Aerosol Hygroscopicity and Cloud Condensation Nuclei Activity during the Ac Exp Campaign: Implications for Cloud Condensation Nuclei Parameterization F. Zhang Et Al.: Aerosol Hygroscopicity and Cloud Condensation Nuclei in the Ac 3 Exp Campaign

Aerosol hygroscopicity and cloud condensation nuclei (CCN) activity under background conditions and during pollution events are investigated during the Aerosol-CCN-Cloud Closure Experiment (AC 3 Exp) campaign conducted at Xianghe, China in summer 2013. A gradual increase in size-resolved activation ratio (AR) with particle diameter (D p) suggests that aerosol particles have different hygroscopicities. During pollution events, the activation diameter (D a) measured at low supersaturation (SS) was significantly increased compared to background conditions. An increase was not observed when SS was > 0.4 %. The hy-groscopicity parameter (κ) was ∼ 0.31–0.38 for particles in accumulation mode under background conditions. This range in magnitude of κ was ∼ 20 %, higher than κ derived under polluted conditions. For particles in nucleation or Aitken mode, κ ranged from 0.20–0.34 for background and polluted cases. Larger particles were on average more hygroscopic than smaller particles. The situation was more complex for heavy pollution particles because of the diversity in particle composition and mixing state. A non-parallel observation CCN closure test showed that uncertainties in CCN number concentration estimates ranged from 30–40 %, which are associated with changes in particle composition as well as measurement uncertainties associated with bulk and size-resolved CCN methods. A case study showed that bulk CCN activation ratios increased as total condensation nuclei (CN) number concentrations (N CN) increased on background days. The background case also showed that bulk AR correlated well with the hygroscopicity parameter calculated from chemical volume fractions. On the contrary, bulk AR decreased with increasing total N CN during pollution events, but was closely related to the fraction of the total organic mass signal at m/z 44 (f 44), which is usually associated with the particle's organic oxidation level. Our study highlights the importance of chemical composition in determining particle activation properties and underlines the significance of long-term observations of CCN under different atmospheric environments, especially regions with heavy pollution.


Introduction
The indirect influence of aerosol particles on the radiative balance of the atmosphere through changes in cloud droplet number and the persistence of clouds (Twomey, 1974;Albrecht, 1989) carries the largest uncertainty amongst the presently known causes of radiative forcing (IPCC, 2013).Thus, better understanding of aerosol particle formation, growth, and activation is essential.
Field and laboratory experiments have been conducted with the aim of better characterizing the particle physical and chemical parameters impacting on cloud condensation nuclei (CCN) activation.Studies have addressed the relative importance of the size distribution, particle composition, and mixing state in determining CCN activation, but there are disagreements on the relative importance of these parameters (e.g., Roberts et al., 2002;Feingold, 2003;Ervens et al., 2005;Mircea et al., 2005;Dusek et al., 2006a;Anttila and Kerminen, 2007;Hudson, 2007;Quinn et al., 2008;Zhang et al., 2008;Deng et al., 2013, Ma et al., 2013).CCN closure studies are a useful approach to test our knowledge of the controlling physical and chemical factors and to help verify experimental results.The CCN number concentration (N CCN ) is usually predicted from measured aerosol properties such as particle number size distribution and composition or hygroscopicity based on Köhler theory.The closure between the measured and estimated N CCN is often achieved under background atmospheric conditions without heavy pollution (Chuang et al., 2000;Dusek et al., 2003;VanReken et al., 2003;Rissler et al., 2004;Gasparini et al., 2006;Stroud et al., 2007;Bougiatioti et al., 2009).
In urban and polluted areas, the particle size distribution is more complex (Lee et al., 2003;Alfarra et al., 2004;Zhang et al., 2004;Salcedo et al., 2006).Particle activation is affected by the composition and the mixing state of aerosol particles.It has been demonstrated that particles are more difficult to activate during biomass burning plume events (Mircea et al., 2005;Lee et al., 2006;Clarke et al., 2007;Rose et al., 2010Rose et al., , 2011;;Paramonov et al., 2013;Lathem et al., 2013).Also, their activation ratios are reduced by secondary organics formed from oxidation of common biogenic emissions (VanReken et al., 2005;Varutbangkul et al., 2006;Mei et al., 2013) and black carbon (Dusek et al., 2006b;Kuwata et al., 2007).Other organic components (e.g., organic acids) have been shown to activate more easily (Raymond and Pandis, 2002;Hartz et al., 2006;Bougiatioti et al., 2011), but still much less than inorganic species.Therefore, testing the understanding of parameters controlling CCN activation and growth is challenging in heavily polluted areas.Furthermore, the main uncertainty in predicting the magnitude of global aerosol indirect effects arises from those regions under the influence of urban emissions (Sotiropoulou et al., 2007).The study of aerosol-CCN closure and relationships within, and in the outflow of, heavily polluted areas is thus important.
Field measurements of CCN have been made in East Asia, where megacities are likely to be major sources of pollutants and CCN (Yum et al., 2007;Rose et al., 2010Rose et al., , 2011;;Gunthe et al., 2009;Yue et al., 2011;Liu et al., 2011;Zhang et al., 2012;Deng et al., 2011;Leng et al., 2013).Despite the significant accomplishments of these studies, limitations and uncertainties exist.As a recent example over the region of interest, Deng et al. (2011) over-predicted concentrations of CCN at a site in the North China Plain by 19 % compared with direct measurements.
The aim of this paper is to study aerosol hygroscopicity and CCN activity under high aerosol loading conditions and to parameterize CCN number concentrations by using CCN activation ratios (AR) as a proxy of the total number of aerosol particles in the atmosphere.A cumulative Gaussian distribution function (CDF) fit model is applied to aerosol data collected under clean and polluted conditions to examine the influence of size distribution, heterogeneity of chemical composition, and mixing state on CCN activity.The hygroscopicity parameter (κ) is derived using Köhler theory to study aerosol hygroscopicity on clean days and during pollution events.In the CCN closure study, in addition to the parallel observation (PO) closure test, we apply the CCN efficiency spectrum to non-simultaneous condensation nuclei (CN) and bulk CCN observations, namely the non-parallel observation (NPO) closure test, to estimate N CCN .Finally, the relationship between bulk AR and aerosol physical and chemical properties is examined.

Measurements and data
An intensive field campaign called the Aerosol-CCN-Cloud Closure Experiment (AC 3 Exp) was conducted during June and July of 2013 at the Xianghe Atmospheric Observatory (39.798 • N, 116.958 • E; 35 m a.s.l.(above sea level)) located about 60 km southeast of the Beijing metropolitan area.This site is surrounded by agricultural land, densely occupied residences, and light industry.Set between two megacities (Beijing to the northwest and Tianjin to the southeast) and less than 5 km west of the local town center (with a population of 50 000), the site experiences frequent pollution plumes.Depending on the wind direction, instruments at the Xianghe site detect pollutants of urban, rural, or mixed origins, experiencing both fresh biomass burning emissions and advected aged aerosols.Details about the measurement location and meteorological conditions at the site can be found in Li et al. (2007bLi et al. ( , 2011)).

Instruments and measurements
Bulk CCN activation was measured from 1 to 25 June 2013.Size-resolved CCN were measured from 7 to 21 July 2013.Aerosol particle size distributions (10-700 nm) were measured from 1 to 25 June 2013 and from 7 to 21 July 2013.During 1-25 June 2013, a scanning mobility particle sizer (SMPS) was used independently for size distribution measurements.From 7 to 21 July 2013, it was combined with a Droplet Measurement Technologies-Cloud Condensation Nuclei Counter (DMT-CCNc) (Lance et al., 2006) and used for size-resolved CCN measurements.The CCN efficiency spectrum was derived from size-resolved CCN observations made from 7 to 21 July 2013.The aerosol particle size distribution data independently measured by the SMPS and bulk CCN measurements from 1 to 25 June 2013 combined with the derived CCN efficiency spectrum (Fig. 1) is used for the NPO CCN closure test.Aerosol chemical composition was measured from 31 May to 30 June 2013.
The aerosol inlet for the size distribution measurements was equipped with a TSI Environmental Sampling System (Model 3031200), which consists of a standard PM 10 inlet, a sharp-cut PM 1 cyclone, and a bundled Nafion dryer.After drying, the sample flow with relative humidity (RH) of < 30 % was drawn into the SMPS for the aerosol size distribution measurements (10-700 nm).Meanwhile, the bulk N CCN at different supersaturation (SS) was measured, using a continuous-flow CCN counter from the DMT-CCNc.Each bulk CCN measurement cycle included three SS levels: 0.23, 0.51 and 0.80 %.The scanning times for those SS levels were set at 7, 5, and 5 min, respectively.
The size-resolved CCN efficiency spectra were measured by coupling the same DMT-CCNc used with the SMPS (Rose et al., 2008).In this setup the particles are rapidly dried with RH < 30 % upon entering the Differential Mobility Analyzer (DMA).Thus, size selection is effectively performed under dry conditions, and the relative deviations in particle diameter should be < 1 % except for potential kinetic limitations (Mikhailov et al., 2009).The sample flow exiting the DMA was split into two parts, with 0.3 L min −1 for the CPC and 0.5 L min −1 for the CCN C .The DMA, controlled by the TSI-AIM software, scanned one size distribution every five minutes.The DMT-CCNc was operated at a total flow rate of 0.5 L min −1 with a sheath-to-aerosol flow ratio of 10.The inlet RH for DMT-CCNc was < 30 %.During the field campaign, the mean sample temperature and pressure measured by the DMT-CCNc sensors were 23.5 ± 1.6 • C and 985.5 ± 3.6 hPa.The SS levels of DMT-CCNc were calibrated with ammonium sulfate before and after the field campaign, following procedures outlined in Rose et al. (2008).During each CCN measurement cycle, calibrated effective SSs were 0.08, 0.11, 0.23, 0.42 and 0.80 %.The overall error (1σ ) for the SS levels was estimated to be < 3.5 %.The completion of a full measurement cycle took 60 min (20 min for SS = 0.08 % and 10 min for the other SS levels).
The measurement of non-refractory submicron aerosol species including organics, sulfate, nitrate, ammonium, and chloride were made with an Aerodyne Aerosol Chemical Speciation Monitor (ACSM) (Sun et al., 2012).The ACSM uses the same aerosol sampling, vaporization and ionization modules as the Aerosol Mass Spectrometer (AMS) (De-Carlo et al., 2006), but removes the size components.During the field campaign, ambient aerosols were drawn inside through a 1/2 in.(outer diameter) stainless steel tube at a flow rate of ∼ 3 L min −1 , of which ∼ 84 cc min −1 was subsampled into the ACSM.An URG cyclone (Model: URG-2000-30ED) was also supplied in front of the sampling inlet to remove coarse particles with a size cut-off of 2.5 mm.Before sampling into the ACSM, aerosol particles are dried by silica gel desiccant.The residence time in the sampling tube was ∼ 5 s.The ACSM was operated at a time resolution of ∼ 15 min with a scan rate of the mass spectrometer of 500 ms amu −1 from m/z 10 to 150.Regarding the calibration of the ACSM, mono-dispersed, size-selected 300 nm ammonium nitrate particles within a range of concentrations were sampled into both the ACSM and a condensation particle counter (CPC).The ionization efficiency (IE) was then determined by comparing the response factors of the ACSM to the mass calculated with known particle size and number concentrations from the CPC.Once the IE was determined, changes in the internal standard naphthalene or air ions, e.g., m/z 28 (N + 2 ) or m/z 32 (O + 2 ), were used to account for the degradation of the detector.Other details including the instrument, aerosol sampling setup, operations, and calibration are given in Sun et al. (2012) and Ng et al. (2011).

F. Zhang et al.: Aerosol hygroscopicity and cloud condensation nuclei in the AC 3 Exp campaign
In addition to the ACSM, the black carbon (BC) in PM 2.5 was simultaneously measured at a time resolution of 5 min by a BC analyzer (Aethalometer, Model AE22, Magee Scientific Corporation).The campaign averaged mass concentrations of BC were ∼ 4.2 µg m −3 , and the averaged mass fraction was about 6 %, with maximum of 18 % and minimum of 2 %.During the experiment period, the campaign area was generally hot and moist, with an average temperature of 23.6 • C and an average ambient RH of 72.3 %.

Data
The raw CCN data for both bulk and size-resolved CCN measurements were first filtered according to the instrument recorded parameters (e.g., temperature and flow).For example, if the relative difference between the actual and set sample flows was larger than 4 %, the data were flagged as invalid.If the "temperature stability" was recorded as "0", the data was also flagged as invalid data due to instrument fluctuations.These flagged data were not used for further analysis.A multiple charge correction and transfer function (Deng et al., 2011) was applied to each CN size distribution spectrum as well as to the CCN efficiency spectrum.The CCN AR is the ratio of N CCN / N CN .To examine CCN activity under different conditions, the size-resolved CCN efficiency data were classified as polluted or as background based on the aerosol loading as well as the synchronous surface horizontal wind data.Polluted conditions were identified when the total CN number concentration (N CN ) was > 15 000 cm −3 and when the airflow came from the southeast or east.Background cases were identified when N CN was < 15 000 cm −3 and when winds were from the west or northwest.N CN is the total aerosol number concentration with a particle size range of 10-700 nm.Here, the background refers to a regional background condition which represents a well-mixed atmosphere unaffected by local emissions, like biomass burning.Bulk measurements of total CCN number concentrations at SS levels of 0.23, 0.51 and 0.80 % could lead to a considerable underestimation of N CCN under polluted conditions (Deng et al., 2011) due to water depletion inside the column (Lathem and Nenes, 2011).Therefore, in this study, data points with N CN > 25 000 cm −3 were excluded.In the closure study, CCN size distributions were calculated by multiplying the fitted campaign-averaged CCN efficiency spectrum (a 3parameter CDF fit) with the aerosol particle number size distribution.The total N CCN was then obtained by integrating the size-resolved N CCN over the whole size range.Aerosol mass concentrations were processed using ACSM standard data analysis software (version 1.5.1.1).Detailed procedures for the data analysis have been described in Ng et al. (2011) and Sun et al. (2012).

Theory
As proposed by Petters and Kreidenweis (2007), κ can be used to describe the ability of particles to absorb water vapor and act as CCN.Based on Köhler theory (Köhler, 1936), κ relates the dry diameter of aerosol particles to the critical water vapor SS.According to measurements and thermodynamic models, κ is zero for insoluble materials like soot or mineral dust.However, their hygroscopicity changes due to the aging process, so the κ value then is > 0. The magnitude of κ is ∼ 0.1 for secondary organic aerosols, ∼ 0.6 for ammonium sulfate and nitrate, 0.95-1 for sea salt (Niedermeier et al., 2008), and 1.28 for sodium chloride aerosols.The effective hygroscopicity of mixed aerosols can be approximated by a linear combination of the κ values of the individual chemical components weighted by the volume or mass fractions (Kreidenweis et al., 2008;Gunthe et al., 2009).In this study, we calculated κ based on both size-resolved CCN measurements and bulk chemical composition observations made during the field campaign.The method to derive κ is described below.

Derivation of the average hygroscopicity parameter
The magnitude of κ was derived from the measured sizeresolved CCN activated fraction using κ-Köhler theory (Petters and Kreidenweis, 2007).In κ-Köhler theory, the water vapor saturation ratio over an aqueous solution droplet, S c , is given by: where D is the droplet diameter, D p is the dry diameter of the particle, M w is the molecular weight of water, σ w is the surface tension of pure water, ρ w is the density of water, R is the gas constant, and T is the absolute temperature.When κ is greater than 0.1, it can be conveniently expressed as: where S c is the particle critical supersaturation and is derived using the approach described by Rose et al. (2008), and A is defined as: The characteristic S c of the size selected CCN is represented by the SS level at which AR reaches 50 %.For the parameters listed above, T = 298.15K, R = 8.315 J K −1 mol −1 , ρ w = 997.1 kg m −3 , M w = 0.018015 kg mol −1 , and σ w = 0.072 J m −2 .Note that values derived from CCN measurement data through Köhler model calculations assume that the surface tension of pure water must be regarded as an "effective hygroscopicity parameter" accounting not only for the reduction of water activity by the solute ("effective Raoult parameter") but also for surface tension effects (Petters and Kreidenweis, 2007).In this study, a parameter called κ a , which characterizes the average hygroscopicity of CCN-active particles in the size range around activated diameters (D a ), is calculated from the data pairs of SS and D a based on the κ-Köhler theory.

Derivation of the particle hygroscopicity
For a given internal mixture, κ can be predicted by a simple mixing rule on the basis of chemical volume fractions, ε i (Petters and Kreidenweis, 2007;Gunthe et al., 2009): where κ i and ε i are the hygroscopicity parameter and volume fraction for the individual (dry) components in the mixture and i is the number of components in the mixture.We derive ε i from the particle chemical composition measured by the ACSM.Measurements from the ACSM show that the composition of submicron particles was dominated by organics, followed by nitrate, ammonium, and sulfate.The contribution of chloride was negligible (with a volume fraction of about < 2 %).The analysis of the anion and cation balance suggests that anionic species (NO − 3 , SO 2− 4 ) were essentially neutralized by NH + 4 over the relevant size range.For refractory species, BC represented a negligible fraction of the total submicron aerosol volume (less than 3 %).Sea salt and dust are usually coarse mode particles with particle sizes > 1 µm (Whitby, 1978).The contribution of such types of aerosols is thus expected to be negligible for sizes < 1000 nm.Therefore, the submicron particles measured by the ACSM mainly consisted of organics, (NH 4 ) 2 SO 4 , and NH 4 NO 3 .The particle hygroscopicity is thus the volume average of the three participating species: (5) The value of κ (NH 4 ) 2 SO 4 is 0.67 and κ NH 4 NO 3 is 0.61, derived from previous laboratory experiments (Petters and Kreidenweis, 2007).The following linear function derived by Mei et al. (2013) was used to estimate κ Org in our study: κ Org = 2.10 × f 44 − 0.11.The mean value of κ Org during the field campaign wasps 0.115 ± 0.019.Species volume fractions were derived from mass concentrations and densities of participating species.The densities of (NH 4 ) 2 SO 4 and NH 4 NO 3 are 1770 kg m −3 and 1720 kg m −3 , respectively.The density of organics is 1200 kg m −3 (Turpin et al., 2001).

Cumulative Gaussian distribution function fit and parameters derived from the cloud condensation nuclei efficiency
The spectra of measured CCN efficiency under both polluted and background conditions were fitted with a CDF (Rose et al., 2008): where the maximum activated fraction (MAF) is equal to 2a; D a is the midpoint activation diameter; and σ a is the CDF standard deviation.These parameters were determined for each spectrum.If MAF = 1 by changing the parameter "a" to 0.5, the spectrum is characteristic for internally mixed aerosols with homogeneous composition and hygroscopicity.The 3-parameter fit results represent the average activation properties of the aerosol particle fraction.During the field campaign, about 1200 size-resolved CCN efficiency spectra for atmospheric aerosols at SS levels of 0.08 to 0.80 % were measured.Figure 1 shows campaign-averaged spectra of both measured and fitted CCN efficiency at SS levels of 0.08, 0.11, 0.23, 0.42, and 0.80 % for background and polluted conditions.The slope of AR with respect to diameters near D a in Fig. 1 provides information about the heterogeneity of the composition for the size-resolved particles.
For an ideal case when all CCN-active particles have the same composition and size, a steep change of AR from 0 to MAF would be observed as D reaches D a .A gradual increase in size-resolved AR with particle diameter suggests that aerosol particles consisted of different hygroscopicities.
The gentler slopes of AR around D a during pollution events show that the particle composition was more heterogeneous than the particle composition under background conditions.Significant differences in size-resolved CCN efficiency spectra under polluted and clean conditions at lower SS levels have been derived.The different shapes suggest that aerosol hygroscopicity and CCN activity would be affected by local emission sources, e.g., biomass burning.

Activation diameter
The three parameters (MAF, D a , and σ ), describing the CCN efficiency spectra derived from the 3-parameter CDF fits, and κ a under polluted and clean conditions, are summarized in Table 1.Activation diameters under polluted and clean conditions are denoted as D a _POL and D a _BG in Table 1, respectively.As expected, D a decreases with increasing SS under both background and polluted conditions.At a given SS, D a _POL is greater than D a _BG, suggesting that particles under polluted conditions would be activated at a larger diameter.As SS increases, the difference between D a _POL and D a _BG decreases.

Maximum activated fraction
In general, aerosols with a more uniform and homogenous chemical composition or with a core-shell structure would have a higher MAF.The MAF under polluted and background conditions are denoted by MAF_POL and MAF_BG in Table 1.Values of MAF_BG and MAF_POL range from 0.95-0.98 and from 0.94-0.98,respectively.No significant discrepancies in MAF are observed between polluted and background conditions Observations show that particles can activate to CCN completely when particle diameters are greater than 300 nm even at SS = 0.08 %.This suggests that a smaller portion of 1− MAF (2-6 %) is caused by the error in the CDF fit, which will lead to a lower MAF than expected.

Cumulative Gaussian distribution function standard deviations
CDF standard deviations (σ ) are general indicators for the extent of external mixing and the heterogeneity of particle composition for aerosols in the size range around D a .CDF σ under polluted and background conditions are denoted as σ _POL and σ _BG in Table 1, respectively.Under ideal conditions, σ equals zero for an internally mixed, fully monodispersed aerosol with particles of homogeneous chemical composition.According to Rose et al. (2008), even after correcting for the DMA transfer function, calibration aerosols composed of high-purity ammonium sulfate exhibit small non-zero σ values that correspond to ∼ 3 % of D a .This can be attributed to heterogeneities of the water vapor SS profile in the CCNc or other non-idealities, such as DMA transfer function and particle shape effects.Thus, "heterogeneity parameter" values of σ/D a = 3 % indicate internally mixed CCN, whereas higher values indicate external mixtures of particles.Heterogeneity parameters under polluted and background conditions are denoted as σ/D a _POL and σ/D a _BG in Table 1, respectively.They range from 17-30 %, which is much higher than the 3 % observed for pure ammonium sulfate, indicating that the particles were externally mixed with respect to their solute content.

Derived average hygroscopicity parameter dependence on activation diameter
Figure 2 shows the dependence of κ a on D a under background and polluted conditions.κ a _POL and κ a _BG are defined as the average hygroscopicity of CCN-active particles in the size range around D a under polluted and background conditions, respectively.For background days, larger particles were on average more hygroscopic than smaller particles, i.e., κ a _BG increases substantially from about 0.22 at 30-60 nm to about 0.38 at a size range of 120-180 nm.This is consistent with field results observed in Guangzhou, South China by Rose et al. (2010).However, compared to κ a _BG, κ a _POL shows a relatively flat trend as the particle size diameter increases and error bars are larger.This suggests that under polluted conditions, particle composition and their mixing state is complex and diverse.In this case, larger particles are even less hygroscopic than smaller particles.One possible reason for changes in κ a under polluted conditions may be the presence of a high amount of organics freshly emitted from biomass burning (Andreae and Rosenfeld, 2008;Petters et al., 2009;Rose et al., 2010) which would coat the larger particles and render them less hygroscopic.Overall, κ for polluted aerosols is about 20 % lower than that for clean aerosol particles in the accumulation size range.For particles in the nucleation or Aitken size range, κ a for polluted particles is slightly higher than that for particles in the background cases.Based on laboratory experiments, Petters et al. (2009) examined the hygroscopic properties of particles freshly emitted from biomass burning.They found that κ was a function of particle size, with 250 nm particles being generally weakly hygroscopic and sub-100 nm particles being more hygroscopic.During the field campaign, polluted cases occurred when particles were mainly biomass burning aerosols.The laboratory results, to some extent, can thus explain our field measurements.Further investigations, including laboratory experiments and field measurements of sizeresolved chemical composition, are needed to confirm and clarify this.extend to large particle sizes indicating the impact by pollution.The largest variation in D a _BG and D a _POL is seen at SS = 0.08 % and 0.11 %, respectively.One reason for this is the weakened impact of chemical composition on CCN activity at high SS levels, i.e., the solute effect.The other reason is the larger uncertainties that arise from making measurements at low SS levels.

Probability distribution functions of activation diameter and average hygroscopicity parameter
Figure 4 shows PDFs of κ a under background conditions and during pollution events.The PDF of κ a _POL has a large variation at each size range around D p and shows two modes.For example, κ a _POL for particles around 48 nm shows two peaks at ∼ 0.15 and ∼ 0.23.Peak values of ∼ 0.26 and ∼ 0.32 are seen for particles around D p = 198 nm.Most κ a _POL values are less than 0.3.This suggests externally mixed, but less hygroscopic particles are prevalent during pollution events.Less variation is seen in the PDFs of κ a _BG.One mode is seen with peak values of 0.23, 0.30, 0.35, 0.35, and 0.38 for D p = 46, 64, 92, 152, and 179 nm, respectively.Most κ a _BG values are greater than 0.3 when D p > 60 nm, indicating that the particles are more hygroscopic with a relatively homogeneous composition.

Cloud condensation nuclei closure tests
In this section, we compare observed total N CCN with corresponding values that were estimated on the basis of aerosol particle number size distributions measured both in parallel and not in parallel and assuming a uniform particle composition.Closure was achieved when estimated and observed N CCN agreed quantitatively within the range of their uncertainties.

Parallel observation closure test
In parallel observation (PO) closure tests, the measured CCN efficiency spectrum is first multiplied by the measured CN size distribution, which yields the CCN size distribution.This is then integrated over the whole size range to obtain the observed total CCN concentration (CCN_Obs).Sizeresolved N CCN are calculated by multiplying the campaignaveraged CCN efficiency spectrum with simultaneously measured CN number size distributions.Estimated total N CCN (CCN_Estimated) are obtained by the stepwise integration of size-resolved N CCN from 10 to 700 nm.With this comparison, the influence of the variation in chemical composition on the CCN concentration can be investigated because the CN size distribution is the same for both parameters.
Figure 5 shows CCN_Estimated as a function of CCN_Obs at SS levels ranging from 0.08 to 0.80 %.Estimated and measured total N CCN agree well.The mean slope and correlation coefficient (R 2 ) are 0.99 and 0.97, respectively, at the five SS levels.A ∼ 3-4 % underestimation occurs at SS levels of 0.08 and 0.11 %.One reason for this slight underestimation is that size-resolved ARs are more variable at low SS levels than at higher SS levels.Also, compared to total activated CCN number concentrations at high SS levels, there are fewer N CCN at low SS levels, leading to larger uncertainties or to a lower correlation.Overall, CCN closure can be achieved by using campaign-averaged CCN efficiency spectra.In this PO closure test, the influence of variations in chemical composition on CCN concentrations is insignificant.

Non-parallel observation closure test
Mean CCN efficiency spectra derived from size-resolved CCN measurements taken on 7-21 July 2013 are used to estimate total CCN number concentrations based on CN size distribution measurements taken on 1-25 June 2013.This is referred to as a non-parallel observation (NPO) closure test.The average measured CCN efficiency spectrum (cor- responding to spectra in Fig. 1) is multiplied by the measured CN size distribution which yields the CCN size distribution.This is integrated over the whole size range (10-700 nm) to obtain the estimated total CCN concentration.The mean CCN efficiency spectra at SS levels of 0.23 and 0.80 % (Fig. 1) is used to estimate the total CCN concentration during 1-25 June 2013.The mean CCN efficiency spectra at SS = 0.51 % is derived using the exponential relationships developed from plotting the three CDF fit parameters as a function of SS (see Fig. 7).
Estimated total N CCN as a function of measured bulk N CCN at SS levels of 0.23, 0.51, and 0.80 % are shown in Fig. 6.The lower slope at SS = 0.23 % indicates that the estimation on the basis of NPO closure underestimates about 7 % of the observed N CCN .The closure is considerably improved at higher SS levels.A reasonable correlation between estimated and measured total N CCN is seen (R 2 = 0.6 − 0.8), which suggests temporal variations in chemical composition and mixing state of aerosol particles.In addition, there are uncertainties due to measuring bulk and size-resolved CCN.Overall, uncertainties in this NPO CCN closure study range from 30-40 %.Caution is needed when using data from any short-term experiment at a single site to do CCN parameterizations for large-scale applications.It is necessary to conduct long-term CCN measurements at more regional sites, especially those that are heavily polluted.

Case study: cloud condensation nuclei activation and chemical composition
The behavior of CCN activation under background and polluted conditions is examined.Two cases are selected: one case with total N CN < 15 000 cm −3 (background) and another case with total N CN > 15 000 cm −3 (polluted).Trends in bulk CCN activation as N CN increases are different for the background and polluted cases.Bulk AR at the three SS levels (0.23, 0.51, and 0.80 %) increases as total N CN increases for background cases (Fig. 8a) and decreases as total N CN increases for polluted cases (Fig. 8b).For the background cases, changes in bulk AR are dependent on changes in κ chem (Fig. 8c).A good correlation between AR_ 0.23 and κ chem (R 2 > 0.7) is seen in Fig. 9.A high correlation between bulk AR and κ chem , when total N CN is low, is observed during the campaign (Fig. 10).In these cases, organics account for ∼ 30 % of the total particle mass concentration and concentrations of soluble inorganics are high (Fig. 8e).In particular, the mass concentration of nitrate is higher than that for organics and accounts for the largest mass fraction when κ chem reaches a maximum with a mean value of ∼ 0.45.The f 44 , which is the fraction of total organic mass signal at m/z 44, is not correlated with AR (Fig. 8c).The m/z 44 signal is mostly due to acids (Takegawa et al., 2007;Duplissy et al., 2011) or acid-derived species, such as esters, and f 44 is closely related to the organic oxidation level, i.e., O : C ratio (Aiken et al., 2008).Oxidized/aged acids are generally more hygroscopic and easily activated.Therefore, the lower correlation between f 44 and AR implies that organics under low N CN conditions are less hygroscopic.CN number concentrations in the nucleation, Aitken, and accumulation modes are shown in Fig. 8g (polluted) and Fig. 8h (background).Under background conditions, bulk AR at SS = 0.23 % is more correlated (R 2 = 0.5) with changes in N CN in the accumulation mode (Fig. 9), suggesting that most aerosol particles with sizes > 100 nm can be activated.Smaller particles with Aitken diameters of < 40 nm at the given SS levels (0.23-0.80 %) are not as easily activated, if at all, so no correlation is seen (Fig. 8g).
Under polluted conditions, there is little dependence of changes in bulk AR with changes in κ chem (Fig. 8d).Bulk AR at SS = 0.23 % is moderately correlated with f 44 (R 2 = 0.5, Fig. 9).As stated above, f 44 is always related to the organic oxidation level.Usually, oxidized/aged acids are more hygroscopic and easily activated.The correlation betweenf 44 and bulk AR suggests that the organics contribution from oxidized or aged aerosols play a significant role in CCN activity (Jimenez et al., 2009).A bias is introduced by using a parameterized function derived from observations made at other sites with different aerosol types to describe the particle hygroscopicity and activation properties due to the complexity of the organic aerosol fraction and its tendency to evolve with atmospheric oxidative processing and aerosol aging (e.g., Padró et al., 2010;Engelhart et al., 2011Engelhart et al., , 2012;;Asa-Awuku et al., 2011).Under polluted conditions, the bulk AR_0.2 is  more correlated with changes in accumulation mode particles (R 2 = ∼ 0.3).Overall, based on the case study, one cannot use a parameterized formula using only total N CN to estimate total CCN number concentrations.If observations such as size-resolved CCN and size-resolved chemical composition are not available, the possibility of using bulk κ chem and f 44 in combination with bulk N CN > 100 nm to parameterize CCN number concentrations is implied by the case study.Further field investigations are needed to demonstrate and confirm the relationship between bulk AR and particle size and composition.

Summary and conclusions
Atmospheric aerosol particles acting as CCN are pivotal elements of the hydrological cycle and climate change.In this study, we measured and characterized N CCN in relatively clean and polluted air during the AC 3 Exp campaign conducted at Xianghe, China during summer 2013.The aim was to examine CCN activation properties under high aerosol loading conditions in a polluted region and to assess the impacts of particle size and chemical composition on the CCN AR, which acts as a proxy of the total number of aerosol particles in the atmosphere.Based on the CDF fit method, a gradual increase in size-resolved AR with particle diameter suggests that aerosol particles have different hygroscopici- ties.D a increased significantly at lower SS levels (< 0.23 %) due to pollution, e.g., biomass burnings.This increase was not observed when SS > 0.4 %.For particles in the accumulation mode, values of κ ranged from 0.31-0.38under background conditions, about 20 % higher than that derived under polluted conditions.For particles in the nucleation or Aitken mode, κ ranged from 0.20-0.34under both background and polluted conditions.Larger particles were on average more hygroscopic than smaller particles.However, the case of particles originating from heavy pollution is more complex, due to the diversity in particle composition and mixing state.The low R 2 for the NPO CCN closure test suggests a 30-40 % uncertainty in total N CCN estimation.Using bulk chemical composition data from ACSM measurements, the relationship between bulk AR and the physical and chemical properties of atmospheric aerosols is investigated.Based on a case study, we conclude that one cannot use a parameterized formula using only total N CN to estimate total N CCN .We have shown the possibility of using bulk κ chem and f 44 in combination with bulk N CN > 100 nm to parameterize CCN number concentrations.Further field investigations or examinations are needed to demonstrate and confirm the relationship between bulk AR and particle size and composition.

Figure 2 .
Figure 2. Derived hygroscopicity parameter, κ a , as a function of particle diameter, D a , under polluted (POL) and background (BG) conditions.Bottom: percent change in κ a due to pollution as a function of D a .Error bars represent one standard deviation calculated over the entire measurement period.

Figure 3 .
Figure 3. Probability distribution functions of D a under background (_BG, left panels) and polluted (_POL, right panels) conditions at five SS levels (0.08-0.80 %) during the size-resolved CCN measurement period.

Figure 3
Figure3shows probability distribution functions (PDFs) for D a under background conditions and during pollution events.D a _POL are mainly distributed in the ranges of about185- 205, 163-180, 95-120, 65-75 and 45-55 nm at SS levels of 0.08, 0.11, 0.23, 0.42 and 0.80 %, respectively.At each SS level, the PDFs of D a _POL have a wider distribution than the PDFs of D a _BG.At each SS level, the PDFs of D a _POL

Figure 4 .
Figure 4. Probability distribution functions of κ a under background (_BG, left panels) and polluted (_POL, right panels) conditions for different particle size ranges during the size-resolved CCN measurement period.

Figure 5 .
Figure 5.Estimated N CCN as a function of observed N CCN at different SS levels in the parallel observation (PO) closure test.The green solid line is the 1:1 line.

Figure 6 .
Figure 6.As in Fig. 5, but for the non-parallel observation (NPO) closure test.The dashed green lines indicate the boundaries representing ± 30 % deviations of N CCN -estimated from N CCNobserved.

Figure 7 .
Figure 7. Activation diameter, D a , maximum activated fraction, MAF, and standard deviation, σ , as a function of super saturation, SS.

Figure 9 .
Figure 9. AR measured at SS = 0.23 % as a function of (a) κ chem under background conditions, (b) accumulation mode N CN under background conditions, (c) f 44 under polluted conditions, and (d) accumulation mode N CN under polluted conditions.The accumulation mode size range is 130-700 nm in this study.

Figure 10 .
Figure 10.Time series of bulk AR at SS = 0.23 %, derived κ chem , and N CN from 19-24 June 2013.Green shaded areas highlight periods with high correlations between bulk AR and κ chem .