Aerosol–Cloud Interaction at the Summit of Mt. Fuji, Japan: Factors Influencing Cloud Droplet Number Concentrations

Abstract

To investigate interactions between aerosols and clouds, the size and number concentrations of the cloud condensation nuclei (CCN) and the cloud droplets (CDs) were measured at the summit of Mt. Fuji (altitude 3776 m), Japan. The CCN number concentrations (N_CCN) are significantly higher in continental air masses than in air masses from the Pacific Ocean. The hygroscopicity parameter κ did not change much for different air mass origins, indicating that aerosol particles in the free troposphere are well mixed. Based on the CD number concentrations (N_CD), the degree of supersaturation in the ambient air during the cloud-shrouded period was estimated to be 0.15% (25th percentile) to 0.44% (75th percentile). To evaluate factors influencing the N_CD, measured N_CD were compared to ones calculated based on the Köhler theory using aerosol number size distributions, κ, and the degree of supersaturation. The results showed that N_CD could not be reproduced satisfyingly when the mean number size distribution or the mean effective supersaturation were used for the calculation. This study highlights the importance of obtaining information about the degree of supersaturation to predict N_CD in the atmosphere.

1. Introduction

Aerosol particles in the atmosphere can act as cloud condensation nuclei (CCN), which affect the Earth’s radiation budget, and consequently the Earth’s climate. The number concentration of CCN is a key factor for determining cloud optical properties and precipitation. The CCN activation of an aerosol particle is explained by the Köhler theory [1], and aerosol particles are more likely to act as CCN if the degree of water vapor supersaturation (SS) in the ambient air is higher, the particle size larger, and the particle more hygroscopic. The hygroscopicity of a particle strongly depends on its chemical composition. The relative importance of the size and chemical composition of aerosol particles in determining CCN concentrations at a given SS has been discussed elsewhere (e.g., [2,3]). Several publications state that the CCN activation of aerosol particles is mainly controlled by their size distribution rather than their chemical composition [2]. On the other hand, it has been reported that the chemical composition also matters, especially in regions with a mix of both fresh and aged aerosols [3]. In order to better understand the factors governing CCN number concentrations, it is important to accumulate CCN observation data at various sites. These data can contribute to the refinement of numerical predictions of both CCN and cloud droplet number concentrations.

With regard to the CCN activation of aerosol particles, both the particle size distribution and chemical composition of atmospheric aerosols have been measured at various locations (e.g., [4]), whereas the knowledge about SS is limited. Water vapor SS is an important parameter for CCN activation of aerosols and is necessary for numerical models predicting cloud/fog formation [5,6,7]. A critical degree of supersaturation (SS_c) can be determined for a single particle based on the Köhler theory. In the atmosphere, where the size and chemical compositions of aerosol particles are complex, a range of critical supersaturations is generated for a number of CCN. The SS at which the CCN concentration equals the cloud droplet number concentration is termed effective supersaturation (SS_eff) and is considered that the SS_eff exceeds the SS_c for all CCN active particles in the atmosphere [1]. Although direct measurement of the SS remains unattainable, it is possible to estimate the SS_eff indirectly from observation data. The reported values of the SS_eff have been limited to Europe [8,9,10,11], and SS_eff values from Asian regions have been limited [12].

Observations using airplanes (e.g., [13]), balloons (e.g., [14]) or helicopters (e.g., [10]) are powerful approaches to observe both aerosol and cloud particles in the atmosphere directly. Although these observations by aircraft have provided important findings of aerosol-cloud interactions, it is difficult to get long-term observation data. At mountain sites, orographic clouds frequently form when air flows upward along a mountain slope and the air temperature drops below the dew point due to adiabatic expansion. Observations at mountain sites, therefore, make it possible to observe the CCN activate into cloud droplets on-site over a long time period. Several studies have reported CCN properties at mountain sites. For example, at Jungfraujoch (altitude 3580 m), Switzerland, the impact of the aerosol size distribution on cloud droplet properties [15], CCN activation over the period of one year [16], temporal variations of aerosol hygroscopicity in the free troposphere [17] and scavenge of black carbon in liquid-phase clouds [18] have been studied. Henning et al. [19] reported changes in the hygroscopicity of aerosols through cloud-processing observed at the upwind and downwind sites of Mt. Schmücke (altitude 937 m), Germany. Friedman et al. [20] reported the impact of anthropogenic aerosols on CCN activation at Storm Peak Laboratory in Colorado (altitude 3210m), United States. Asmi et al. [9] compared CCN and actually activated particles inside a cloud by using a two-inlet system at Puy-de-Dôme, France (altitude 1465 m), and Hsiao et al. [21] characterized CCN at two different mountain sites in Asia, Doi Ang Kang (altitude 1536 m), Thailand and Mt. Lulin (altitude 2862 m), Taiwan.

In this study, we observed aerosols and clouds at the summit of Mt. Fuji (altitude 3776 m; Figure 1a), Japan, at the highest atmospheric monitoring site downwind of East Asia, during the summer of 2014. Because the summit of Mt. Fuji is mostly in the free troposphere, the observation site is less affected by local air pollutants and thus is a suitable site to observe differences in CCN characteristics of different air mass origins. During the observation period, the number concentrations of aerosol particles, CCN, and cloud droplets were measured simultaneously. Using the observation data of cloud-shrouded periods, the SS_eff of the ambient atmosphere is evaluated. In addition, the factors determining the cloud droplet number concentrations at the summit of Mt. Fuji are discussed, including the SS_eff, aerosol size distribution, and the chemical composition of aerosol particles.

2. Materials and Methods

2.1. Observation

Atmospheric observations were conducted at the Mt. Fuji Research Station (MFRS) (35.36° N, 138.73° E, altitude 3776 m) at the summit of Mt. Fuji (Figure 1a) from 18 July to 20 August of 2014. More details on the MFRS can be found elsewhere (e.g., [22,23]). Briefly, the MFRS consists of five buildings: Building 1, building 2, building 3, building 4 and a temporary building (Figure 1b). Aerosol observations were conducted at buildings 1 and 3 during summer only, when the commercial power supply was available. The instrumental setup used in this study is shown in Figure 1c.

Ambient air was drawn from an air inlet installed on the 2nd floor of building 1 using a conductive silicon tube (I.D. 3/8 in.) and introduced into a diffusion dryer filled with silica gel to dehumidify the sample air down to a relative humidity (RH) of 30% or less. Temperature and RH of the sampled air both upstream and downstream of the diffusion dryer were measured by a temperature and humidity sensor (HygroPalm23, Rotronic AG). After the diffusion drying, the size distribution of aerosol particles ranging in size from 10 to 500 nm in diameter was determined by a scanning mobility particle sizer (SMPS; SMPS3034, TSI Inc., Shoreview, MN, USA). Number concentration of total aerosol particles (condensation nuclei, CN) (N_CN) was calculated by the sum of the aerosol concentrations of particles in the range of 10 to 500 nm obtained by the SMPS. The particle loss at the diffusion dryer was calculated to be about 7% for the minimum size bin of the SMPS (the diameter midpoint is 10.4 nm), and less than 3% above 20 nm in diameter. The number size distributions were corrected on the basis of this calculation. Note that SPMS data are missing from 18:00 LT (LT = UTC + 9 h) 30 July to 07:00 LT 31 July, from 23:00 LT 1 August to 12:00 LT 3 August, and from 07:00 LT 7 August to 11:00 LT 14 August, due to maintenance and instrumental failure. Simultaneously, number concentrations of CCN (N_CCN) at 6 different SSs (0.02, 0.06, 0.16, 0.21, 0.36 and 0.55%) were measured using a CCN counter (CCNC, CCN-100, DMT Inc., Loveland, CO, USA). The CCN data obtained at the 0.02% and 0.06% SS conditions were not used for the following analysis because those SSs are below the operating limit (0.07%; [24]). The SSs were calibrated on-site by using (NH₄)₂SO₄ particles. For the calibration, the (NH₄)₂SO₄ particles were generated by atomization of an aqueous solution of (NH₄)₂SO₄ (Kanto Chemical Co., Inc., Tokyo, Japan, >99.5%) and introduced to a diffusion dryer filled with silica gel to dehumidify the (NH₄)₂SO₄ aerosol to RH of 30% or less. A differential mobility classifier (Model 3081, TSI Inc., Minnesota, USA) was used to classify the particles in the range of 30 to 200 nm every 10 nm, and the N_CN and N_CCN of the monodisperse aerosol particles were measured by a condensation particle counter (Model 3025A, TSI Inc., Shoreview, MN, USA) and the CCNC, respectively, to obtain CCN activation spectra. For the calibration, a Köhler model similar to the Reference Köhler model described by Brechtel and Kreidenweis [25] was applied.

Number concentrations (N_CD) and size distributions (from 2 to 50 μm in diameter) of cloud droplets (CD) were measured by a fog monitor (FM, FM-100, DMT Inc., Loveland, CO, USA) which was installed outside of building 3. The most frequent wind direction at the summit of Mt. Fuji was west [26]. Therefore, the FM was installed with its inlet facing the west slope because orographic clouds are likely to be generated by the wind ascending the westerly slope. Meteorological data, including air temperature, RH, atmospheric pressure, precipitation, wind speed and wind direction, were obtained by a meteorological data acquisition system.

2.2. Data Processing

The SMPS measured the number size distribution of the aerosol particles every 3 min. The hourly-averaged values of N_CN and the size distributions were used in the following analyses. The CCN counter measured N_CCN every second. It was operated at 6 different SS conditions and each SS condition continued for 10 min. Thus, it took 1 h to complete one cycle. Averaged values of N_CCN at each SS condition were calculated for every hour. The N_CCN data obtained at the 0.16, 0.21, 0.36 and 0.55% SSs were used for the following analysis. The FM measured the N_CD every 10 s and hourly-averaged values of N_CD were used for the following analyses.

When aerosol particles are mixed well internally, the particles with diameters larger than the activation diameter (d_a) can activate as CCN at a given SS. Then, N_CCN can be written as:

$$N_{\mathrm{CCN}}={{\displaystyle \int }}_{d_{\mathrm{a}}}^{d_{\max }}n\left(\log d\right)\mathrm{dlog}d,$$

(1)

where d and n(logd) are the dry diameter and the number size distribution of the aerosol obtained by SMPS, respectively. Then d_a can be obtained by integrating the particle size distributions. The maximum particle size measured by SMPS, 500 nm, was used as d_max.

The relationship between the saturation ratio, S, and the droplet diameter is described by the Köhlor theory [1]. The theory can be re-written by using the hygroscopicity parameter κ [27] as follows:

$$S=\frac{d_{\mathrm{drop}}^3-d^3}{d_{\mathrm{drop}}^3-d^3\left(1-\kappa \right)}\exp \left(\frac{4{\sigma }_{\mathrm{s}/\mathrm{a}}{M}_{\mathrm{w}}}{RT{\rho }_{\mathrm{w}}{d}_{\mathrm{drop}}}\right),$$

(2)

where d_drop is the droplet diameter, σ_s/a is the surface tension at the solution–air interface, M_w is the molecular weight of water, R is the universal gas constant, T is the droplet temperature, and ρ_w is the density of pure water. Note that the relationship between supersaturation, SS, and saturation ratio, S, is as below:

$$SS=\left(S-1\right)\times 100\%.$$

(3)

Equation (2) can be simplified as follows [27,28]:

$$S=\exp \left(\sqrt{\frac{4A^3}{27\kappa d^3}}\right),$$

(4)

The parameter A represents Kelvin’s term and can be written as:

$$A=\frac{4{\sigma }_{\mathrm{s}/\mathrm{a}}{M}_{\mathrm{w}}}{RT{\rho }_{\mathrm{w}}},$$

(5)

In Equation (5), A$\approx$0.66/T (in mm) can be used [29]. The κ of ambient aerosol was then calculated for each SS condition (0.16, 0.21, 0.36 and 0.55%) of the CCN counter by using d as d_a in Equation (4):

$$\kappa =\frac{4A^3}{27d_a^3{\ln }^{2}S}.$$

(6)

By applying Equation (1) and replacing N_CCN by N_CD, the activation diameters (d_{a, CD}) under ambient SS conditions were calculated based on the aerosol size distributions:

$$N_{\mathrm{CD}}={{\displaystyle \int }}_{d_{\mathrm{a},\mathrm{CD}}}^{d_{\max }}n\left(\log d_{\mathrm{p}}\right)\mathrm{dlog}{d}_{\mathrm{p}}.$$

(7)

To estimate the SS_eff of the clouds observed at the MFRS, we used the regression curve of the relationship between mean d_a, calculated by Equation (1), and SS obtained by the CCN measurements. Using the data collected by the FM, the liquid water content (LWC) and the effective radius (R_eff) of the cloud droplets were calculated using the following equations [15], respectively:

$$LWC=\frac{\pi }{6}{\rho }_{\mathrm{w}}{{\displaystyle \sum }}_{\mathrm{i}}{n}_{\mathrm{i}}{d}_{\mathrm{i}}^3,$$

(8)

$$R_{\mathrm{eff}}=\frac{1}{2}{{\displaystyle \sum }}_{\mathrm{i}}{n}_{\mathrm{i}}{d}_{\mathrm{i}}^3/{{\displaystyle \sum }}_{\mathrm{i}}{n}_{\mathrm{i}}{d}_{\mathrm{i}}^2,$$

(9)

where, n_i is the number concentration of cloud droplets and d_i is the geometric mean diameter of each size bin set by the FM.

2.3. Classification of Air Mass Transport Pathways

Backward air mass trajectories for the past 72 h were calculated every hour using the hybrid single-particle Lagrangian integrated trajectory (HYSPLIT) model [30,31]. The trajectories were classified based on their vertical and horizontal transport pathways. Vertical pathways that transported constantly above 2000 m altitude were classified as free troposphere (FT). For all others, the trajectories were classified as originating from the planetary boundary layer. For the horizontal pathways, the trajectories were classified into four categories: Continental, Pacific, and Others, according to where each trajectory started (Figure 2). The trajectories that started from the Asian continent and Pacific Ocean were classified as Continental and Pacific, respectively. The trajectories classified as Others included those originating from the marginal seas (East China Sea, Sea of Japan and Okhotsk Sea) or those positioned over the Japanese mainland for more than 48 h.

3. Results and Discussion

3.1. Overview

Figure 3 shows temporal variations in (a) air mass origins, (b) air pressure, precipitation, (c) air temperature, ambient RH (d) wind speed and wind direction throughout the observation period. About 84% of the backward trajectories were classified as FT (Figure 3a), indicating that the MFRS was in the free troposphere for most of the observation period. Thus, it is possible to see the effect of long-range transport of aerosols of different air mass origins on CCN characteristics. The trajectories classified as Continental are frequently observed in the early part of the observation period (20 to 29 July), whereas, those classified as Pacific are observed from about halfway through the observation period onwards (2 to 20 August) (Figure 3a). The mean air pressure during the observation period was 648.7 hPa. The drops of air pressure were observed on 27 July and 9 to 10 August, respectively, due to the passing of a cold front and the effect of the typhoon Halong (Figure 3b). The mean air temperature was 5.9 °C, with a range from 0.4 to 14.4 °C (Figure 3c). At the summit, northerly winds were dominant in July. In August, westerly winds prevailed except around 8 to 11 August, when easterly winds were observed along with the typhoon Halong passing through Japan islands (Figure 3d).

Figure 4 shows temporal variations in (a) air mass origins, (b) number size distributions of aerosol particles with 10 to 500 nm in diameter, (c) N_CN and N_CCN at each SS condition, and (d) LWC. Cloud-shrouded periods were defined as periods when the LWC is greater than 0.01 g m⁻³. As a result, 40% of the observation period was classified as cloud-shrouded (Figure 4d). In the following analysis, the data collected when (1) the backward trajectories were classified as FT, and (2) when the SMPS was in operation, were used.

3.2. CCN Properties and Their Relationship to Air Mass Origins

The mean value of N_CN for the entire observation period was 527.7 cm^–3. By air mass origin, the mean values of N_CN of the Continental, Pacific and Others periods were 617.9, 391.5 and 524.5 cm^–3, respectively. The 25th percentile, median, mean and 75th percentile values of N_CN and N_CCN are summarized in

Table 1. The 25th percentile, median, mean and 75th percentile values of N_CN. Number of data (N) for each air mass category is also indicated.

		NCN (cm−3)
	N	25th	Median	Mean	75th
Entire period	502	231.0	414.2	527.7	590.7
Continental	167	383.3	459.4	617.9	595.1
Pacific	188	153.8	205.1	391.5	433.4
Others	147	272.9	450.6	524.5	610.7

and

Table 2. The 25th percentile, median, mean and 75th percentile values of N_CCN at each SS. Number of data (N) for each air mass category is also indicated.

		NCCN (cm−3) (0.16% SS)					NCCN (cm−3) (0.21% SS)
	N	25th	Median	Mean	75th	N	25th	Median	Mean	75th
Entire period	487	14.2	41.3	80.9	117.8	491	24.4	61.3	107.8	160.8
Continental	167	44.5	82.6	176.1	117.2	167	68.1	114.9	155.3	228.0
Pacific	179	7.6	12.8	22.4	24.8	184	14.1	22.9	34.0	39.4
Others	141	30.1	62.3	112.3	181.2	140	43.1	90.4	147.9	230.2
		NCCN (cm−3) (0.36% SS)					NCCN (cm−3) (0.55% SS)
	N	25th	Median	Mean	75th	N	25th	Median	Mean	75th
Entire period	491	50.2	113.3	169.7	260.2	492	76.2	181.5	224.2	341.1
Continental	167	135.1	204.2	239.9	324.4	167	206.2	269.7	310.0	395.2
Pacific	185	27.9	46.2	64.4	73.3	185	37.8	64.6	92.5	103.0
Others	139	91.4	158.2	225.4	355.6	140	138.2	229.9	295.7	460.5

, respectively. The mean values of N_CN, N_CCN and the activation ratio (N_CCN/N_CN) at each SS condition showed clear differences between the periods classified as Continental and Pacific; those values were significantly higher for the Continental air masses than for the air masses from the Pacific (Figure 5a,b). The same trend was observed in a previous study at the MFRS [32]. The mean values of κ were calculated by Equation (6), to be 0.25 to 0.28 at each SS condition over the analysis period. These values were similar to values observed at other mountain sites, Jungfraujoch (0.13 to 0.29 at SSs 0.12% to 1.18%; [16]) and Storm Peak Laboratory (0.2 ± 0.15 at SSs 0.07% to 0.72%; [20]). The recent study summarized aerosol chemistry at alpine sites and aircraft measurements and reported that the main components of submicron non-refractory aerosol in high-altitude sites are sulfate, ammonium and organics [33]. Additionally, it was reported that ammonium and sulfate are the main inorganic ion species in the aerosol collected at the summit of Fuji [34]. The κ values observed at MFRS during summer can be explained by the mixtures of inorganics (κ = 0.61 to 1.28; [27]) and organics (κ = 0 to 0.2; [35]). Black carbon (κ $\cong$0) may affect the aerosol composition at MFRS especially when the airmass comes from the planetary boundary layer [36] although it could not be a major component. For the different air mass origins, the mean values of κ are 0.26 to 0.30 (Continental), 0.19 to 0.28 (Pacific) and 0.24 to 0.30 (Others) (Figure 5d). In the free troposphere, aerosol particles have a longer lifetime because they are less susceptible to deposition. Therefore, particles in the free troposphere are more aged and well-mixed than those in the planetary boundary layer (e.g., [33]). This could cause the relatively uniform chemical composition made up of mainly sulfate and aged (oxidized) organics regardless of airmass origins. Because smaller particles require larger SS to activate, each κ value at a certain SS reflects the chemical composition of the particles around d_a. The κ values are relatively uniform regardless of the particle size (Figure 5c,d), indicating that no size dependency of the chemical composition of the aerosol at the MFRS during summer exists. This result differs from other CCN studies at non-alpine sites (e.g., [37,38,39]), in which smaller particles were shown to be less hygroscopic, suggesting that the aerosol in the free troposphere is well-mixed compared to the aerosol sampled closer to the sources.

The κ values are similar between air mass origins, while the particle size distribution is strongly dependent on the particle source (Figure 6). The mode diameter of the aerosol particles of Continental origin (58 nm) is significantly larger than that of Pacific origin (24 nm). It is thought that aging processes, such as condensation, coagulation and cloud processing, change number size distributions (e.g., [40]). The aerosols of Continental origin are more likely to be affected by these aging processes because an abundant supply of gaseous and particulate substances from these sources is likely. The larger the size of the aerosol, the easier it is to become activated CCN. From the above results, it can be concluded that the particle size distribution is the most important factor determining N_CCN in the free tropospheric airmass at the MFRS during summer.

3.3. Evaluation of SS_eff and Cloud Droplet Properties

The cloud droplet concentration (N_CD) during the cloud-shrouded periods ranged from 1.6 to 389.1 cm⁻³ (mean: 78.3 cm⁻³). By using Equation (7), the activation diameters (d_{a, CD}) under ambient SS conditions were calculated and they ranged from 31.6 to 379.6 nm. To estimate the SS_eff of the clouds observed at the MFRS, we used the regression curve of the relationship between mean d_a and SS obtained by the CCN measurements (Figure 5c):

$$d_{\mathrm{a},\mathrm{CD}}=46.32\times {{SS}_{\mathrm{p}}}^{-0.65}.$$

(10)

Based on these calculations, the SS_eff during the cloud-shrouded periods is estimated to be 0.15% (25th percentile) to 0.44% (75th percentile) (Figure 7). Asmi et al. [9] estimated the in-cloud SS at the Puy-de-Dôme station (altitude 1465 m) to be 0.1% to 0.6%, and Hammer et al. [11] reported an SS_eff of cumulus and shallow-layer clouds at Jungfraujoch (attitude 3580 m) of 0.37% to 0.5% and 0.17% to 0.3%, respectively. The values of SS_eff calculated in this study were comparable to the values of SS of clouds at mountain sites in previous studies. When compared to air mass origins, no significant differences among the SS_eff of Continental, Pacific and Others were apparent (Figure 7). A positive correlation (R = 0.54, p <0.01) between N_CD and N_CCN under the 0.21% SS condition, which is close to the median value of SS_eff, is observed. The points are scattered around an 1:1 line and large scatter is observed when LWC is lower (Figure 8).

Parameters characterizing the clouds at the MFRS are summarized in

Table 3. Mean values of SS_eff, w, N_CD, LWC and R_eff for clouds at the MFRS classified according to air mass origin. Standard deviations are indicated in parentheses.

		Air Mass Origin
	Entire Period	Continental	Pacific	Others
SSeff (%)	0.34 (0.34)	0.24 (0.14)	0.44 (0.41)	0.26 (0.26)
w (m s−1)	1.29 (1.21)	1.03 (1.32)	1.40 (0.83)	1.37 (1.63)
LWC (g m−3)	0.05 (0.05)	0.08 (0.06)	0.03 (0.02)	0.06 (0.05)
NCD (cm−3)	78 (78)	144 (92)	36 (32)	86 (65)
Reff (μm)	8.1 (2.5)	6.5 (1.7)	9.3 (2.5)	7.6 (2.1)

. Note that these parameters were obtained only during the cloud-shrouded periods. It was reported that SS tends to be higher at higher updraft wind velocities (e.g., [8]), because adiabatic expansion is an important factor for the determination of SS in a cloud. In this study, the updraft wind velocity was not measured. Hammer et al. [11] proposed a method to calculate the updraft velocity by multiplying the horizontal wind speed by the tangent of the angle of the mountain slope:

$$w=v_{\mathrm{h}}\times \tan \theta ,$$

(11)

where $w$, $v_{\mathrm{h}}$ and $\theta$ are updraft velocity, horizontal velocity, and the angle of mountain slope, respectively. Equation (11) assumes that the airmass rises along the slope of the mountain, otherwise it is overestimated. Here, it was used as a standard of the maximum value of the updraft velocity. As for the angle of the mountain slope, 38° (the angle of west side) was used. Note that Mt. Fuji is a stratovolcano with conical shape, and the angle of the slope is almost same regardless of the direction. The mean updraft velocity was slightly higher when the airmass origin was Pacific than when it was Continental. On the whole, however, the difference of updraft velocities by the airmass origin was not remarkable. The topography of the summit of Mt. Fuji is complicated due to the existence of a crater. This will disturb the air stream, so the actual updraft velocity was likely to be different from the estimate. In the future, it will be necessary to monitor the updraft velocity on site by installing multiple ultrasonic anemometers to understand the factor controlling SS.

The values of LWC range from 0.01 to 0.26 g m⁻³, and were similar to those of stratocumulus, stratus and fog (0.05–0.3 g m⁻³; [41]) (Table 3) There was a clear difference in N_CD between air masses of Continental and Pacific origin as observed in N_CCN (Figure 5a). The mean N_CD in Continental airmass (144 cm⁻³) was similar to the annual mean value in East Asia region estimated by the satellite observation (136.9 cm⁻³) [42], however, the mean N_CD during the entire period (78 cm⁻³) was less than the annual mean value. This could be caused by the seasonality of prevailing winds in East Asia: winds come from the Pacific under the Pacific high-pressure system during summer, but in other seasons, winds come mostly from the Asia continent. It is considered that the period (summer) in which we conducted our observations was the season with the lowest N_CD of the year in the region. The normalized size distributions of N_CD were illustrated in Figure 9. The shapes of size distributions resembled each other, but the CDs larger than 20 μm were more in the Pacific airmass. The calculated R_eff were 6.5 and 9.3 μm, respectively, in the Continental and Pacific airmasses. If the water content is constant in the atmosphere, R_eff of cloud particles decreases as N_CD increases. This results in a change of the optical properties of the cloud and is known as the Twomey effect [43]. In this study, LWC of Continental airmasses is larger than that of Pacific airmass. This could be due to the higher N_CD of Continental air masses, because LWC was a function of both number and size. Although the values of LWC differed, the inversely proportional relationship between N_CD and R_eff was still observed between Continental and Pacific airmasses.

3.4. Factors Controlling the Cloud Droplet Concentration

To investigate factors controlling the N_CD, N_{CD, cal} was calculated by using the aerosol number size distribution, the hygroscopicity parameter (κ), and the SS_eff. Where κ and the ambient SS were known, the activation diameter (d_{a, cal}) can be calculated based on the Köhler theory (Equations (2) or (4)). Then, the N_CD can be calculated by inputting the calculated d_{a, cal} values and the size distribution into Equation (1). Here, each N_{CD, cal} was calculated for the following three patterns: (1) mean size distribution shape, (2) mean SS_eff, and (3) mean hygroscopicity parameter κ (at 0.21% SS of the CCN measurements) for the cloud-shrouded periods. Then the calculated value was compared to the measured value (N_CD). As a result, relative errors were 278.7%, 139.4% and 43.4%, respectively, when size distributions, SS_eff and κ were fixed to their mean values. Therefore, the N_{CD, cal} can be estimated with ±50% error by using the mean value of κ during cloud-shrouded periods (Figure 10c). However, the N_{CD, cal} cannot be reproduced well if the mean number size distribution or the mean SS_eff are used in the calculations (Figure 10a,b). These results indicate that the size distribution is the most important factor controlling the cloud droplet number concentration at the MFRS, but that the concentration can be reproduced reasonably well when using a representative aerosol chemical composition. In addition, the evaluation of the SS_eff seems to be of greater importance than the chemistry for accurately predicting cloud droplet number concentrations at the MFRS during summer, where well-mixed, free tropospheric aerosol is observed.

4. Concluding Remarks

Atmospheric observations at a high mountain site located in the free troposphere, where the influence of long-range transport is significant, revealed clear differences in CCN and cloud droplet characteristics for different air mass origins. Both CCN and cloud droplet concentrations were significantly higher in continental air masses compared to those in air masses from the Pacific Ocean. The effective radius of the cloud droplets was also affected by the number concentration of cloud droplets, with smaller radii in the continental air masses. Based on the CCN characteristics and cloud droplet number concentration obtained by our observations, the water vapor supersaturation of the atmosphere was estimated to be equivalent to that estimated at other alpine sites. A calculation based on the Köhler theory to investigate the factors controlling the cloud droplet concentration showed that the most important factor is the aerosol particle size distribution, followed by the supersaturation and the chemical composition.

In evaluating the climatic impact of aerosol–cloud interactions, it is highly important to predict the cloud droplet number concentration which determines the optical properties of clouds. In this study, we clarified for the first time in East Asia that it is important to know the supersaturation degree of cloud rather than the chemical composition of aerosol particles for the prediction of cloud particle number concentration by the simultaneous observation of aerosol and cloud. This would be true in the case of well-mixed aerosols at locations remote from the sources like the free troposphere. Although it is difficult to directly measure the degree of cloud supersaturation, it will lead to a more precise prediction of cloud droplet number concentrations if the cloud supersaturations in various atmospheric environments are known by using the indirect method proposed in this study. For further understanding, it is also necessary to know factors controlling the degree of supersaturation simultaneously. Updraft velocity is an important factor, but is difficult to reproduce with a meteorological model in field observation of alpine sites where air stream is complicated. When conducting similar observations in mountain sites, it is necessary to acquire observation data of updraft velocity by installing multiple ultrasonic anemometers or releasing radiosondes from the foot of the mountain periodically.