Cloud condensation nuclei activity at Jeju Island, Korea in spring 2005

We measured the number concentrations of cloud condensation nuclei (CCN) and the size distributions of CCN/CN (CN: condensation nuclei) ratios at supersaturations ( SS s) of 0.097, 0.27, 0.58, and 0.97% at Jeju Island, Korea during March-April 2005. We made simultaneous measurements of aerosol inorganic ions, water-soluble organic carbon (WSOC), organic carbon (OC), and elemental carbon (EC) in PM 2.5 . The CCN/CN ratios increased with increasing particle diameter, and the diameter at CCN/CN=0.5 was defined as D 50 . D 50 represents the activation dry diameter of atmospheric particles. The average D 50 at SS =0.097% and 0.97% was 136±17 nm and 31±3 nm, respectively. The temporal variation of D 50 at SS =0.097% was correlated with the mass fraction of water-soluble components (inorganic ions + WSOC), indicating that the temporal variation of CCN activity was mainly controlled by changes in the water-soluble components fraction. The critical dry diameter ( D crit ), which is the threshold dry diameter for CCN activation, was calculated from the observed aerosol chemical compositions by Kohler theory for comparison with D 50 . The D 50 at SS =0.097% was correlated ( r 2 =0.48) with calculated D crit , although D crit was larger than D 50 by 20–29% on average. The systematic difference between D 50 and D crit could be caused by the size dependence of the aerosol chemical compositions or surface tension lowering caused by the mixing of water-soluble organic compounds. This difference corresponds to a 27±14% uncertainty in the CCN number concentration estimated from the observed particle number size distribution.


Introduction
A subset of atmospheric particles acts as cloud condensation nuclei (CCN). An increase in CCN number concentration causes an increase in cloud droplet concentration and a decrease in droplet size, which in turn impacts cloud albedo and precipita-threshold diameters for CCN activation are compared with the mass fractions of aerosol chemical compositions. Threshold diameters for the activation (D crit ) are calculated and compared with the observed threshold diameter to investigate the controlling factors of D crit . 10 The equilibrium water vapor pressure (S) of an aerosol particle can be calculated by Köhler theory. According to the theory, S is described as follows (e.g., Roberts et al., 2002;Mochida et al., 2006):

Theory
where d dry is the dry diameter of a particle and d wet is the diameter of a droplet under 15 equilibrium conditions. The suffix i denotes the properties of the i -th solute compound. M w is the molecular weight of water; M i is the molecular weight of solute;ρ w is the density of water; ρ i is density of solute; σ is the surface tension; R is the gas constant; T is the temperature; ν i is the stoicheiometric number of ions and molecule; φ i is the osmotic coefficient; ε i is the degree of dissolution; and m i is the mass mixing 20 ratio of the i -th solute. The first term on the right-hand side represents an increase in the equilibrium vapor pressure of water due to surface tension (the Kelvin effect). The second term on the right-hand side denotes the decrease in the equilibrium vapor 15808 If we assume that A and B are constants and d wet is much larger than d dry , SS c can be expressed as follows: This equation shows that SSc decreases with particle diameters. Thus, at a certain SS, there exists a threshold value of d dry above which all aerosol particles act as CCN. 10 We denote this diameter as the critical dry diameter (D crit ). Figure 1 shows the CCN observation system used for this study. In this system, ambient particles were dried to a relative humidity (RH) <5% using two diffusion dryers in 15 series (TSI Model 3062). Dried particles were charged with a 241 Am bipolar neutralizer. Particles were then introduced to a differential mobility analyzer (DMA: TSI Model 3081). The DMA classified particles by their electrical mobility. The voltage applied to the DMA was scanned stepwise to change the diameter of the classified particle ( Table 1). The sheath to sample flow ratio of the DMA was set to 10:1. Classified calibrated with ammonium sulfate particles as described by Kuwata et al. (2007). The calibration was performed at the observation site before and after the campaign. There are several different methods for the calculation of the water activity of ammonium sulfate particles (Kreidenweis et al., 2005, and references therein). We chose the ideal solution approximation (φ=1) in calculating the D crit of atmospheric particles (Sect. 5). 10 Therefore we used this approximation for the interpretation of the calibration results to ensure consistency. For comparison, SSs calculated using the osmotic coefficient of Clegg et al. (1996) are also shown in parentheses in Table 1. The differences of these two values give a measure of the uncertainty in the estimation of the SS in the CCN counter. The Debye -Hückel constant is needed to calculate the osmotic coefficient 15 of Clegg et al. (1996) because it is based on the Pitzer model. The Debye -Hückel constant at 300 K was calculated by the equation given by Clegg et al. (1994). Multiply charged particles were included in the classified particles, therefore an inverse analysis was performed for CCN and CN size distribution data for multiple-charge correction. The STWOM algorithm (Markowski, 1987) was used to obtain size distribu-20 tions of CN and CCN. In this calculation, the equilibrium charge distribution (Wiedensohler, 1988) and the DMA transfer function derived by Knutson and Whitby (1975) were included in the kernel function. The raw data for CN and CCN were linearly interpolated so that the interval of each bin was ∆logd p =0.015. Size-resolved CCN/CN ratios were calculated using the data after the inverse analysis. 25 The three-way valve in Fig. 1  , Cl − ) were measured by a particle-into-liquid sampler combined with ion chromatography (PILS-IC) (Orsini et al., 2003;Takegawa et al., 2005). The concentration of water-soluble organic carbon (WSOC) was measured by PILS combined with a total organic carbon analyzer (PILS-5 WSOC) (Sullivan et al., 2004;Miyazaki et al., 2006Miyazaki et al., , 2007. Elemental carbon (EC) and organic carbon (OC) were measured by a semi-continuous thermal-optical carbon aerosol analyzer (Sunset Laboratory, Inc.) Kondo et al., 2006;Miyazaki et al., 2006). PM 2.5 cyclones were used for these instruments. The detection limits of the PILS-IC, PILS-WSOC, OC, and EC were estimated to be 0.01 µg/m 3 , 10 0.1 µg/m 3 , 1.0 µg/m 3 , and 0.2 µg/m 3 , respectively (Takegawa et al., 2005;Kondo et al., 2006;Miyazaki et al., 2006). Aerosol number size distribution (10-300 nm) was measured with a scanning mobility particle sizer (SMPS 3936, TSI). The SMPS used in this study comprised a DMA (TSI Model 3081) and a CPC (TSI Model 3010). In addition, number concentration 15 of particles larger than 10 nm (CN) was measured by another CPC (TSI Model 3010) (Yum et al., 2007). The concentration of carbon monoxide (CO) was measured using a non-dispersive infrared analyzer (Horiba APMA-360 model) (Sawa et al., 2007).

Measurement site
The observations were performed between 18 March and 5 April 2005 at Gosan 20 (33.2 • N, 126.1 • E) on Jeju Island, Korea, as part of the Atmospheric Brown Cloud -East Asian Regional Experiment 2005 campaign. The location of Gosan is shown in Fig. 2. The instruments were placed in a container located about 10 m back from the edge of a cliff. The sampling inlets were made of stainless steel tubes with an inner diameter of 7 mm. The top of the inlets was located about 4 m above the ground level. 25 The meteorological parameters at the Gosan site were observed by Korean Meteorological Administration (KMA

EGU
were northerlies and north-northwesterlies (more than 50% of the observation period) associated with the Siberian high-pressure system. This led to the frequent transport of anthropogenic pollutants from the Korean Peninsula and China to Gosan. Sawa et al. (2007) attributed the high concentrations of CO at Gosan to the emissions from these regions using a chemical transport model. More detailed description of the mete-5 orological conditions and the transport of anthropogenic pollutants can be found elsewhere (Sawa et al., 2007;Miyazaki et al., 2007). Previous studies have also shown that air masses observed at Gosan are heavily influenced by anthropogenic emissions from East Asia (Carmichael et al., 1997;Lee et al., 2006). on 26 March, the CCN/CN size distribution of ambient particles was very similar to that of ammonium sulfate, indicating that almost all particles were composed of ammonium sulfate at that time. On the other hand, on 28 March, the CCN/CN size distribution shifted to a larger diameter, and the rate of increase was lower than ammonium sulfate. This shift indicates that the bulk aerosol chemical composition on 28 March was 20 significantly different from ammonium sulfate. The slower increase rate indicates the co-existence of different types of aerosol particles with different activation curves. This can be interpreted as co-existence of aerosol particles with different chemical composition considering that each activation curve depends on the chemical composition composing each type of particles.

Results
EGU fraction of chemical compounds and theoretically calculated D crit . In this study, the threshold diameter was defined as the diameter corresponding to CCN/CN=0.5 (D 50 ) because it represents the bulk chemical composition. If aerosol particles were not fully activated, D 50 would not necessary corresponds to the threshold diameter. In all cases shown in Fig. 3 Table 2. The average CCN number concentration at SS=0.097% and 0.97% were 1200 cm −3 and 4000 cm −3 , respectively. These concentrations are higher than those of other remote areas of the world such as the Island of Tasmania in Australia and Mace Head in Ireland by about 20 an order of magnitude (Covert et al., 1998;Reade et al., 2006). CCN concentrations measured at Gosan during this observation period were extensively compared with those obtained in other regions by Yum et al. (2007). The CCN concentrations observed in this study are close to those of Anmyeon (Korea: see Fig. 2) in springtime of 2004 (SS=1%)   Figure 4b shows the time series of the number size distribution of aerosol particles and D 50 . The D crit of ammonium sulfate at each SS is also shown as dashed lines in this figure. The average values of D 50 at SS=0.097% and 0.97% were 136 nm and 31 nm, respectively (Table 2). In general, the D 50 is almost equal to or slightly larger (by ∼25%) 5 than the D crit of ammonium sulfate (125 nm and 27 nm, respectively). This means that the D crit of ammonium sulfate is the smallest D crit of the atmospheric particles during the observation period.
The temporal variation of D 50 at different SSs did not always correlate. As shown in Eq. (2), the threshold diameter for CCN activation depends on A and B, which are 10 determined by the aerosol chemical compositions. Thus, this difference in the temporal variations of D 50 at each SS indicates the difference in temporal variation of chemical composition in different size ranges.
Using Eq.
(2), we calculated B assuming the surface tension of water. This parameter gives information on the chemical composition (approximate number of solute ions 15 and molecules included in a unit volume) at D 50 of each SS. The results are summarized in Table 2. The average values of B did not depend on SS significantly, indicating that the chemical composition averaged over the observation period was rather uniform in the diameter range considered (30 to 160 nm).

Number size distribution 20
New particle formation events can have significant impact on CCN number concentration (e.g., O'Dowd et al., 2002;Laaksonen et al., 2005) and aerosol hygroscopicity, which is closely related to CCN activity (Buzorius et al., 2004). However, there has been no clear observational evidence of this effect. In Fig. 4b, new particle formation is clearly identified on 19, 25, 29, 30, and 31 March. It can be seen more clearly in EGU of particle number concentrations larger than 30 nm. Thus, the difference of the number concentrations (CN−CCN (SS=0.97%)) represents the number concentration of particles between 10 and 30 nm. During the periods of new particle formation events described above, the enhancement of number concentration of small (10-30 nm) particles was observed. In particular, the events occurring on 29 and 30 March were 5 important in that newly formed particles influenced the CCN number concentration as a consequence of particle growth beyond D 50 . Figure 5a and b show the CCN number concentrations, particle size distributions, and D 50 during this event. The peak diameter of the size distribution obtained by bimodal lognormal fitting is shown as red lines in Fig. 5b. 10 At 14:00 local time (LT) on 29 March, small (<20-nm) particles appeared and began to grow. The peak diameter grew to 25 nm by 18:00 LT, and some particles grew larger than the D 50 at SS=0.97% (28 nm). At this time, the CCN number concentration (SS=0.97%) began to increase. The peak diameter and D 50 (SS=0.97%) were equal at 21:00 LT (blue dashed vertical line in Fig. 5b). At this time, the majority of 15 newly formed particles began to act as CCN at SS=0.97%. In the case of SS=0.58%, the peak diameter equaled D 50 at 02:00 LT on 30 March, and the CCN number concentration increased from 1700 cm −3 (01:30 LT) to 5800 cm −3 (11:30 LT). At 03:00 LT, some portion of the particles grew large enough to act as CCN at SS=0.27%. Then, CCN number concentration at SS=0.27% increased from 1000 to 4700 cm −3 with the 20 increase in the peak diameter. For this SS, the peak diameter reached D 50 at 10:00 LT. At the same time, another new particle formation event occurred, and this event also clearly affected the CCN number concentration at SS=0.97% and 0.58%. At 12:00 LT, some fraction of the particles grew larger than the D 50 at SS=0.097%, and they affected the CCN number concentration at this SS. A similar phenomenon was also ob-25 served on 25 March (Figs. 4a and b). These results clearly show that the newly formed particles significantly increased the CCN number concentration. Buzorius et al. (2004) have shown that the deliquescence relative humidity and the hygroscopic growth of newly formed particles at Gosan during ACE-Asia campaign EGU were similar to those of ammonium sulfate. In the present study, the observed D 50 of newly formed particles was also similar to ammonium sulfate. For instance, the D 50 of SS=0.97% was 27 nm at 21:00 LT, 29 March, and that of SS=0.27% was 66 nm at 10:00 LT, 30 March. The D crit of ammonium sulfate was 27 nm and 63 nm, respectively (Table 1). This result indicates that newly formed particles observed in this event were 5 mainly composed of inorganic compounds such as ammonium sulfate, consistent with Buzorius et al. (2004).
Previous studies have suggested that new particle formation has an impact on the CCN number concentration from the measurements of number size distributions (e.g., O'Dowd et al., 2002;Laaksonen et al., 2005) and modeling (e.g., Arnold, 2006;Sotiropoulou et al., 2006). The present observations clearly demonstrate that new particle formation is one of the important processes of CCN formation at Gosan. Figure 6 shows the number size distribution of all particles and CCN averaged over the whole observation period. Averaged CCN/CN ratios were multiplied by the number size distribution measured by the SMPS averaged over the observation period to obtain 15 the rough estimate of the CCN size distribution. The peak diameter of the average CCN size distribution was about 150 nm at SS=0.097% and shifted to about 80 nm at SS=0.97%. Detailed discussion of number size distribution of particles during the observation period has been given by Yum et al. (2007).

Ion balance of inorganic components 20
The average concentrations of the inorganic components measured by PILS-IC are summarized in Table 3 EGU the following discussion of CCN activity. Na + , Mg 2+ , and Cl − were also ignored for the same reason. Figure 7 shows the ion balance of NH + 4 , NO − 3 , and SO 2− 4 . The cation and anions balance very well (slope = 0.94). This result shows that sulfate and nitrate in PM 2.5 were neutralized by ammonium at Gosan. The addition of K + to the ion balance causes 5 the balance to deviate from the 1:1 line. This suggests that neither NO − 3 nor SO 2− 4 were the counter ions of K + . In addition, because the molar concentration of K + was only 8% of NH + 4 , we ignored K + . In the following discussion, we assume that the inorganic component of sub-micron particles was composed only of ammonium sulfate and ammonium nitrate.

EGU
sulfate. This indicates that B is mainly determined by inorganic components because hygroscopicity is mainly determined by B (e.g., Kreidenweis et al., 2005;Mochida et al., 2006). This is consistent with the present observations because the mass fraction of inorganic compounds was significantly larger than that of WSOC during this period (Fig. 8).

5
The D 50 at SS=0.097% did not necessarily correlate with those of higher SSs (Fig. 4b), as discussed in Sect. 4.2.2. Correlation of D 50 at SS=0.097% with those at higher SSs decreased with the increase in SS (r 2 =0.25 and 0.00 for SS=0.27% and 0.97%, respectively). This indicates that the temporal variation of the mass fraction of PM 2.5 was reflected in that of D 50 at SS=0.097% (100-200 nm) but was not reflected in D 50 at higher SSs (<100 nm). This is because the contribution to the PM 2.5 mass concentration of particles with diameters smaller than 100 nm was much smaller than that of 100-200 nm.

Comparison of D 50 and calculated D crit
In this section, we compare the observed D 50 and the D crit calculated from the simul- 15 taneously measured chemical composition. We first discuss the assumptions used to calculate A and B in the Köhler theory (Sect. 2). The surface tension of water (72 mN/m) was assumed for the calculation of A. The calculation was performed at T =300 K. For the calculation of B, the ideal solution approximation (φ=1) was used. We used the simultaneously measured aerosol chemical composition (PM 2.5 ) assum-20 ing that chemical composition was not size-dependent. Assumptions regarding the chemical composition and chemical properties of organic compounds (e.g. molecular weight (MW), elemental ratio, density) are also required for the calculation of B. Kawamura et al. (2003) and Mochida et al. (2003) measured dicarboxylic acids and other water-soluble organic compounds over the Sea of Japan and Yellow Sea during the 25 ACE-Asia campaign. The concentration of oxalic acid was higher than that of other compounds by an order of magnitude. Simoneit et al. (2004) also measured dicarboxylic acids at Gosan during ACE-Asia campaign. Adipic acid was the largest linear 15818 tion. The average molecular weight of WSOC should be higher than that of oxalic acid because dicarboxylic acids and saccharides with larger molecular weights were also observed at Gosan (Simoneit et al., 2004). In fact, Miyazaki et al. (2007) indicated that a significant portion of WSOC measured during ABC-EAREX2005 campaign could be attributed to organic compounds having a MW larger than oxalic acid. To test the sen-10 sitivity of D crit on the assumed WSOC composition, we also calculated D crit assuming that the average properties (MW, elemental composition, and density) of water-soluble organic compounds are equal to that of adipic acid. In this study, this assumption is called the "adipic acid assumption." In addition, recent studies have shown that significant fractions (20-60%) of WSOC are high-molecular weight compounds such as . In these calculations, water-soluble organic compounds were assumed to be completely dissolved in water (ε=1). 25 The chemical properties of WIOC are also required for the calculation of B. EGU (-CH2-). Therefore, we assumed that all carbon atoms in WIOC originated from methylene groups. The density of WIOC was assumed to be 0.8 g/cm 3 because the densities of hydrocarbons and fatty acids are typically 0.7-0.9 g/cm 3 (Pang et al., 2006). The properties of EC were assumed to be equal to those of graphite. The values used for the calculation are summarized in Table 4.

5
The observed D 50 and calculated D crit at SS=0.097% are compared in Fig. 9. This figure shows three calculated results based on the oxalic acid, adipic acid, and HULIS assumptions. In all cases, the temporal variations of the calculated D crit correlated with the observed D 50 . However, the calculated D crit values are systematically larger than the observations by 20-29%. 10 The correlations of D 50 and the calculated D crit (adipic acid assumption) are shown in Fig. 10. At all SSs, the calculated diameters are larger than the observations. The differences between the calculated D crit and the observed D 50 (calc -obs) are summarized in Table 5. In all cases, D crit was overestimated by 16-29%.
The r 2 values decrease with increasing SS (diminishing D 50 ). This is likely due to 15 the size dependence of the temporal variation of the aerosol chemical compositions as discussed in Sect. 4.2.2.

Possible causes of the discrepancy
In this section, we discuss the possible causes of the difference between D 50 and D crit and identify critical assumptions for the calculation of D crit .

20
In calculating B, we assumed the composition and chemical properties of each component. The average chemical properties of water-soluble organic compounds were assumed to be identical to those of oxalic acid, adipic acid, or HULIS. Nevertheless, the D crit values were larger than D 50 even in the case of the oxalic acid assumption (the maximum estimation of B). Therefore, the assumption on the chemical composition of 25 WSOC is not the main cause of this discrepancy.
We assumed a density of 0.8 g/cm 3 for water insoluble organic compounds. If the actual density was larger than this value, this leads to the overestimation of D crit be-15820 culation. If the discrepancy is due to this effect, the osmotic coefficient at the critical droplet diameters needs to be increased by a factor of about 2, considering the magnitude of the discrepancy (Table 5) and Eq. (2). Nevertheless, such a large change in the osmotic coefficients is unlikely, as the solution was very dilute at the critical droplet diameters. 10 In this study, the aerosol chemical composition was measured at PM 2.5 . If the chemical compositions of PM 2.5 do not represent those at D 50 (<200 nm), it causes an error in the calculation of D crit . We discuss this point in detail in Sect. 5.3.1.
The surface tension of water was assumed for the calculation of A in Eq.
(2). If the decrease in surface tension due to WSOC was significant, it may affect the D crit 15 (Facchini et al., 1999). This point is discussed in Sect. 5.3.2.

Size dependence of chemical composition
Some studies have used size-resolved aerosol chemical composition for CCN closure studies (e.g., Cantrell et al., 2001;Roberts et al., 2002). These studies have shown the size-dependence of aerosol chemical composition. In addition, Medina et al. (2007) 20 have shown that the use of size-resolved chemical composition can decrease the error in the closure of the CCN number concentration. Topping et al. (2004) measured sizeresolved chemical composition at Gosan during ACE-Asia campaign using a Berner impactor. They reported the size distribution of mass fractions of inorganic compounds and WSOC, and the results clearly showed the size-dependence. However, it is diffi-25 cult to use their results to estimate the effect of size-dependent chemical composition because they did not measure water-insoluble components of the impactor samples. Mochida et al. (2007)  cal composition was affected by super-micron particles by comparison with the ratio of Mochida et al. (2007). The D crit for each SS was calculated using the chemical composition of the submicron mode particle given in Table 2 in Mochida et al. (2007) assuming that OC was entirely composed of WIOC because the fraction of WSOC was much smaller than that of WIOC (Fig. 8) and WSOC concentration was not reported. The major difference of this calculation is the higher sulfate fraction ([OC]/[nss-SO 2− 4 ]≈0.2). At SS=0.097%, the observed D 50 was 136±17 nm, and the calculated D crit using the data of Mochida et al. (2007) was 142 nm, as summarized in Table 6. In the case of other SSs, the D crit also agree with the average value of D 50 . This result shows possible effects of the 15 size-dependent chemical composition on CCN activation.
The size dependence of aerosol chemical composition is clearer for the period of new particle formation. The D 50 values of newly formed particles were very close to that of ammonium sulfate, as discussed in Sect. 4.2.3. In particular, newly formed particles grew larger than D 50 at SS=0.097% on 31 March (Fig. 4), and D 50 values were nearly 20 equal to the D crit of ammonium sulfate on that day (Fig. 9). However, the mass fraction of carbonaceous aerosols (WSOC + WIOC + EC) at PM 2.5 was about 30% at this time (Fig. 8), and this led to the overestimation of D crit (Fig. 9). The decrease of D 50 associated with new/secondary particle formation was also observed on 25 March. In this case, the mass fraction of water-insoluble compounds was also 30-40%, and 25 this caused the overestimation of D crit . These results suggest that the PM 2.5 mass concentration was biased by large (>200-nm) particles at least during these periods. Introduction As described above, the surface tension of water was assumed for the calculation in this study. Nevertheless, if the decrease in surface tension due to organic compounds was significant, it may affect the critical SS of the particles (Facchini et al., 1999). Surface tension lowering effects have been observed in various regions of the world 5 such as Po Valley in Italy (Facchini et al., 1999), Mace Head in Ireland (Cavalli et al., 1999), and the Great Hungarian Plain (Kiss et al., 2005). In particular, Decesari et al. (2005) measured the surface tension lowering effect of aerosol, cloud water, and wet deposition samples at Jeju Island during the ACE-Asia campaign. They showed that the decrease of the surface tension due to aerosol was relatively small, whereas the 10 surface tension decrease of cloud water and wet-deposition samples was significant. McFiggans et al. (2006) compared these results, and they showed that the magnitude of the effect for cloud water at Jeju Island was the most significant, and the effects for Po Valley and Mace Head samples were not as large as that of cloud water at Jeju (the red and blue lines in Fig. 11). They summarized the fitted parameters of Szyszkowski-15 Langmuir (Eq. 3) obtained for these observation results, where σ 0 is the surface tension of pure water, C is the WSOC concentration of the solution, and a and b are the empirical parameters obtained by fitting the observational results. In order to investigate the sensitivity of D crit to surface tension, we calculated 20 the D crit at each SS using the parameters given by McFiggans et al. (2006). In this calculation, the adipic acid assumption was used, and the values for Jeju cloud water and Po Valley fog were employed. Substituting σ in Eq. (1) by that expressed by Eq. (3), the D crit values were obtained by numerical calculation. The results are summarized in Fig. 12. In the case of Jeju cloud water, the calculated diameters for all SSs are smaller 25 than the observed D 50 by 10-47%, and the use of the equation for Po Valley decreased the differences between D crit and D 50 (7 to −13%). This indicates that the decrease of surface tension can potentially explain the discrepancy.

EGU
critical SS was about 46-56 mN/m (Jeju cloud water) and 51-62 mN/m (Po Valley), respectively (Fig. 11). The discussion in this section and Sect. 5.3.1 shows that the size-dependence of chemical composition and the decrease in surface tension are the critical parameters in estimating D crit . For a more quantitative assessment of these effects, we need si-5 multaneous measurements of these parameters and D 50 .

Impact of D crit on CCN number concentration
The main purpose of the calculation of D crit is the precise estimation of CCN number concentration from the particle size distribution. We estimated the difference in CCN number concentration due to the difference in D 50 and calculated D crit . The following 10 equation was used for the assessment of the difference: where N CCN is the number concentration of CCN measured by the CCN counter, ∆N CCN is the difference of the CCN number concentration caused by the error in the estimation of D crit , and f N (logD p ) is the number size distribution measured by the SMPS. Therefore, ∆N CCN /N CCN is the ratio of the difference of the CCN number concentration due to the difference in D crit and the observed CCN number concentration. ∆N CCN /N CCN depends on the number size distribution (f N (logD p )) and the uncertainty of the chemical composition (D crit -D 50 ). For this calculation, the calculated D crit (adipic acid assumption, surface tension of water) was used. In addition to the adipic acid as-20 sumption, we also calculated ∆N CCN /N CCN using the ammonium sulfate assumption (D crit =D crit of ammonium sulfate) because this assumption has frequently been used in previous CCN studies (e.g., VanReken et al., 2003), including those for Gosan (Yum et al., 2007).

EGU
The results are summarized in Fig. 13. In general, the adipic acid approximation underestimates the CCN number concentration due to the overestimation of D crit , whereas the ammonium sulfate approximation overestimates the CCN number concentration because of the underestimation of the D crit . At SS=0.097%, ∆N CCN /N CCN was −0.27±0.14 and 0.16±0.18 for the adipic acid and ammonium sulfate approximations, 5 respectively. These values give a measure of the uncertainty associated with the CCN prediction based on the results of this study. The absolute value of ∆N CCN /N CCN increased with decreasing SS. During the observation period, the average values of B did not show a significant dependence on the value of SS as discussed in the Sect. 4.2.2, and the magnitude of error associated with the estimation of D crit does not depend on 10 SS significantly. Thus, this trend was not mainly due to the size dependence of the chemical composition. D crit decreased with increasing SS. The particle number concentration between D 50 and D crit was smaller at higher SS in comparison with the CCN number concentration because at higher SS, a larger fraction of CCN is in the size range larger than D 50 , as can be seen from Fig. 6. These results show that the effect 15 of chemical composition on N CCN was more important at lower SS, and the aerosol number size distribution was important for higher SS.

Summary
We measured the CCN number concentration (N CCN ) and the size-resolved CCN/CN ratios at SS=0. 097, 0.27, 0.58, and 0 EGU particles and those of ammonium sulfate particles were almost identical for SS=0.97, 0.58 and 0.27%, indicating that these particles were mainly composed of ammonium sulfate. The peak diameter of the CCN size distribution averaged for the whole observational period was about 150 nm at SS=0.097% and shifted to about 80 nm at SS=0.97%.

5
The temporal variation of D 50 at SS=0.097% was negatively correlated with the variation of the water-soluble fraction (inorganics + WSOC) of the aerosol components. For quantitative comparison, the threshold diameters for CCN activation (D crit ) were calculated by Köhler theory assuming the surface tension of water and PM 2.5 aerosol chemical composition. The calculated D crit values were correlated with D 50 at SS=0.097% 10 (r 2 =0.48). However, D crit was systematically larger than D 50 by about 16-29%. Sensitivity studies have shown that this discrepancy can be explained by possible differences in aerosol chemical composition between sub-micron and super-micron size ranges. In addition, a decrease of the surface tension due to the existence of WSOC can also significantly decrease D crit . 15 The particle number concentrations in the size range between D 50 and D crit (∆N CCN ) were calculated using the observed size distribution. The ratios of∆N CCN to N CCN (∆N CCN /N CCN ) were estimated to be −0.27±0.14 and −0.10±0.13 at SS=0.097% and 0.97%, respectively, assuming that water-soluble organic compounds are represented by adipic acid. The∆N CCN /N CCN ratio gives a measure of the uncertainty in estimat-20 ing CCN number concentrations using particle number size distributions and PM 2.5 chemical compositions in the East Asia region. Introduction