Cloud condensation nuclei ( CCN ) from fresh and aged air pollution in the megacity region of Beijing

Atmospheric aerosol particles serving as cloud condensation nuclei (CCN) are key elements of the hydrological cycle and climate. CCN properties were measured and characterized during the CAREBeijing-2006 campaign at a regional site south of the megacity of Beijing, China. Size-resolved CCN efficiency spectra recorded for a supersaturation range of S = 0.07 % to 0.86 % yielded average activation diameters in the range of 190 nm to 45 nm. The corresponding effective hygroscopicity parameters ( κ) exhibited a strong size dependence ranging from ∼0.25 in the Aitken size range to∼0.45 in the accumulation size range. The campaign average value ( κ = 0.3± 0.1) was similar to the values observed and modeled for other populated continental regions. The hygroscopicity parameters derived from the CCN measurements were consistent with chemical composition data recorded by an aerosol mass spectrometer (AMS) and thermo-optical measurements of apparent elemental and organic carbon (EC and OC). The CCN hygroscopicity and its size dependence could be parameterized as a function of only AMS based organic and inorganic mass fractions ( forg, finorg) using the simple mixing ruleκp ≈ 0.1 · forg + 0.7 · finorg. Correspondence to: S. S. Gunthe (s.gunthe@iitm.ac.in) When the measured air masses originated from the north and passed rapidly over the center of Beijing (fresh city pollution), the average particle hygroscopicity was reduced ( κ = 0.2± 0.1), which is consistent with enhanced mass fractions of organic compounds ( ∼50 %) and EC ( ∼30 %) in the fine particulate matter (PM1). Moreover, substantial fractions of externally mixed weakly CCN-active particles were observed at low supersaturation ( S = 0.07 %), which can be explained by the presence of freshly emitted soot particles with very low hygroscopicity ( κ < 0.1). Particles in stagnant air from the industrialized region south of Beijing (aged regional pollution) were on average larger and more hygroscopic, which is consistent with enhanced mass fractions ( ∼60 %) of soluble inorganic ions (mostly sulfate, ammonium, and nitrate). Accordingly, the number concentration of CCN in aged air from the megacity region was higher than in fresh city outflow ((2.5–9.9)× 103 cm−3 vs. (0.4–8.3)× 103 cm−3 for S = 0.07–0.86 %) although the total aerosol particle number concentration was lower (1.2 × 104 cm−3 vs. 2.3× 104 cm−3). A comparison with related studies suggests that the fresh outflow from Chinese urban centers generally may contain more, but smaller and less hygroscopic aerosol particles and thus fewer CCN than the aged outflow from megacity regions. Published by Copernicus Publications on behalf of the European Geosciences Union. 11024 S. S. Gunthe et al.: CCN in Beijing

When the measured air masses originated from the north and passed rapidly over the center of Beijing (fresh city pollution), the average particle hygroscopicity was reduced (κ = 0.2 ± 0.1), which is consistent with enhanced mass fractions of organic compounds (∼50 %) and EC (∼30 %) in the fine particulate matter (PM 1 ).Moreover, substantial fractions of externally mixed weakly CCN-active particles were observed at low supersaturation (S = 0.07 %), which can be explained by the presence of freshly emitted soot particles with very low hygroscopicity (κ < 0.1).Particles in stagnant air from the industrialized region south of Beijing (aged regional pollution) were on average larger and more hygroscopic, which is consistent with enhanced mass fractions (∼60 %) of soluble inorganic ions (mostly sulfate, ammonium, and nitrate).Accordingly, the number concentration of CCN in aged air from the megacity region was higher than in fresh city outflow ((2.5-9.9)× 10 3 cm −3 vs. (0.4-8.3) × 10 3 cm −3 for S = 0.07-0.86%) although the total aerosol particle number concentration was lower (1.2 × 10 4 cm −3 vs. 2.3 × 10 4 cm −3 ).A comparison with related studies suggests that the fresh outflow from Chinese urban centers generally may contain more, but smaller and less hygroscopic aerosol particles and thus fewer CCN than the aged outflow from megacity regions.

Introduction
Atmospheric aerosol particles that enable the condensation of water vapor and the formation of cloud droplets at a given level of water vapor supersaturation are called cloud condensation nuclei (CCN).These aerosol particles play an important role in the formation of clouds and precipitation, and thus influence atmospheric chemistry and physics, and the hydrological cycle and climate (Seinfeld and Pandis, 2006;Lohmann and Feichter, 2005).Elevated concentrations of CCN tend to increase the concentration and reduce the size of the droplets in a cloud.Apart from changing the optical properties and the radiative effects of clouds on climate, this may lead to the suppression of precipitation in shallow and short lived clouds and to a greater convective overturning and more precipitation in deep convective clouds (Rosenfeld et al., 2008).The response of cloud characteristics and precipitation processes to increasing anthropogenic aerosol concentrations, however, represents one of the largest uncertainties in the current understanding of climate change (Andreae et al., 2005;IPCC, 2007).To incorporate the effects of CCN in meteorological models at scales from large eddy simulations (LES) to global climate models (GCMs), knowledge of the spatial and temporal distribution of CCN in the atmosphere is essential (Huang et al., 2007;Andreae and Rosenfeld, 2008;Stevens and Feingold, 2009;Pöschl et al., 2010).
Beijing, the capital of China, is a megacity with a population of about 20 million people (http://www.stats.gov.cn/english/newsandcomingevents/t20110429 402722516.htm).The city is located in Northeastern China, it is surrounded by the Yanshan Mountains from the west to the northeast, and has heavily industrialized areas from the southwest to the east (Chen et al., 2007;Streets et al., 2007).Due to the rapid economic growth and the increase in motor traffic during the last decade, the pollution in Beijing is a complex mixture of local domestic, industrial, and traffic emissions and there is regional pollution originating from the highly industrialized areas to the south (i.e.Hebei province and Tianjin Municipality) (Xia et al., 2007;Wehner et al., 2008).Episodically there is also influence of dust or biomass burning, but this was not observed during our campaign.The "Campaign of Air Quality Research in Beijing andSurrounding Region 2006" (CAREBeijing-2006) took place in summer 2006.Its objectives were to study the origin of the regional air pollution and to advise the local authorities on measures to reduce local particulate pollution during the Olympic Games to be held in Beijing in August 2008.A suburban measurement location, Yufa, located ∼50 km south of the center of Beijing, enabled investigations of the southerly inflow into the city to probe the impact of aged regional pollution on the city's air quality, as well as investigations of the city outflow when winds were northerly (fresh city pollution).
In this study we report on the first size-resolved measurements of CCN in the megacity region of Beijing, and we parameterize the effective aerosol particle hygroscopicity and CCN activity as a function of chemical composition determined by aerosol mass spectrometry (AMS).

Measurement location, aerosol inlets and supporting data
The measurements were performed during the CAREBeijing-2006 campaign (10 August to 9 September 2006) at a suburban site on the campus of Huang Pu University in Yufa (39.51467 • N, 116.30533 • E), which is located ∼50 km south of the city center of Beijing (Fig. 1).The instruments were located on the third and the top floor of a four story school building with the sample inlets and a meteorological station mounted on the roof (Achtert et al., 2009;Garland et al., 2009;Takegawa et al., 2009a,b;Wiedensohler et al., 2009).The average meteorological parameters (arithmetic mean ± standard deviation) recorded during the campaign period at the aerosol inlet were: 24.6 ± 4.4  et al. (2009).
The main aerosol inlet used in this study was equipped with a Rupprecht & Patashnick PM 10 inlet that was optimized for isokinetic flow conditions with a cyclone for the 10 µm size cut-off (flow rate 16.7 l min −1 ).The sample flow passed through stainless steel tubing (1.9 cm inner diameter, 5.1 m length) to a diffusion dryer with silica gel/molecular sieve cartridges (alternating regeneration with dry pressurized air, regeneration cycles 15-50 min, average RH = 29 ± 5 %; Tuch et al., 2009).After drying, the sample flow was split into separate lines.One line (stainless steel, 0.94 cm inner diameter, ∼7 m length, 1.5 l min −1 flow rate) was for the CCN measurement setup described below, and another one was used for the aerosol size distribution measurements with a Twin Differential Mobility Particle Sizer (TDMPS, 3-900 nm particle diameter; Birmili et al., 1999;Achtert et al., 2009;Wiedensohler et al., 2009).The aerosol mass spectrometer (AMS) measurements presented in this paper were performed using a separate inlet and sampling line as described by Takegawa et al. (2009a,b).
The mass concentrations of apparent elemental carbon (EC) (Andreae and Gelencsér, 2006) and organic carbon (OC) were measured using a Sunset Laboratory semicontinuous EC/OC analyzer operated by the University of Tokyo and Peking University (Kondo et al., 2006).The time resolution was 1 h, which included a sampling time of 40 min and an analysis time of 15 min.The sample line for the EC/OC analyzer was a stainless steel tube with an inner diameter of 10 mm and a length of ∼9 m.The ambient particles were collected on a quartz filter and then analyzed on the basis of the thermal-optical-transmittance method with a temperature protocol that was based on that proposed by the National Institute for Occupational Safety and Health (NIOSH).More details regarding EC/OC measurement techniques and results during CAREBeijing-2006 are discussed elsewhere (Takegawa et al., 2009a).The campaign average values of EC/OC data reported in this study include only data points for which also CCN data were available.

CCN measurements and data analysis
Size-resolved CCN efficiency spectra (activation curves) were measured with a Droplet Measurement Technologies continuous flow CCN counter (DMT-CCNC, Roberts and Nenes, 2005;Lance et al., 2006) coupled to a Differential Mobility Analyzer (DMA; TSI 3071; sheath flow 10 l min −1 ) and a condensation particle counter (CPC; TSI 3762; sample flow 1.0 l min −1 ; Frank et al., 2006;Rose et al., 2008).The CCNC was operated at a total flow rate of 0.5 l min −1 with a sheath-to-aerosol flow ratio of 10.
The effective water vapor supersaturation (S) was regulated by the temperature difference between the upper and lower end of the CCNC flow column ( T ) and calibrated with ammonium sulfate aerosol as described by Rose et al. (2008Rose et al. ( , 2010, Köhler model AP3; calibration line S = k S T + S 0 with k S = 0.0793 % K −1 , S 0 = −0.0909%, R 2 = 0.9083; relative uncertainty S/S < 10 %).Note that high CCN number concentrations in the CCNC column can significantly depress the supersaturation generated in the column (Lathem and Nenes, 2011).For our measurements, however, we do not expect this effect since the highest CCN counts in the column were ∼400 cm −3 .
For each CCN measurement cycle, T was set to 5 different levels (2.0-11.9K) corresponding to S values in the range of 0.07-0.86%.For each T and the respective S, the diameter of the dry aerosol particles selected by the DMA (D) was set to 9 different values in the range of 20-290 nm depending on the supersaturation selected.At each D, the number concentration of total aerosol particles (condensation nuclei, CN), N CN , was measured with the CPC, and the number concentration of CCN, N CCN , was measured with the CCNC.The integration time for each measurement data point was 30 s, the recording of a CCN efficiency spectrum (N CCN /N CN vs. D) took ∼16 min (including a 50 s adjustment time for each new particle size and 4 min for adjustment to the next supersaturation level), and the completion of a full measurement cycle comprising CCN efficiency spectra at 5 different supersaturation levels took ∼85 min (including 5 min of settling time for the changeover from highest to lowest S).
Approximately 300 measurement cycles, corresponding to ∼1500 CCN efficiency spectra, were performed during the CAREBeijing-2006 campaign, with occasional short term interruptions and one long term interruption (24 August to 1 September) for instrument calibration and maintenance.
The measurement data of the CCN efficiency spectra were corrected for the effects of multiply charged particles, differences in the CCNC and CPC counting efficiencies with a correction factor f corr (f corr = x 1 − x 2 • x D 3 ; x 1 = 0.9108, x 2 = 0.2489, and x 3 = 0.9746, with D being the diameter in nm and R 2 = 0.95), and the DMA transfer function as described in detail by Rose et al. (2010).For the multiple charge correction (Frank et al., 2006), we used the total aerosol particle number size distributions that were measured in parallel with CCN measurements by the TDMPS.For several CCN measurement cycles, TDMPS data were not available, and hence no charge correction was applied.However, the CCN data from these cycles remained comparable with the corrected ones, because the effect of the charge correction was generally small (< 5 % change in the activation diameters and other parameters used for further analysis; Rose et al., 2008Rose et al., , 2010)).The uncertainties of N CCN , N CN and N CCN /N CN were estimated to be < 20 % for individual measurement data points and < 10 % for average values and fit parameters (Rose et al., 2008(Rose et al., , 2010)).By fitting with a cumulative Gaussian distribution function (CDF), the following parameters were derived from each measured CCN efficiency spectrum (Gunthe et al., 2009;Rose et al., 2010): the maximum activated fraction MAF f , the midpoint activation diameter D a , and the standard deviation σ a of 3-parameter CDF fits, as well as the midpoint activation diameter D t and the standard deviation σ t of 2parameter CDF fits with MAF f set to 1.Note that the CCN efficiency measured at the largest diameter of each spectrum (MAF m = N CCN /N CN at D max ) was generally in good agreement with MAF f as derived from the 3-parameter CDF fit (deviations < 10 %).Key parameters derived from the measured CCN spectra and other related parameters used in the present study are listed and briefly explained in Table 1.
As detailed by Rose et al. (2010) the activation diameters and standard deviations derived from the 3-parameter and 2parameter CDF fits are not the same for CCN efficiency spectra with MAF f < 1: the 3-parameter fit results represent the average properties of the CCN-active aerosol particle fraction, whereas the 2-parameter fit results are proxies for the effective overall properties of the external mixture of CCNactive and CCN-inactive particles.In the remainder of this study we will focus on the 3-parameter fit results.
The difference between unity and the maximum observed CCN efficiency (1-MAF m or 1-MAF f ) represents the fraction of externally mixed CCN-inactive particles measured at D max or fitted over the whole diameter range of each scan, respectively.Though derived from a fit to the whole scan, MAF f is only valid at D max and the nearby diameter range with a plateau of activated particle fraction N CCN /N CN .Particles that are CCN-inactive at D max can be regarded as weakly CCN-active particles with hygroscopicity parameters below the threshold value κ c = κ(D max ,S) (Su et al., 2010).
The CDF standard deviations are general indicators for the mixing state and heterogeneity of particle composition in the investigated aerosol: σ a characterizes the CCN-active particles in the size range around D a , and σ t is a measure for the overall heterogeneity of CCN-active and -inactive particles in the size range around D t .Under ideal conditions, the CDF standard deviations should be zero for an internally mixed, fully monodisperse aerosol with particles of homogeneous chemical composition.Even after correcting for the DMA transfer function, however, calibration aerosols composed of high purity ammonium sulfate exhibit small non-zero σ a values that correspond to ∼3 % of D a and can be attributed to heterogeneities of the water vapor supersaturation profile in the CCNC or other non-idealities, such as DMA transfer function or particle shape effects.Thus, increasing values of the normalized CDF standard deviation or "heterogeneity parameter" indicate increasing heterogeneity of particle composition (Rose et al., 2010;Gunthe et al., 2009;Su et al., 2010).
For all data pairs of supersaturation and activation diameter derived from the CCN efficiency spectra measured in this study, effective hygroscopicity parameters κ were calculated using the κ-Köhler model equations and parameters specified in Rose et al. (2010) (surface tension 0.072 J m −2 , temperature 298 K).Note that the κ values derived from CCN measurement data through Köhler model calculations assuming the surface tension of pure water have to be regarded as "effective hygroscopicity parameters" that account not only for the reduction of water activity by the solute ("effective Raoult parameters") but also for surface tension effects (Petters and Kreidenweis, 2007;Mikhailov et al., 2009;Pöschl et al., 2009;Gunthe et al., 2009;Rose et al., 2010).
The parameter κ a calculated from the data pairs of S and D a characterizes the average hygroscopicity of CCN-active particles in the size range around D a .κ t calculated from D t is an approximate measure (proxy) for the effective hygroscopicity of mixtures of CCN-active and -inactive particles in the size range around D t (Rose et al., 2010).As discussed in earlier studies (Gunthe et al., 2009;Rose et al., 2010Rose et al., , 2011)), κ a is better suited for comparison with κ values predicted from AMS measurements, because κ a is not influenced by weakly CCN-active particles consisting mostly of insoluble and refractory materials like mineral dust and soot (or biopolymers that tend to char upon heating), which are also not (or less efficiently) detected by AMS.On the other hand, κ t is better suited for the single parameter prediction of CCN number concentrations when CCN-active particles are externally mixed with CCN-inactive particles (Gunthe et al., 2009;Rose et al., 2010Rose et al., , 2011)).
The statistical uncertainty in the determination of activation diameters by curve fitting (standard error of the CDF fit parameters) was on average ∼2-3 nm (∼1-5 %).According to the relative sensitivities specified by Kreidenweis et al. (2009), the uncertainty of 1-5 % in diameter corresponds to an uncertainty of 3-15 % in κ, and the uncertainty of < 10 % in supersaturation reported above corresponds to an uncertainty of < 20 % in κ (Su et al., 2010).The reported campaign average values of CCN parameters include all available data points unless mentioned otherwise.

AMS measurements and data analysis
An Aerodyne quadrupole aerosol mass spectrometer (Q-AMS, hereafter referred to as "AMS" for brevity) was used to measure the size-resolved chemical composition of nonrefractory (vaporized at ∼600 • C under vacuum) submicron aerosol particles (Jayne et al., 2000).Sampling information, data analysis technique, and performance of the AMS during the CAREBeijing-2006 campaign is described in detail by Takegawa et al. (2009a).The AMS data used in this study comprise a time series of mass size distributions (dM/dlogD va , ∼10 min time resolution) where D va is the vacuum aerodynamic diameter (DeCarlo et al., 2004).The mass size distributions were calculated for inorganic ions (SO 2− 4 , NH + 4 , NO − 3 , Cl − ) and organic matter (Org) in the size range between 45 nm and 1.4 µm.
The mass concentration data of every size bin were averaged with the two adjacent size bins to minimize the influence of noise.To make the size-resolved AMS results directly comparable with the CCN measurement results, all calculations and plots in this study using AMS data were based on approximate mobility equivalent diameters that have been calculated by division of the AMS vacuum aerodynamic diameter through a density scaling factor of 1.6.The scaling factor is based on the assumption of an effective particle density of 1.6 g cm −3 (Achtert et al., 2009;van Pinxteren et al., 2009).Accordingly, the scaled AMS mass size distributions span a mobility size range of 29 nm to 900 nm.
Size-resolved mass fractions were calculated diameterwise from the size distributions of mass concentration.The mass fraction of one chemical component is its mass concentration divided by the sum of the masses of all components at this diameter.To describe the chemical composition of the particles activated at the midpoint of the CCN efficiency spectrum, the AMS mass concentrations were integrated over the size interval of D a −σ a to D a +σ a (activation diameter ± CDF standard deviation).The mass concentrations within these intervals (m D ) were used to calculate mass fractions of inorganic and organic compounds (f inorg , f org ) for correlation analysis.For this purpose, the AMS data set was adjusted to the time resolution of the CCN data.The AMS data point corresponding to a CCN data point was chosen to be the one that is closest in time within a time interval of ±10 min.If, for a CCN data point no corresponding AMS data point was available, the CCN data point was excluded from the analysis.The campaign average values of AMS data reported in this study include only data points for which also CCN data were available (see time series plots).
Table 2. Characteristic CCN parameters (arithmetic mean values ± standard deviations) listed for the entire campaign, for the focus periods of aged regional pollution and of fresh city pollution, and for the data from the PRIDE-PRD2006 campaign (entire campaign excluding biomass burning event, Rose et al., 2010): midpoint activation diameters (D a , D t ), maximum activated fractions (MAF f , MAF m ), CDF standard deviations (σ a , σ t ), heterogeneity parameters (σ a /D a , σ t /D t ), hygroscopicity parameters (κ a , κ t ), number concentrations of total aerosol particles (N CN,tot ) and with D > 30 nm (N CN,30 ), number concentrations of cloud condensation nuclei (N CCN,S ), integral CCN efficiencies (N CCN,S /N CN,tot , N CCN,S /N CN,30 ).n ES and n SD are the numbers of averaged CCN efficiency spectra and size distributions, respectively.Subscripts a and t stand for parameters derived from 3-parameter and 2-parameter CDF fits to the measured CCN efficiency spectra, respectively.For corresponding median values see Table S1.

Campaign averages
Figure 2 shows campaign averages of the CCN efficiency spectra (Fig. 2a), the effective hygroscopicity of CCN-active particles (Fig. 2b), the CN and CCN number size distributions (Fig. 2c), and the mass size distributions as obtained by the AMS (Fig. 2d).The average parameters derived from the CCN efficiency spectra are summarized in Table 2 (arithmetic mean ± standard deviation), and the corresponding median values are given in the Supplement (Table S1).
The midpoint activation diameters, D a , increased with decreasing S as expected from Köhler theory (Fig. 2a).At medium to high supersaturation levels (S = 0.26 % to 0.86 %), the CCN efficiency spectra generally reached up to one (maximum activated fraction MAF f ≈ 1) and the normalized standard deviations were small (σ a /D a ≈ 10 %), which implies that nearly all aerosol particles larger than the midpoint activation diameter (D > D a ) were CCN-active.At the lowest supersaturation level (S = 0.07 %), however, the maximum activated fraction remained on average below one, which indicates the existence of externally mixed weakly CCN-active particles with much lower hygroscopicity.The average value of MAF f was ∼0.8 and the minimum values were as low as ∼0.4,implying that on average ∼20 % and up to ∼60 % of the aerosol particles with diameters around ∼250 nm were not CCN-active at S = 0.07 %.According to the κ-Köhler model, particles larger than ∼250 nm that are not activated at S = 0.07 % must have effective hygroscopicity parameters κ < 0.1.Most of these particles, however, can activate at S = 0.26 %, corresponding to a minimum value of κ = 0.01.Similarly high proportions of externally mixed weakly CCN-active particles in polluted megacity air have been reported by Rose et al. (2010Rose et al. ( , 2011) ) for Guangzhou in the Pearl River Delta (PRIDE-PRD2006).Most likely these are externally mixed soot particles freshly emitted from (d) Mass size distributions of organic matter and inorganic ions as determined by AMS and averaged over all AMS data available during the CCN measurement period (Fig. 3).
strong local and regional sources, which is consistent with the relatively high EC concentrations and EC/OC ratios observed in this study (∼6 µg m −3 EC, EC/OC ≈ 1, Table 4) and with the findings of related recent studies (Garland et al., 2008(Garland et al., , 2009;;Achtert et al., 2009;Cheng et al., 2009;Wehner et al., 2009;Su et al., 2010;Rose et al., 2010Rose et al., , 2011)).Dust particles, which would have a similarly low hygroscopicity, are not likely to represent a significant fraction in the size range we investigated in our study (<300 nm), except for dust storm episodes (e.g.Wu et al., 2009).Remote sensing data, however, do not show any evidence for dust storm episodes during our measurement campaign (Fig. S1).The abundance, properties, and effects of the externally mixed, weakly CCN-active particles shall be specifically addressed in follow-up studies.
Figure 2b shows that the campaign average effective hygroscopicity of CCN-active particles increased strongly with particle size.In the Aitken size range (∼40-70 nm), κ a was on average between 0.2 and 0.3, which is consistent with inorganic mass fractions < 50 %.In the accumulation size range (∼70-200 nm), κ a increased up to ∼0.5, which is due to enhanced mass fractions of sulfate and other soluble inorganic ions as illustrated in Fig. 2d and discussed in Sect.3.3.The observed increase of CCN hygroscopicity with increasing particle diameter is also consistent with the size dependence of aerosol hygroscopic growth factors reported by Achtert et al. (2009)  hygroscopicity parameters for the CCN-active particles (κ a ) as well as for the total ensemble of aerosol particles (κ t , Sect.2.2.2) were κ a ≈ κ t ≈ 0.3 ± 0.1 (Table 2).This is consistent with other measurements, estimates and model calculations for the effective average hygroscopicity of aerosols in populated regions (Andreae and Rosenfeld, 2008;Bougiatioti et al., 2009;Pöschl et al., 2009;Shinozuka et al., 2009;Chang et al., 2010;Kammermann et al., 2010;Pringle et al., 2010;Rose et al., 2010Rose et al., , 2011;;Jurányi et al., 2011, and references therein).
As illustrated in Fig. 2c, the campaign average number size distribution of total aerosol particles (CN) peaked near ∼100 nm, and the mean and median values of N CN,tot were 1.6 × 10 4 cm −3 (Table 2) and 1.0 × 10 4 cm −3 (Table S1), respectively.At S = 0.07 % the average CCN activation diameter was much larger than the modal diameter of the CN size distribution (∼200 nm), and the CCN size distributions accounted only for ∼10 % of N CN,tot (∼20 % of N CN,30 , Table 2).At the higher supersaturation levels, the CCN activation diameters were smaller, and the integral CCN efficiencies reached up to ∼70 % of N CN,tot (∼80 % of N CN,30 , Table 2).Compared to the global mean value of N CCN,S /N CN,tot ≈ 0.36 at S ≈ 0.4 % suggested by Andreae (2009), the campaign average value was higher (N CCN,S /N CN,tot ≈ 0.5) but the average value for the fresh city pollution period discussed below was lower (N CCN,S /N CN,tot ≈ 0.2).
The best fit parameters for a monomodal lognormal distribution fitted to the average CN size distribution are listed in Table 3 (mean) and in Table S2 (median), respectively.The campaign average number concentrations of CCN were in the range of N CCN,S ≈ 2 × 10 3 cm −3 at S = 0.07 % to N CCN,S ≈ 1×10 4 cm −3 at S = 0.86 % (Table 2).This is consistent with the values recently reported by Deng et al. (2011) for Wuqing, a town located between Beijing and Tianjin in the Northern China plain.
The campaign average mass size distributions measured by AMS (Fig. 2d) peaked at mass equivalent diameters around ∼200 nm for organic matter and around ∼350 nm for sulfate, nitrate and ammonium ions.As discussed by Takegawa et al. (2009a), the peak diameter of the sulfate mass size distribution exhibited an increase with aging time due to strong secondary sulfate formation.On average, the total mass concentration determined by AMS and integrated over the investigated size range (29-900 nm) was ∼26 µg m −3 with an organic mass fraction of ∼40 % and a sulfate mass fraction of ∼30 % sulfate.The mass fractions of other inorganic ions are listed in Table 4.

Time series and focus periods
Figure 3 shows the temporal evolution and variability of several parameters characteristic for the meteorological conditions, CCN properties, and aerosol composition observed throughout the measurement campaign.For clarity, the CCN parameters are plotted only for the lowest and the highest supersaturation level investigated.The temporal evolution of these parameters determined at intermediate supersaturation levels was generally similar to the one at S = 0.86 %.
Most parameters exhibited more or less pronounced diurnal cycles and some episodes with drastically different characteristics.For example, the integral CCN efficiencies (N CCN /N CN,tot , N CCN /N CN,30 ), the maximum activated fraction at low supersaturation (MAF f at S = 0.07 %), and the inorganic mass fraction determined by AMS exhibited a strong decrease on 20 August and on 3 September.Similar temporal patterns and episodes were reported by Garland et al. (2009), Achtert et al. (2009), andTakegawa et al. (2009a), who found that major changes in aerosol optical properties and chemical composition during CAREBeijing-2006 could be attributed to changes in the prevailing meteorological conditions, e.g.wind speed, wind direction, and air mass origin (back trajectories).To explore the relationship  8. 16.8. 20.8. 24.8. 28.8.1.9.5.9.9.9.
Fig. 3. Time series of meteorological parameters, CCN properties, and aerosol composition: (a) ambient temperature (black), and rain fall (blue), (b) wind speed (black) and ambient relative humidity (red), (c) wind direction for all observed wind speeds (grey shade) and for wind speed >2 m s −1 (blue); (d) effective hygroscopicity of CCN-active particles, (e, f) heterogeneity (normalized width) and maximum activated fraction (amplitude) of 3-parameter CDF fit to observed CCN efficiency spectra, (g) number concentrations of total aerosol particles (N CN,tot ), particles with D > 30 nm (N CN,30 ), and of cloud condensation nuclei, (h) integral CCN efficiencies relative to N CN,tot and to N CN,30 ; (i) organic and inorganic mass fractions determined by AMS and integrated over the investigated size range (29-900 nm).All the parameters are plotted against the date in August-September 2006 (local time).The CCN parameters are plotted for the lowest and the highest investigated supersaturation level (S = 0.07 % and 0.86 %).The red and green shaded areas indicate the aged regional and fresh city pollution episodes, respectively.
between air mass origin and CCN properties, we highlight two distinct focus periods: 1.A focus period of "aged regional pollution", when stagnant air masses came from the highly industrialized regions south of Beijing, where numerous power plants and factories are located, and passed over the measurement site with relatively low wind speed (17 August 00:00 to 18 August 23:59 LT; shaded red in Fig. 3).
2. A focus period of "fresh city pollution", when clean air masses came from the mountain regions in the north, passed over the city center of Beijing with relatively high wind speed (frontal passage), and arrived at the measurement site from wind directions ranging from northeast to northwest (20 August 00:00 to 21 August 23:59 LT, and 3 September 15:00 to 4 September 15:00 LT; shaded green in Fig. 3).The dates of these focus periods were selected according to the back trajectory analysis and the air mass classification in Table 2 of Garland et al. (2009), taking into account the availability of the CCN and AMS data.For both periods, Fig. 4 shows the average number size distributions of CN and CCN as well as the mass size distributions of organic matter and inorganic ions determined by AMS.Characteristic parameters of aerosol composition and CCN activity are summarized in Tables 2, 3, and 4 (mean values), and in the supplementary Tables S1, S2, and S3 (median values).
During the focus period for aged regional pollution (shaded red in Fig. 3), the meteorological parameters as well as some of the aerosol and CCN properties were qualitatively similar to the campaign average conditions outlined above (Sect.3.1, Fig. 2).Differences could be observed in the total aerosol particle number size distribution that peaked at a larger diameter (∼120 nm, Fig. 4a), and in the mass concentrations determined by AMS that were enhanced by factors of 1.5-2 (Fig. 4b, Table 4) compared to the campaign average.The concentrations of sulfate and other inorganic ions were also more strongly enhanced than that of organic compounds, which led to an increase in the hygroscopicity of CCN active particles (κ a = 0.35 ± 0.05).Moreover, the maximum activated fractions of CCN measured at low supersaturation were enhanced (MAF f = 0.91 at S = 0.07 %), which indicates a relatively low abundance of externally mixed weakly CCN-active particles.Due to the increased particle size, increased hygroscopicity, and decreased fraction of externally mixed weakly CCN-active particles, the integral CCN efficiencies were also higher than the campaign average values: from ∼20 % of N CN,tot (∼20 % of N CN,30 ) at S = 0.07 % up to ∼80 % of N CN,tot (∼90 % of N CN,30 ) at S = 0.86 % (Table 2).
During the focus period for fresh city pollution (shaded green in Fig. 3), the aerosol and CCN properties differed drastically from the focus period for aged regional pollution.The total aerosol particle number size distribution peaked at a much lower diameter in the Aitken size range (∼50 nm, Fig. 4c), and the mass concentrations determined by AMS were reduced by factors of 2 to 8 (Fig. 4d, Table 4).The concentrations of sulfate and other inorganic ions were more strongly reduced than that of organic compounds, which led to a decrease in the hygroscopicity of CCN active particles (κ a = 0.22 ± 0.07).Moreover, the maximum activated fractions of CCN measured at low supersaturation were reduced (MAF f = 0.66 at S = 0.07 %), which indicates a relatively high abundance of externally mixed weakly CCN-active particles.Due to the strongly decreased particle size, decreased hygroscopicity, and increased fraction of externally-mixed weakly CCN-active particles, the integral CCN efficiencies were substantially lower than during the aged regional pollution period: from ∼2 % of N CN,tot (∼4 % of N CN,30 ) at S = 0.07 % up to ∼40 % of N CN,tot (∼60 % of N CN,30 ) at S = 0.86 % (Table 2).Thus, the number concentrations of CCN were lower by as much as a factor of ∼7 at S = 0.07 % to a factor of ∼1.2 at S = 0.86 % although the total aerosol particle number concentrations were higher by a factor of ∼2.

Relationship between aerosol chemical composition and hygroscopicity
As outlined above and in earlier studies, the effective hygroscopicity of CCN-active particles can be efficiently linked to the organic and inorganic mass fractions determined by AMS measurements (Gunthe et al., 2009;Dusek et al., 2010;Rose et al., 2011).
Figure 5 shows the correlation of κ a with the organic mass fraction determined by integration over the relevant diameter range in the AMS mass size distribution (D a ±σ a , Sect.2.2.3) using the data points for which the mass concentration in the integration interval was larger than 1 µg m −3 (R 2 = 0.78).Using all available data points, we obtained a weaker correlation (R 2 = 0.37), which can be explained by the limited reliability of mass fractions calculated from AMS data with low concentration in the relevant size range (m D < 1 µg m −3 , Rose et al., 2011).
Extrapolation of the fit line in Fig. 5 to f org = 1 yields an estimate of κ org = 0.06 ± 0.01 for the effective average hygroscopicity of the non-refractory organic particulate matter in the investigated aerosols.This is consistent with the average value of κ org ≈ 0.1 derived from earlier field and laboratory studies (Gunthe et al., 2009;Jimenez et al., 2009;Shinozuka et al., 2009;King et al., 2009;Dusek et al., 2010;Roberts et al., 2010;Rose et al., 2011).Extrapolation of the fit line to f org = 0 yields an estimate of κ inorg = 0.68 ± 0.02 for the effective average hygroscopicity of the non-refractory inorganic particulate matter.This is similar to the characteristic values for ammonium sulfate and related compounds (Petters and Kreidenweis, 2007) and consistent with the average values of κ inorg ≈0.6-0.7 derived from earlier field studies (Gunthe et al., 2009;Dusek et al., 2010;Rose et al., 2011).
Based on the observed correlation, and in analogy to earlier studies, we can use the following simple mixing rule to predict effective hygroscopicity parameters from AMS measurement data (Gunthe et al., 2009;Dusek et al., 2010;Rose et al., 2011):  . 5. Correlation between the effective hygroscopicity parameter of CCN-active particles (κ a ) and the size-integrated organic mass fraction (f org ) as determined by AMS measurements for the aged regional pollution period (red), the fresh city pollution period (green), and the rest of the campaign (black).Different symbols indicate different supersaturation levels.The black line is a linear least square fit with the following equation, correlation coefficient, and number of data points: y = 0.684−0.623x,R 2 = 0.78, n = 271.

Fig
As illustrated in Fig. 6, the size dependence of the particle hygroscopicity as predicted from the AMS measurement data with rounded average values of κ org = 0.1 and κ inorg = 0.7 generally agrees well with the effective average hygroscopicity parameters derived from the CCN measurements.An apparent overprediction of κ a in the Aitken size range during the focus period of fresh city pollution (<100 nm, Fig. 6c) is likely due to enhanced mass fractions of internally mixed soot and elemental carbon (Table 4), which are not captured by the AMS data.In addition, the chemical composition and effective hygroscopicity parameter of the organic and inorganic fractions may be different during the fresh city pollution.Possibilities for improvement using additional data from complementary measurement techniques (e.g. from volatility or hygroscopicity tandem differential mobility analyzers; Rose et al., 2011;Su et al., 2010) and the impact on the prediction of CCN and cloud droplet number concentrations will be explored in follow-up studies.

Comparison between the megacity regions of Beijing and Guangzhou
Directly before the CAREBeijing-2006 campaign, the same instrumentation and experimental setup had been used in the PRIDE-PRD2006 campaign to measure the CCN properties and chemical composition of the aerosol particles at a rural site (Backgarden) ∼60 km northwest of the emerging megacity Guangzhou in the Pearl River Delta (PRD) (Rose et al., 2010(Rose et al., , 2011)).the air masses generally originated from the southeast, and passed over the highly industrialized PRD region, which includes Hong Kong, Guangzhou and several other cities, before arriving at the Backgarden measurement site (monsoon circulation prevailed during the campaign).
During PRIDE-PRD2006 the meteorological conditions were less variable than during CAREBeijing-2006, and the sampled air masses contained a mix of city pollution from Guangzhou, and regional pollution from the Pearl River Delta (Garland et al., 2008;Rose et al., 2010Rose et al., , 2011)).For the comparison with Beijing, we used the PRIDE-PRD2006 data set "entire campaign excluding the biomass burning event" from Rose et al. (2010Rose et al. ( , 2011)).In this data set, the highly polluted period of 23-26 July 2006, which was characterized by intense local biomass burning, was excluded.A similar comparison between data from CAREBeijing-2006 and PRIDE-PRD2006 has been done for the aerosol optical properties by Garland et al. (2009).
Overall, both the meteorological conditions and the CCN properties of PRIDE-PRD2006 represent a mix of aged regional pollution from the Pearl River Delta and fresh city pollution from Guangzhou, whereas the meteorological conditions and CCN properties of CAREBeijing-2006 exhibited a clear separation between aged regional pollution in stagnant air masses, and fresh city pollution in the outflow from Beijing during frontal passages.

Summary and conclusions
During the CAREBeijing-2006 campaign, we performed the first size-resolved CCN measurements in the megacity region of Beijing.The CCN hygroscopicity exhibited a strong size dependence, ranging from κ a = 0.25 to 0.45, i.e. particles in the Aitken size range were much less hygroscopic than in the accumulation size range.Averaged over the entire range of measured aerosol particle size and water vapor supersaturation, the mean value and standard deviation of the CCN hygroscopicity were κ a ≈ 0.3 ± 0.1 This is consistent with other measurements, estimates and model calculations for the effective average hygroscopicity of aerosols in populated continental regions (Andreae and Rosenfeld, 2008;Bougiatioti et al., 2009;Pöschl et al., 2009;Shinozuka et al., 2009;Chang et al., 2010;Kammermann et al., 2010;Pringle et al., 2010;Rose et al., 2010Rose et al., , 2011;;Jurányi et al., 2011, and references therein).
The CCN hygroscopicity and its size dependence could be parameterized as a function of AMS-based organic and inorganic mass fractions using the simple mixing rule κ p ≈ 0.1 • f org + 0.7 • f inorg .An apparent overprediction of κ a in the Aitken size range during the focus period of fresh city pollution was likely due to enhanced mass fractions of internally mixed soot and elemental carbon, which are not captured by the AMS.Possibilities for improvement by additional data from complementary measurement techniques (e.g. from volatility or hygroscopicity tandem differential mobility analyzers; Rose et al., 2011;Su et al., 2010) and the impact on the prediction of CCN and cloud droplet number concentrations will be explored in follow-up studies.
When the measured air masses originated from the north and passed rapidly over the center of Beijing (fresh city pollution), the average particle hygroscopicity was reduced (κ = 0.2±0.1),which is consistent with enhanced mass fractions of organic compounds (∼50 % of PM 1 ) and elemental carbon (∼30 % of PM 1 ).Moreover, substantial fractions of weakly CCN-active particles with much lower hygroscopicity were observed at low supersaturation (S = 0.07 %), which can be explained as being from freshly emitted, externally mixed soot particles.Particles in stagnant air from the industrialized megacity region south of Beijing (aged regional pollution) were on average larger and more hygroscopic, which is consistent with enhanced mass fractions (∼60 %) of soluble inorganic ions (mostly sulfate, ammonium, and nitrate).Accordingly, the number concentration of CCN in aged air from the megacity region was higher ((2.5-9.9)× 10 3 cm −3 vs. (0.4-8.3) × 10 3 cm −3 for S = 0.07-0.86%) although the total aerosol particle number concentration was lower than in the fresh city outflow (1.2 × 10 4 cm −3 vs. 2.3 × 10 4 cm −3 ).
Prior to CAREBeijing-2006, the same experimental equipment and scientific approach had been used in the PRIDE-PRD2006 campaign to characterize the CCN properties and chemical composition of aerosol particles at a rural location in the megacity region of Guangzhou in the Pearl River Delta (Rose et al., 2010(Rose et al., , 2011)).Both the meteorological conditions and the CCN properties of PRIDE-PRD2006 were characteristic for a mix of aged regional pollution from the Pearl River Delta and fresh urban pollution from Guangzhou.The systematic analysis and comparison of the data sets from CAREBeijing-2006 and PRIDE-PRD2006 suggest that the fresh outflow from Chinese megacity centers generally contains more, but smaller and less hygroscopic aerosol particles, and thus fewer CCN than the aged outflow from megacity regions.

Fig. 1 .
Fig. 1.Measurement location Yufa (blue cross) in the megacity region of Beijing, China.The map is color-coded by population density.

Fig. 2 .
Fig. 2. Campaign-average CCN efficiency spectra (a), hygroscopicity parameters (b), number size distributions (c) and mass size distributions (d).Data points and lines represent median values; error bars represent lower and upper quartiles.(a) Size-resolved CCN efficiency spectra at different supersaturation levels (S = 0.07-0.86%): activated particle fraction N CCN /N CN plotted against mobility equivalent diameter D. The lines are 3-parameter cumulative Gaussian distribution function (CDF) fits.(b) Effective hygroscopicity of CCN-active particles (κ a ) plotted against the midpoint activation diameter (D a ).The black line represents the modeled/predicted hygroscopicity parameter (κ p ) as calculated from organic and inorganic mass fractions derived from AMS measurements averaged over the CCN measurement period and using rounded off values as obtained by extrapolating a linear fit in Fig.5.(c) Number size distributions of total aerosol particles (CN) and of cloud condensation nuclei (CCN) averaged over the entire campaign (CCN measurement period where TDMPS data were available).The CCN size distributions were calculated by multiplying the median CN size distribution with the median CCN efficiency spectra.(d) Mass size distributions of organic matter and inorganic ions as determined by AMS and averaged over all AMS data available during the CCN measurement period (Fig.3).

Fig. 4 .
Fig. 4. Number size distributions of total aerosol particles (CN) and of cloud condensation nuclei (CCN) for different supersaturation levels (left column; panels a, c), and mass size distributions of organic and inorganic compounds determined by AMS (right column; panels b, d).The upper panels (a, b) are averages over the aged regional pollution period and the lower panels (c, d) are averages over the fresh city pollution period.

Fig. 6 .
Fig.6.Size distributions of mass fractions of organic (green), inorganic (blue), and sulfate (red) as determined by the AMS (cut-off at 50 nm), and the observed effective hygroscopicity parameter of CCN-active particles (κ a , black data points, median values with error bars extending to upper and lower quartiles) averaged over the entire campaign (a), the aged regional pollution period (b), and the fresh city pollution period (c).The grey line in each panel indicates the effective hygroscopicity parameter (κ p ) modeled/predicted from the corresponding median organic and inorganic mass fractions as determined by size-resolved AMS measurements calculated by using rounded off values as obtained by extrapolating a linear fit in Fig.5(κ p = f org • 0.1 + f inorg • 0.7).Data points at which m D < 1 µg m −3 were excluded.

Table 1 .
List and short description of symbols and parameters frequently used in this study.Details about definitions and derivations are described in the text.

Table 3 .
(Rose et al., 2010) of monomodal lognormal size distribution functions fitted to the mean number size distribution of aerosol particles (CN) for the entire campaign, the focus periods (aged regional pollution, fresh city pollution), and the mean number size distribution observed during PRIDE-PRD2006(Rose et al., 2010): integral number concentration (N CN ), count median or geometric mean diameter (D g ), and geometric standard deviation (σ g ).

Table 4 .
Mass concentrations and corresponding mass fractions of organic matter and inorganic ions determined by aerosol mass spectrometry (AMS).Apparent elemental and organic carbon (EC, OC) mass concentrations determined by thermo-optical measurements and mass fractions of EC and OC relative to the estimated total concentration of fine particulate matter (PM 1 ≈ AMS+EC).Arithmetic mean values and standard deviations are listed for the entire campaign and for the focus periods of aged regional pollution and of fresh city pollution.