) and B-1 (59%). As shown in Figures 10a and 10b, the maximum concentration of the median size distribution was found to be about 50 nm for type A-1 but much larger (>100 nm) for type B-1. The contrast in the particle size distribution between marine and continental air masses differs entirely from what was observed in the boundary layer. Statistical investigations of particle size distributions observed in the boundary layer in Germany [Birmili et al., 2001] and in northern Europe [Tunved et al., 2005] indicated that continental air masses have greater accumulation and Aitken mode concentrations, on average, than marine air masses.
 A high particle-number concentration (N9–300, 1200 cm−3) dominated by Aitken mode particles (890 cm−3) is an interesting characteristic of type A-1. In addition to 5-day backward trajectories, we calculated 10-day backward trajectories for type A-1 air masses. Backward trajectories of type A-1 indicate that most type A-1 air masses subsided from the altitude of 4–5 km for the preceding 5 days (Figure 3a) and originated from altitudes of 5–10 km above the tropical Pacific Ocean during the preceding 10 days.
 In the upper troposphere (8–12 km) over the tropical Pacific Ocean, high concentrations (5000–20,000 cm−3 STP) of newly formed aerosol particles with Dp < 10 nm were observed frequently [Clarke and Kapustin, 2002]. Evolution of the number-size distribution of aerosol particles during subsidence from tropical upper troposphere to the subtropical lower troposphere was investigated using aerosol dynamic models [Clarke, 1993; Raes, 1995; Raes et al., 2000]. According to the model studies during subsidence for 1–2 weeks from the upper troposphere, the aerosol particle concentration might decrease to 1000–2000 cm−3 by coagulation, and its mode diameter of less than 10 nm might grow into Aitken mode range through coagulation and condensation of gas-phase sulfuric acid. The total particle concentration and peak diameters of the dominant mode (40–60 nm) predicted by the model studies agree with type A-1 in this study. The models typically predict a monomodal distribution, but the size distribution of type A-1 was bimodal distribution with Aitken and accumulation modes (Figure 10a). In the upper troposphere, concentrations of accumulation mode particles were reported to be very low (approximately <10 cm−3, STP) [e.g., Clarke et al., 1999; de Reus et al., 2001; Zaizen et al., 2004]. The manner in which accumulation mode particles are added to the Aitken mode particles during subsidence from the tropical upper troposphere remains enigmatic. However, the following processes are considered to contribute to the formation of the bimodal size distribution.
 First, in-cloud sulfate production might add aerosol particles of accumulation mode size through the so-called cloud processes [Hoppel et al., 1986, 1994a, 1994b]. However, cloud formation is not realistic in a subsiding air mass. In fact, according to meteorological data (objective analysis data) along backward trajectories, relative humidities (RHs) of type A-1 are mostly lower than 60% (average RH, 55%) during transportation for the preceding 5 days, suggesting that cloud processes play a minor role in adding the accumulation mode particles. Next, the mixing process with an air mass that is rich in accumulation mode particles might add accumulation mode particles to the subsiding air mass under the monomodal Aitken mode particles. According to meteorological data along backward trajectories of type A-1 masses, water mixing ratios in type A-1 masses show a gradual increase from 3 to 6 g·kg−1 during transportation for the preceding 5 days, suggesting gradual mixing of subsiding air with low-altitude air containing water vapor and accumulation mode particles during transportation. This process might not cause a great decrease of Aitken mode concentration because gradual dilution of Aitken mode particles leads to a slow decrease of Aitken mode concentration through coagulation if Aitken mode particles originated in the upper troposphere were diluted through gradual mixing with low-altitude air during subsidence.
 In type B-1, monomodal size distributions account for 46% of the entire distribution. Most (80%) of the monomodal size distribution is of monomodal accumulation mode with a large average σg (approximately 1.7; Table 2). Monomodal accumulation modes were observed frequently, especially during 29 July–5 August (Figure 7), when warm humid air masses (Figure 4a) were transported from over the coastal area of the Yellow Sea in China (Figure 6a). The coastal area of the Yellow Sea is a highly industrialized, major source of air pollutants in Asia [Streets et al., 2003].
 To our knowledge, in industrialized areas of China, measurement of particle number-size distributions of Dp < 100 nm has not been reported. However, in an extremely polluted Asian city, New Delhi, India, monomodal accumulation modes with high particle concentration (order of 104 cm−3) were observed during nighttime when the influence of primary particle emissions from traffic, biomass and refuse burning, and cooking are low [Mönkkönen et al., 2005]. The particle size distributions in industrialized areas in China might resemble those of New Delhi. The monomodal accumulation mode observed in type B-1 might be explained by air mass transportation from the highly polluted area and dilution with clean air during transportation to Mt. Norikura. On the other hand, because monomodal accumulation modes are observed frequently in humid air masses, the lack of Aitken mode in the size distributions might also be explained by impaction scavenging of Aitken mode particles by cloud droplets in nonprecipitation clouds (see below).
 Regarding type A-2 masses, FO of accumulation mode and median concentration of accumulation mode particles were low (54%, 100 cm−3). As described in section 3.2, it is inferred that in-cloud scavenging mainly accounts for low accumulation mode concentration in type A-2 air masses. The particles activated as cloud condensation nuclei will be removed by precipitation from clouds. That might engender the low FO of accumulation mode and low accumulation mode concentration in type A-2 masses. On the other hand, Aitken mode concentration was also low (100 cm−3) in type A-2, although the FO of Aitken mode was high (100%). Smaller particles, which are inactive as cloud condensation nuclei (Dp approximately < 0.1 μm), might not be removed by nucleation scavenging at the first stage of cloud formation. However, they might be removed via diffusion to the vapor droplets in the clouds. We estimate scavenging speed of cloud interstitial aerosol particles inside the cloud by the Brownian diffusion to cloud droplets [Fuchs, 1964], assuming that the cloud droplets’ diameter is 20 μm, the cloud droplet concentration is 50–500 cm−3 [Pruppacher and Klett, 1997, pp. 15–24], the dry diameters of cloud interstitial particles are 50 nm (corresponding to peak diameter of the Aitken mode of type A-2), and the hygroscopic growth factor of interstitial particle diameter is 2.7 (growth factor of (NH4)2SO4 particle at RH 98% [Tang and Munkelwitz, 1994]). The e-folding time of a cloud interstitial aerosol particle concentration is estimated as about 10–100 h, which is consistent with the decay time estimated for N9–100 in section 3.2 (20 h). The low Aitken mode concentration of the type A-2 air mass might be caused by scavenging through diffusion to cloud droplets.
 In type B-2, the FO of accumulation mode and concentration of accumulation mode particles were low (38%, 100 cm−3). The FO of the Aitken mode was high (approximately 100%) in type B-2, but unlike type A-2, the Aitken mode concentration (340 cm−3) was not so low as in type B-2. As mentioned in section 3.2, the lack of accumulation mode particles in type B-2 masses might result mostly from nucleation scavenging in fog, which is usually associated with local precipitation at the site.
 The results of this study suggest that the precipitation amount experienced by an air mass is an important factor determining the number concentration and shape of the size distribution of submicrometer particles. The importance of precipitation scavenging process as a controlling factor for number-size distribution of aerosol particles was pointed out also in the boundary layer by Tunved et al.  on the basis of statistical analyses using meteorological data along backward trajectories of air masses. Statistical information of number-size distributions of aerosol particles combined with the degree of precipitation scavenging would be useful for model verification of various models that address or incorporate aerosol parameters. However, to our knowledge, no such study has provided statistical information of size distribution parameters with the degree of precipitation scavenging experienced by air mass. A long-term observation of number-size distribution of free tropospheric aerosol particles has been performed at Jungfraujoch in the Swiss Alps (3580 m a.s.l.) [Weingartner et al., 1999]. Without data classification according to cumulative precipitation amounts, as used in this study, meaningful aerosol parameters might not be obtained.
3.4. Occurrence of Nucleation Mode Related With Precipitation Scavenging
 Occurrence of nucleation mode particles was not frequent (14%) at Mt. Norikura under free tropospheric conditions. Geometric mean diameters of the observed nucleation modes were mostly larger than 20 nm, and concentrations of nucleation mode particles were 20–200 cm−3. As one example, a nucleation mode observed on 10 August 2002 is depicted in Figure 11.
 Nucleation modes often appeared under conditions in low Aitken mode and accumulation mode concentrations without local precipitation. Nucleation mode occurred most frequently in type A-2 air masses (30%). The surface of preexisting particles functions as a strong sink of condensable gases that are required to form new particles by homogeneous nucleation (for example, gas-phase H2SO4). Moreover, coagulation scavenging onto preexisting particles is a strong sink of newly formed particles [Kerminen et al., 2001]. Median particle surface concentrations of type A-2 (4 μm2 cm–3) were much lower than those of types A-1 (54 μm2·cm−3) and B-1 (34 μm2·cm−3). Precipitation scavenging might reduce the particle surface concentration. Therefore new particle formation is enhanced in such low-particle conditions. Precipitation scavenging might be an important process that provides favorable conditions for new particle formation by decreasing preexisting particle concentrations.
4. Summary and Conclusions
 Number-size distributions for 9 nm < Dp < 300 nm of free tropospheric aerosol particles were observed at Mt. Norikura (2770 m a.s.l.) in central Japan, in August and September 2002. During the observation period, we observed two air masses transported from over the Pacific Ocean and continental Asia. Observed size distributions were analyzed in terms of their relation with the air mass transportation pathway and meteorological parameters, especially for cumulative precipitation amount in air mass for the preceding 24 h before arrival and local precipitation rate.
 Aerosol particle concentrations in accumulation mode range (N100–300) were decreased by an order of magnitude when the cumulative precipitation amount increased from 0 to 20 mm. The cumulative precipitation amount prior to arrival at the observation site was suggested as an important factor that considerably decreased the number concentration of accumulation mode particles. In the case of local precipitation with fog at the site, low N100–300 was also observed, possibly because of nucleation scavenging by local fog droplets. Considering air masses without precipitation before arrival, accumulation mode concentration was almost the same in air masses transported from over continental Asia (400 cm−3) and the Pacific Ocean (350 cm−3). Aitken mode concentrations were high (890 cm−3) in the air mass that arrived from over the subtropical Pacific Ocean, but lower in the air mass that had been transported from continental Asia (230 cm−3). Backward trajectories indicate that air masses with high Aitken mode concentration mainly originated from the tropical upper troposphere. The nucleation mode (Dp,g < 30 nm) was not observed frequently (14%) in free tropospheric condition at Mt. Norikura, but showed a tendency of occurrence after removal of Aitken and accumulation mode particles because of precipitation scavenging.
 Further long-term observation and statistical analyses at various high mountain sites are necessary to evaluate the factors determining the number-size distribution of aerosol particles and to elucidate the mechanism that maintains aerosol particle concentration in the free troposphere.
 The authors are grateful for the great support and help of the staff of the Norikura Observatory of ICRR, University of Tokyo, and Prof. T. Shibata, Drs. M. Kido and Y. Inomata. We acknowledge the NOAA Air Resources Laboratory (ARL) for provision of the HYSPLIT transport model and READY website at http://www.arl.noaa.gov/ready.html. Weather charts in this paper were provided by the Weather Chart CD published by Japan Meteorological Association (JMA). This work was supported by the Ministry of Education, Culture, Sports, Science and Technology Grants-in-Aid for Scientific Research on Priority Areas (10144104, 10144211, 1131210, 12018207), by a Ministry of Education, Culture, Sports, Science and Technology Grant-in-Aid for Scientific Research (c) 13680601, and by a grant from the 21st Century COE Program, no. G-4, Dynamics of the Sun-Earth-Life Interactive System.