Interactive comment on “Size-resolved and integral measurements of cloud condensation

Abstract. As part of the CLACE-6 campaign we performed size-resolved CCN measurements for a~supersaturation range of S = 0.079 % to 0.66% at the high-alpine research station Jungfraujoch, Switzerland, in March~2007. The derived effective hygroscopicity parameter κ describing the influence of particle composition on CCN activity was on average 0.23–0.30 for Aitken (50–100 nm) and 0.32–0.43 for accumulation mode particles (100–200 nm). The campaign average value of κ = 0.3 is similar to the average value of κ for other continental locations. When air masses came from southeasterly directions crossing the Po Valley in Italy, particles were much more hygroscopic (κ a 0.42) due to large sulfate mass fractions. The κ values obtained at S = 0.079 % exhibited a good negative correlation with the organic mass fractions derived from PM 1 aerosol mass spectrometer (AMS) measurements. Applying a simple mixing rule the organic and inorganic mass fractions observed by the AMS could be used to reproduce the temporal fluctuations of the hygroscopicity of accumulation mode particles quite well. We show how during a cloud event the aerosol particles were activated as cloud droplets and then removed from the air by precipitation leaving behind only a small amount of accumulation mode particles consisting mainly of weakly CCN-active particles, most likely externally mixed unprocessed soot particles. During the campaign we had the opportunity to directly compare two DMT CCN counters for a certain time. The total CCN concentration ( N CCN,tot ) obtained by the two instruments at equal supersaturations agreed well for both possible operating modes: detecting N CCN,tot directly by sampling the polydisperse aerosol with the CCNC, or indirectly by combining size-resolved measurements of the activated fraction with parallel measurements of the particle size distribution (e.g., by SMPS). However, some supersaturation setpoints differed between the two CCNCs by as much as 20% after applying the instrument calibrations, which resulted in differences of the corresponding N CCN,tot of up to 50%. This emphasizes that it is extremely important to carefully calibrate the supersaturation of the instrument, especially at low S .


Abstract
As part of the CLACE-6 campaign we performed size-resolved CCN measurements for a supersaturation range of S = 0.079 % to 0.66 % at the high-alpine research station Jungfraujoch, Switzerland, in March 2007.The derived effective hygroscopicity parameter κ describing the influence of particle composition on CCN activity was on average 0.23-0.30for Aitken (50-100 nm) and 0.32-0.43 for accumulation mode particles (100-200 nm).The campaign average value of κ = 0.3 is similar to the average value of κ for other continental locations.When air masses came from southeasterly directions crossing the Po Valley in Italy, particles were much more hygroscopic (κ ≈ 0.42) due to large sulfate mass fractions.The κ values obtained at S = 0.079 % exhibited a good negative correlation with the organic mass fractions derived from PM 1 aerosol mass spectrometer (AMS) measurements.Applying a simple mixing rule the organic and inorganic mass fractions observed by the AMS could be used to reproduce the temporal fluctuations of the hygroscopicity of accumulation mode particles quite well.
We show how during a cloud event the aerosol particles were activated as cloud droplets and then removed from the air by precipitation leaving behind only a small amount of accumulation mode particles consisting mainly of weakly CCN-active particles, most likely externally mixed unprocessed soot particles.
During the campaign we had the opportunity to directly compare two DMT CCN counters for a certain time.The total CCN concentration (N CCN,tot ) obtained by the two instruments at equal supersaturations agreed well for both possible operating modes: detecting N CCN,tot directly by sampling the polydisperse aerosol with the CCNC, or indirectly by combining size-resolved measurements of the activated fraction with parallel measurements of the particle size distribution (e.g., by SMPS).However, some supersaturation setpoints differed between the two CCNCs by as much as 20 % after applying the instrument calibrations, which resulted in differences of the corresponding N CCN,tot of up to 50 %.This emphasizes that it is extremely important to carefully calibrate the supersaturation of the instrument, especially at low S.

Introduction
Atmospheric aerosol particles that enable the condensation of water vapor and thus form cloud droplets are called cloud condensation nuclei (CCN).They indirectly influence the Earth's climate by determining cloud microphysics, the formation of precipitation and cloud radiative properties (cloud albedo) (e.g., Lohmann and Feichter, 2005).
The response of cloud characteristics and precipitation processes to increasing anthropogenic aerosol concentrations represents one of the largest uncertainties in the current understanding of climate change (IPCC, 2007).Therefore it is crucial to understand the factors that influence the properties and abundance of CCN.Since the temporal and spatial distribution of CCN is highly variable it is necessary to perform measurements at various seasons and locations to improve the incorporation of CCN in meteorological cloud and climate models.
The high-alpine research station Jungfraujoch in Switzerland provides the opportunity to measure continental background aerosols since most of the time the station is Figures located in the free troposphere but sometimes it is also influenced by injections from the planetary boundary layer (Collaud Coen et al., 2011).The hygroscopicity and CCN activity of aerosol particles as well as their activation behavior in ambient clouds have been investigated at the Jungfraujoch for several years (e.g., Henning et al., 2002;Nessler et al., 2003;Cozic et al., 2007;Sjogren et al., 2008;Kammermann et al., 2010;Jurányi et al., 2010Jurányi et al., , 2011;;Hammer et al., 2013).These studies found that there is only little seasonal trend in the hygroscopicity and CCN activity.On the other hand the number fraction of particles that can be activated under a certain water vapor supersaturation exhibits a significant seasonal cycle due to strong seasonal variations of the number fraction of larger particles.During influence from the planetary boundary layer, the hygroscopicity and CCN activity are slightly reduced.Measurements on the mixing state of particles have been investigated only under sub-saturated conditions yet.
Overall, it has been found that the observed chemical composition, the hygroscopic growth, and the CCN activity are generally consistent.
In this study, we present size-resolved CCN measurements on the Jungfraujoch performed during the CLACE-6 campaign in March 2007.We relate the CCN properties to the meteorological conditions and compare them with measurements of the particles' chemical composition.1998).Nevertheless, convective anticyclonic and convective indifferent weather situations, which occur mostly from April to September, influence the JFJ through injections of air parcels from the planetary boundary layer.Therefore aerosol concentrations exhibit a seasonal cycle with a maximum in summer and a minimum in the winter (Collaud Coen et al., 2011).More detailed information about the JFJ site and long term measurements of aerosols can be found in Baltensperger et al. (1997); Cozic et al. (2008) and Collaud Coen et al. (2007, 2011).

Meteorological conditions
The average meteorological parameters (arithmetic mean ± standard deviation) recorded during the CCN measurement period near the aerosol inlet were: (262 ± 3) K ambient temperature, (65 ± 33) % ambient relative humidity, and (653 ± 5) hPa ambient pressure.The time series of the basic meteorological parameters displayed in Fig. A1 do not indicate any diurnal cycle.Except during two periods with clear south easterly winds (6-7 March and 11-13 March), the local wind direction fluctuated a lot.The measurement site was in clouds on 60-65 % of the time during the campaign period.
Three-day backward trajectories were calculated from the HYSPLIT Trajectory Model by NOAA ARL (http://ready.arl.noaa.gov/HYSPLIT.php)using the global GDAS meteorological data with a start height of 3500 m a.s.l., which were in accordance with the Alpine Weather Statistic (AWS) synoptic weather classification system (Schüpp, 1979).Figure 1 shows that the air masses came from several directions during the measurement period.In the beginning (3-6 March) air masses came mainly from westerly directions predominantly crossing central and south France.

Aerosol inlet
During the CLACE-6 campaign three different types of aerosol inlets were used sampling ice nuclei (Ice-CVI, Mertes et al., 2007), interstitial particles (non-activated particles inside clouds < 2.5 µm in diameter) and total particles (interstitial particles and cloud residuals).In this study we used only data recorded by sampling through the total aerosol inlet and interstitial inlet.The total aerosol inlet consisted of a heated and insulated vertical stainless-steel tube (length: 2 m, diameter: 6 cm) and a heated snow-hood (Weingartner et al., 1999).The sampled air was heated to 25 • C in order to evaporate any water from hydrometeors (cloud droplets and ice crystals), thereby releasing the residual particles.During cloud events, the aerosol sampled at the total inlet thus consisted of both the residual particles from dried hydrometeors as well as the inactivated (interstitial) particles.
The interstitial inlet consisted of a cyclone with a 2.5 µm cut-off (PM 2.5 ) segregating all hydrometeors with a diameter > 2.5 µm and letting through only smaller, unactivated particles.During cloud-free periods the aerosol sampled through the interstitial inlet corresponded to the total aerosol.Right below the inlet, the sampling line was split into separate lines.One led to the CCN measurement setup (∼ 8 m length of sampling line), and to the aerosol mass spectrometer setup (∼ 5 m length of sampling line).Another one was used for aerosol particle size distribution measurements (SMPS, ∼ 1 m length of sampling line).

Cloud condensation nuclei (CCN)
Two continuous flow CCN counters (DMT CCNC, Roberts and Nenes, 2005;Lance et al., 2006)  campaign the PSI CCNC was used solely in a polydisperse mode detecting the total number concentration of CCN (N CCN,tot ) at a defined supersaturation.The MPI CCNC was mainly operated in a monodisperse mode, for which it was coupled to a differential mobility analyzer (DMA; TSI 3071; sheath flow 10 L min −1 ; sample flow 1.8 L min −1 ) and a condensation particle counter (CPC; TSI 3762; sample flow 1.0 L min −1 ).In the monodisperse mode, the aerosol first passed through a bipolar charger (Ni-63, 555 MBq) to reach charge equilibrium.Then, the charged aerosol entered the DMA where particles of a certain electrical mobility (corresponding to a certain particle size) were selected and sent to the CPC and the CCNC to measure in parallel the aerosol particle number concentration (N CN ) and the number concentration of particles that can act as CCN (N CCN ), respectively, (i.e., size-resolved CCN efficiency spectra or activation curves; Frank et al., 2006;Rose et al., 2008).Only at the end of the campaign the MPI CCNC was also used in a polydisperse mode for about 15 h.The MPI CCNC and PSI CCNC were operated at a total flow rate of 0.8 L min −1 and 1.0 L min −1 , respectively, both with a sheath-to-aerosol flow ratio of 10.The water vapor supersaturation (S) in the CCNC, which is determined by the temperature difference between the upper and lower end of the CCNC flow column (∆T ), was calibrated using size-selected ammonium sulfate particles according to the procedure described by Rose et al. (2008) and using their Köhler model AP3 for calculation of the thermodynamic properties.The calibrated supersaturation set points were 0.079 %, 0.17 %, 0.27 %, 0.46 %, and 0.66 % for the MPI CCNC and 0.099 %, 0.20 %, 0.29 %, 0.48 %, and 0.67 % for the PSI CCNC (relative uncertainty ∆S/S < 10 % for each CCNC).The two CCNCs sampled aerosol particles only through the total aerosol inlet.

Monodisperse CCN measurements
For each measurement cycle in the monodisperse mode, ∆T was set to 5 different levels (2.4-9.The measurement data of the CCN efficiency spectra were corrected for the effects of multiply charged particles as described by Rose et al. (2008Rose et al. ( , 2010)), and the corrected data were used for further analysis.For the charge correction we used aerosol particle number size distributions that were measured in parallel by a scanning mobility particle sizer (SMPS; custom-built and validated during an SMPS intercomparison workshop (Wiedensohler et al., 2012); scanning between 17 nm and 470 nm mobility equivalent diameter), which was operated directly below the total aerosol inlet (see Sect. 2.3.1).For the times at which no SMPS data were available (∼ 35 % of the scans) we performed no charge correction.Nevertheless, the CCN data from these periods remained comparable with the rest of the days, as the effects of the charge correction were generally small (< 5 % change in activation diameter and other parameters used for further calculations).We also corrected the CCN efficiency spectra for differences in the CCNC and CPC counting efficiencies by using a constant correction factor f corr = 1.04, which we obtained from the supersaturation calibration experiments (Rose et al., 2010).The uncertainties of N CCN , N CN and N CCN /N CN are 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)).Due to the low particle number concentrations Figures prevailing at the JFJ, the CCN concentrations measured at the smallest particle diameters (< 40 nm) were affected by low signal-to-noise ratios and were therefore omitted.Each corrected CCN efficiency spectrum was fitted with a 3-parameter cumulative Gaussian distribution function (CDF), from which the following parameters were derived (Rose et al., 2010): the maximum activated fraction MAF, the midpoint activation diameter D a , and the standard deviation σ a .The CDF standard deviation is a general indicator for the extent of mixing, i.e., for the heterogeneity of the particle composition in the investigated aerosol.Under ideal conditions, σ a should be zero for an internally mixed aerosol with particles of homogeneous chemical composition.Even 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, finite width of the DMA's transfer function, or particle shape effects.Thus, normalized CDF standard deviation or "heterogeneity parameter" values of σ a /D a ≈ 3 % indicate internally mixed homogeneous CCN whereas higher values indicate mixtures of particles with different chemical composition and hygroscopicity (Rose et al., 2010(Rose et al., , 2011)).
For all data pairs of supersaturation S and activation diameter D a derived from the CCN efficiency spectra measured in this study, the effective hygroscopicity parameter κ a (Petters and Kreidenweis, 2007;Pöschl et al., 2009) was calculated using the κ-Köhler model equations and parameters specified in Rose et al. (2010) (surface tension of 0.072 J m −2 , temperature of 303 K).The parameter κ a characterizes the average hygroscopicity of CCN-active particles in the size range around D a .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.
The deviation of MAF from unity represents the number fraction of externally mixed CCN-inactive particles at the largest diameter measured (D max ) and the nearby diam-Figures

Back Close
Full eter range with a plateau of activated particle fraction (Gunthe et al., 2011).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).CCN number size distributions (dN CCN /d log D) at a certain supersaturation were calculated by multiplying the CCN efficiency spectra (3-parameter CDF fits of N CCN /N CN ) measured at this S with the aerosol particle (CN) number size distributions (dN CN /d log D) measured in parallel.From the monodisperse measurements, total CCN number concentrations at a certain supersaturation (N CCN,tot,m ) were calculated by stepwise integration of the CCN number size distributions across the whole diameter range (for further details see Rose et al., 2010).Note that a list of frequently used symbols can be found in Table A1.

Polydisperse CCN measurements
For each measurement cycle in the polydisperse mode, the temperature difference in the CCNC, ∆T , was set to 5 different levels corresponding to S values in the range of 0.099-0.67% for the PSI CCNC and 0.079-0.66% for the MPI CCNC.The PSI CCNC was operated in the polydisperse mode during the entire campaign, whereas the MPI CCNC was operated in this mode only from 13 March 18:00 LT to 14 March 09:30 LT.The total CCN concentration from polydisperse measurements (N CCN,tot,p ) at a certain supersaturation was recorded with 1 Hz.It was averaged over 8 min sampling time and the recording of a full cycle of 5 supersaturations took 60 min including 3 min adjustment time for switching to the next supersaturation level and 5 min extra time for switching from highest to lowest supersaturation.
Since the sampling line between the inlet and the CCN instruments was rather long (see Sect. 2.3.1) it would be necessary to correct the measurements of N CCN,tot,p for diffusional losses.This is, however, not possible, because these kind of losses are size dependent and no information about the particle size is provided in the direct measurement of N CCN,tot,p .Therefore, in this paper, the directly measured total CCN concentration (N CCN,tot,p ) is only used for the comparison of the two CCNCs.Note Figures

Back Close
Full that the total CCN number concentration that is calculated from dN CCN /d log D, i.e.N CCN,tot,m , is independent of any sampling losses between total aerosol inlet and the CCNC, because it is referred to CN size distributions as obtained with the SMPS right below the aerosol inlet (Sects.2.3.1 and 2.3.2).

Aerosol mass spectrometry (AMS)
An Aerodyne High-Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-ToF-AMS, hereafter short AMS) was used to measure the chemical composition of nonrefractory submicron aerosol particles.The data set used in this study includes all times when the AMS was either attached to the total inlet or to the interstitial inlet during cloud-free passages, i.e., the sampled aerosol could be always regarded as total aerosol.
The AMS was described in detail by DeCarlo et al. (2006) and Canagaratna et al. (2007).Briefly, it samples particles in the size range between 40 and 1000 nm (vacuum aerodynamic diameter (D va ), see DeCarlo et al., 2004) through a critical orifice (120 µm for Jungfraujoch pressure conditions) and an aerodynamic lens and evaporates the particles on a hot surface (∼ 600 • C).The evaporated molecules are ionized by electron  − ).The collection efficiency with respect to particle bounce (e.g., Matthew et al., 2008;Middlebrook et al., 2012) was set to 0.5 for this dataset.Due to the low particle mass concentration levels at the JFJ, it was necessary to integrate over the entire particle size range to obtain time series of the mass concentrations of individual species as well as organic and inorganic mass fractions (f org , f inorg ).Size-resolved chemical composition was determined only as an average for the Figures

Back Close
Full measurement period discussed in this study.Due to the particularly high scatter, the ammonium size distribution was smoothed by scaling with the nitrate size distribution.Complementary information on the chemical composition of particles observed at the JFJ can be found e.g., in Henning et al. (2003); Cozic et al. (2007Cozic et al. ( , 2008)); Cziczo et al. (2009); Kamphus et al. (2010); Liu et al. (2010) and Ebert et al. (2011).

Intercomparison of the two CCN counters
For a certain time during the CLACE-6 campaign we operated two DMT CCNCs in parallel (MPI CCNC, and PSI CCNC) to test and compare our used calibration and data analysis methods.The two instruments were operated at equal (uncalibrated) nominal supersaturation, unfortunately resulting in different true supersaturation setpoints.For the highest three S levels, the (calibrated or true) supersaturation in the MPI CCNC was only slightly lower than in the PSI CCNC (< 10 %), but for the two lowest S levels, the difference in S was more significant (15-20 % lower in MPI than in PSI CCNC). Figure 2a displays a scatter plot of the total CCN concentrations measured by the MPI and the PSI CCNC when both instruments were operated in the polydisperse mode (N CCN,tot,p ), i. e., when they both measured the number concentration of CCN directly, without size-segregation of the particles.For all S, the CCN number concentrations measured by the MPI and the PSI CCNC exhibited a very good correlation (R 2 > 0.99).For the three highest S levels the values of N CCN,tot,p agreed very well (on average MPI ∼ 6 % smaller than PSI, Table 2).For the two lowest S, however, the MPI CCNC measured significantly lower concentrations (on average up to ∼ 50 % at lowest S).This enormous difference can be explained by comparing N CCN,tot,p measured at different supersaturations (∼ 20 % difference in S between MPI and PSI CCNC at the lowest level).The strong dependence of N CCN,tot on S at low levels implies that careful CCNC calibration is crucial, particularly at low S (Rose et al., 2008), similar to the fact that Figures N CCN,tot is more sensitive to the mean particle hygroscopicity in this supersaturation range (Jurányi et al., 2010).
Figure 2b shows the comparison of the CCN concentrations measured by MPI and PSI CCNC when the MPI CCNC was operated in a monodisperse/size resolved mode (Sect.2.3.2) and the PSI CCNC was operated in a polydisperse mode.Note that for this comparison the aerosol particle size distribution (dN CN /d log D; measured by SMPS right below the aerosol inlet) used for the calculation of the total MPI CCN number concentration (N CCN,tot,m (MPI)) was corrected for diffusional losses along the sampling line of the PSI CCNC.As expected the data points exhibited a linear correlation, but with a smaller correlation coefficient (R 2 ∼ 0.84 over all S levels) than for the comparison shown in Fig. 2a.The main reason for additional scatter is most likely associated with problems to get different measurements with low time resolution to match.The SMPS, polydisperse CCNC (PSI), and monodisperse CCNC (MPI) have time resolutions of 6 min, 1 h (all S), and 2.5 h (all S), respectively.
Except for the lowest S, the MPI CCN number concentration was larger than the PSI concentration (on average ∼ 20 % for the three highest S, Table 2).Accounting for the difference in the supersaturation of the two CCNCs, i.e., subtracting the average difference in the CCN number concentration as determined in the comparison of polydisperse measurements (N CCN,tot,p ), the MPI CCN number concentration was on average 20 % to 30 % higher than the PSI concentration.This overestimation was inde- To date only a few studies have compared total CCN number concentrations observed by poly and monodisperse CCN measurements, all of them reporting similarly good agreement between the two approaches as our study: Moore et al. (2010) observed relative deviations of ∼ 10-20 % at a total concentration level of ∼ 1000 cm −3 and a supersaturation of 0.3 %.In a supersaturation range of 0.06 % to 0.7 %, Deng et al. (2011) found on average a 19 % over-estimation of mono compared to polydisperse measurements, which they attributed to water depletion in the CCNC due to the high particle concentrations (on the order of 10 3 -10 4 cm −3 ).Zhang et al. ( 2012) operated a DMT CCNC-200 using one supersaturation column for polydisperse and one for monodisperse CCN measurements.In a supersaturation range of 0.1 % to 0.4 % and a concentration range of 2000-15 000 cm −3 , they reported correlations with a slope of 0.7 to 1.In Jurányi et al. (2013), the two methods correlated with slopes of 0.96-1.05(R 2 = 0.75-0.84)for a supersaturation range of 0.2 % to 0.4 % and a concentration range of 100-5000 cm −3 .

CCN efficiency spectra and size distributions
Figure 3a shows the average measured CCN efficiency spectra over the entire measurement period (3 to 13 March 2007) at different supersaturation levels.The average parameters derived from the CCN efficiency spectra are summarized in Table 2 (arithmetic mean value ± standard deviation).
The midpoint activation diameters D a increased with decreasing S and were larger than the critical diameters for CCN activation of pure ammonium sulfate particles at the given supersaturation levels.The heterogeneity parameter (σ a /D a ) increased from 0.09 to 0.20 with increasing S (i.e., with decreasing particle size) indicating that larger particles were more homogeneously mixed than smaller ones.The average CCN efficiency spectra reached a MAF of unity also for the lowest S, which is in contrast to what has often been observed in the past for other size-resolved CCN measurements Introduction

Conclusions References
Tables Figures

Back Close
Full in continental regions (e.g.Gunthe et al., 2011;Rose et al., 2010Rose et al., , 2011;;Paramonov et al., 2013), whose average spectra approached only ∼ 0.9 and lower.In Fig. 3b, the number size distributions averaged over the entire measurement period are shown for aerosol particles (CN) and for CCN at different supersaturation levels.The corresponding average values (mean value ± standard deviation) of the total CN number concentration (N CN,tot , integrated between 17 and 470 nm), the total CCN number concentration (N CCN,tot,m ), and the CCN efficiency (N CCN,tot,m /N CN,tot ) for each supersaturation level are summarized in Table 2.
The average CN number size distribution for the entire measurement period was dominated by two modes, one with a peak at 50 to 60 nm and another one at 120 to 140 nm.The mean total CN number concentration (between 17 and 470 nm) was ∼ 400 cm −3 .As seen from the time series of the CN number size distribution (Fig. A2), the shape of the size distribution varied strongly during the measurement period.Occasionally, only an Aitken mode (peak around 40-60 nm, e.g., morning of 3, 9, and 10 March), or only an accumulation mode (peak around 100-200 nm, e.g., noon of 4 March, evening of 5 March) was present.
At S = 0.079 %, the total CCN number concentration (∼ 50 cm −3 ) accounted on average for ∼ 13 % of N CN,tot because only a minor fraction of particles was larger than the activation diameter at this supersaturation (D a ∼ 170 nm).At S = 0.17-0.66%, the activation diameters were at or below the diameter at which the accumulation mode peaks and the integral CCN efficiencies were substantially higher (∼ 0.3-0.6;with N CCN,tot,m ≈ 100-250 cm −3 ).The observed CN and CCN number concentrations are similar to other integral measurements at the JFJ (Jurányi et al., 2010(Jurányi et al., , 2011)); they are about the same as in other remote clean regions (Amazon Basin: Roberts et al., 2001;Gunthe et al., 2009); about one order of magnitude smaller than in polluted regions

Hygroscopicity parameter
The campaign average of the observed hygroscopicity parameter κ a derived from the CCN efficiency spectra over all measured size ranges was 0.32 (Table 2).This is consistent with other measurements, estimates and model calculations for the effective average hygroscopicity of aerosols in continental regions (e.g., Andreae and Rosenfeld, 2008;Shinozuka et al., 2009;Pöschl et al., 2009;Chang et al., 2010;Pringle et al., 2010;Rose et al., 2010Rose et al., , 2011;;Gunthe et al., 2011;Wu et al., 2013, and references therein).Figure 4 presents the average κ a plotted vs. particle size, i.e., vs. the mean activation diameter observed for the different S-levels.It shows a pronounced size-dependence ranging from κ a ≈ 0.2 for Aitken mode particles around 50 to 70 nm up to κ a ≈ 0.43 for accumulation mode particles of ∼ 170 nm diameter.The observed κ values and size-dependence are comparable with other observations at the JFJ by Jurányi et al. (2011) and Bukowiecki et al. (2011), who found an increasing particle hygroscopicity from κ ∼ 0.15 to 0.3 within a size range of 50 to 150 nm.The size dependence in κ a resulted from differences in the chemical composition of the particles: as evident from Fig. 5 particles of diameters below ∼ 100 nm consisted on average mainly of organic material with relatively low hygroscopicity, whereas particles with D > 100 nm incorporated larger amounts of inorganic ions leading to higher particle hygroscopicity as will be discussed in Sect.

Time series and meteorological conditions
Figure 6 shows the time series of several characteristic CCN parameters over the time period of 3 to 13 March with some interruptions due to technical difficulties and instrument maintenance.For better readability the parameters are plotted only for 0.079 %, 0.17 %, and 0.66 % supersaturation.The values for S = 0.27 % and 0.46 % were generally in between those for S = 0.17 % and 0.66 %.Note that the CCN parameters measured at different supersaturation levels represent the activation properties of the particles of different size ranges (S = 0.079 % to 0.17 %: 100 nm to 200 nm, S = 0.27 % to 0.66 %: 50 nm to 100 nm).
All CCN parameters exhibited strong temporal variations: the hygroscopicity parameter κ a varied between 0.1 and 0.7; the maximum activated fraction MAF was on average 1 but exhibited also very small values of ∼ 0.3 (see Sect. 3.5 for detailed explanation); and σ a /D a varied between 0.03 (quasi homogeneously mixed aerosol) to 0.4 (very heterogeneously mixed aerosol).The CCN concentration measured at every individual supersaturation level varied over one order of magnitude, from ∼ 10 cm −3 at S = 0.079 % to ∼ 1000 cm −3 at S = 0.66 %.Also the total CCN efficiency varied strongly around the mean value obtained for the individual supersaturation levels.
To find out the reason for the temporal variations of the CCN parameters we compared them with different meteorological variables.We did not find any clear difference in the CCN properties with regard to the weather classification (convective cyclonic and advective or indifferent/anticyclonic convective weather types) as described in Collaud Coen et al. (2011).With regard to the origin of air masses (analysis of back trajectories), however, we could find distinct differences in the CCN properties.As detailed in Sect.2.2 and shown in Fig. 1 we could identify different flow directions that occurred during our measurement period.When air masses came from easterly directions, particles were much more CCN active than during the rest of the campaign (κ a of 0.42 compared to 0.27, Table 2 and Fig. 4).This increase in κ a can be explained by large sulfate mass concentrations (Fig. 7a) coming probably from the densely populated and Figures

Back Close
Full highly industrialized Po Valley resulting in a high inorganic mass fraction (up to 80 %, Fig. 7b).The higher hygroscopicity was particularly evident for the Aitken mode particles (50 % larger κ a than for Aitken particles during the rest of the campaign).Moreover, the maximum activated fraction at S = 0.079 % was significantly reduced during the easterly air masses (on average MAF = 0.9) indicating that there was a substantial fraction (10 %) of particles that could not be activated as cloud droplets.The same feature has been reported earlier (Rose et al., 2011;Gunthe et al., 2011) and can be explained by a significant fraction of relatively fresh externally mixed weakly CCN-active soot particles with effective hygroscopicity parameters around κ a = 0.01.On the other hand, the particles that were CCN active at this supersaturation (particle diameter of ∼ 160 nm) were more homogeneously mixed (σ a /D a = 0.06) than during the rest of the campaign (σ a /D a ≈ 0.09).

Dependence of particle hygroscopicity on chemical composition
Figure 8 shows the correlation of the hygroscopicity parameter κ a with the organic mass fraction f org for all times when both CCN and AMS data were available.Note that the individual f org data points were calculated from the total mass concentrations over the size range of D va = 40-1000 nm (Sect.2.3.3).As already shown in Fig. 7b, the organic mass fractions varied between ∼ 0.2 (minimum on 12 March) and almost 1.0 (maximum on 7 March) with an average of 0.43 over the investigated time.As expected the particle hygroscopicity decreased with increasing f org .The best correlation (R 2 = 0.5) we could find for κ a values observed at the smallest supersaturation level, which corresponds to the largest particle size range.The correlation deteriorated with increasing supersaturation, i.e., with decreasing size of the investigated particles.We expected this result as larger particles contribute more to the total mass than smaller ones do and therefore the measured bulk chemical composition is more representative for the larger diameters.When extrapolating the linear least squares fit line in Fig. 8a to f org = 0, we obtained κ inorg = 0.63 for the hygroscopicity of the inorganic fraction, which Figures

Back Close
Full is characteristic for ammonium sulfate and related compounds (Petters and Kreidenweis, 2007).Extrapolation to f org = 1, yielded κ org = 0.18 for the hygroscopicity of the organic fraction.The values for κ org and κ inorg became smaller for the larger supersaturation levels.Similar values for κ inorg have been also derived from other recent studies comparing AMS with CCN measurements (e.g., Gunthe et al., 2009Gunthe et al., , 2011;;Rose et al., 2011;Dusek et al., 2010;King et al., 2009;Shinozuka et al., 2009;Roberts et al., 2010;Jimenez et al., 2009;Duplissy et al., 2011;Sjogren et al., 2008), while κ org is slightly higher.This is likely due to longer aging and higher oxygenation of the particulate organic matter sampled in free tropospheric air at the Jungfraujoch compared to the boundary layer air sampled in the preceding studies in rainforest (Gunthe et al., 2009) and megacity environments (Rose et al., 2011;Gunthe et al., 2011).
In analogy to Gunthe et al. (2009Gunthe et al. ( , 2011) ) and Rose et al. (2011), κ org and κ inorg can be inserted in a simple mixing rule: κ p = κ inorg • f inorg + κ org • f org to predict the size and time dependence of aerosol hygroscopicity based on the inorganic and organic mass fractions determined by AMS. Figure 4 displays κ p as calculated from the campaign mean size distribution of organic and inorganic mass fractions (Fig. 5b).It shows an increase of particle hygroscopicity with increasing size and reproduces the campaign average of size dependent κ a quite well.Figure 6a displays the time series of κ p , which was calculated from the organic and inorganic mass fractions integrated over the entire particle size-range of AMS measurements (Fig. 7b).In Fig. 6a is shown that κ p followed the temporal evolution of κ a at S = 0.079 % quite well for most of the time during the campaign.This result was expected since already f org correlated best with κ a observed for the lowest supersaturation level (S = 0.079 %, Fig. 8).
Note that the standard ZSR mixing rule for volume based hygroscopicity parameters would require the use of volume rather than mass fractions.For simplicity, however, we refrain from making assumptions on particle density here as these introduce additional uncertainties but do not improve the correlation.This has already been discussed by Gunthe et al. (2009) and Mikhailov et al. (2013) and will be further addressed in followup studies.

Precipitation event
As evident from Fig. 6b, the aerosol and CCN properties during the night of 9/10 March seemed to be exceptionally different from the rest of the time.For a more detailed investigation of this event, Fig. 9 shows the temporal evolution of selected aerosol and CCN parameters at S = 0.66 % between 9 March 21:00 LT and 10 March 12:00 LT.
In the late evening of 9 March, the total particle number concentration at S = 0.66 % was very high (up to ∼ 1400 cm −3 , which is 3-4 times higher than the campaign average value; Fig. 9a), and the particle size distribution exhibited two modes, an Aitken mode with a maximum around ∼ 80 nm, and an accumulation mode with a maximum around ∼ 200 nm (Fig. 9b).At 22:46 LT, the CCN efficiency spectrum at S = 0.66 % reached from zero to MAF = 0.91 (Fig. 9d), which is considerably lower than typical at this supersaturation (MAF f ≈ 1).The activation diameter (∼ 70 nm) was larger than the campaign average (∼ 50 nm), i.e., particles in this size range were less hygroscopic than on average (κ a ≈ 0.1 vs. average κ a ≈ 0.23).About 60 % of all particles were CCNactive at S = 0.66 %, which is also similar to the value obtained for the entire campaign (N CCN,tot, m /N CN,tot = 0.60).
Shortly after midnight, the total particle number concentration suddenly dropped down to ∼ 470 cm −3 (at 01:21 LT) because the accumulation mode disappeared almost completely.Only ∼ 20 % of all particles (N CN,tot ) were still CCN-active at S = 0.66 %.The CCN efficiency spectrum reached only up to half of unity (i.e., MAF = 0.48), which means that ∼ 50 % of the particles of 110 nm in diameter were CCN-inactive even at this high level of supersaturation, i.e., they had a κ of ≤ 0.03.As shown by Rose et al. (2011) it can be assumed that this fraction of CCN-inactive particles consisted of externally mixed weakly CCN-active soot particles with very low hygroscopicity.In a polluted megacity region in China, Rose et al. (2011) also observed a large fraction of these particles (up to 60 %) but only at low supersaturations (≤ 0.27 %).At S = 0.66 %, those particles were already activated indicating that the particles we observed here at the JFJ must have been even less hygroscopic.A possible explanation of this event (rapidly Introduction

Conclusions References
Tables Figures

Back Close
Full decreasing particle concentration together with increasing fraction of CCN-inactive particles) could be that most of the accumulation mode particles we observed at 22:46 LT on 9 March were activated as cloud droplets, precipitated and were thus removed from the sampled air we measured at 01:21 LT on 10 March.
If we calculate the absolute number concentration of particles in the 110 nm size bin that are CCN-inactive at S = 0.66 %, we obtain a value of 116 cm −3 at 22:46 LT (∼ 10 % of 1160 cm −3 ) and an almost equal value of 135 cm −3 at 01:21 LT (∼ 50 % of ∼ 270 cm −3 ). Figure A3 shows the number size distribution of particles that are CCNinactive at S = 0.66 %.As can be seen, indeed the whole size distribution of inactive particles at 22:46 LT and 01:21 LT are similar to each other.This supports the argument that the change of CCN properties was not caused, e.g., by a different air mass, but that the CCN-active particles were activated, grew to cloud droplets and were then removed by precipitation, whereas the CCN-inactive particles remained unaffected by the precipitation formation.Until 4 a.m. the particle concentration in the Aitken mode decreased even more leading to a total CN number concentration of ∼ 350 cm −3 (Fig. 9a).MAF increased very little and the hygroscopicity of the CCN-active particles (κ a ) became slightly higher (∼ 0.2) (Fig. 9e).The total CCN efficiency remained very low at ∼ 0.2-0.3 (Fig. 9a).
Until noon of 10 March the aerosol and CCN properties reached their campaign average behavior again (cf.Table 2): the total CN number concentration was around 380 cm −3 , the total CCN efficiency reached ∼ 0.6 again, and the CCN efficiency spectra went up to MAF f ≈ 1).The particle hygroscopicity exhibited a value of κ a ≈ 0.3, which was even a little bit higher than the average value at this supersaturation.The number concentration of CCN-inactive particles, however, decreased and the peak of their number size distribution shifted towards smaller diameters (Fig. A3).
Summarizing, the evolution of the parameters shown in Fig. 9 presents an example of the development of CCN properties during a cloud and precipitation event: most of the accumulation particles (here ∼ 85 %) and a part of the Aitken particles (here ∼ 30 %) are activated into cloud droplets.Even if they are not very hygroscopic like in this case Introduction

Conclusions References
Tables Figures

Back Close
Full they are CCN-active because of their size (Kelvin effect).Through precipitation they are removed from the air, so that only those particles remain that have a significantly lower than average hygroscopicity (i.e., externally mixed weakly CCN-active soot particles) or that have an activation diameter that is smaller than the critical activation diameter particles would have at the ambient supersaturation present in the cloud.The remaining "clean" air (mostly consisting of particles that were CCN-inactive before) will then age by condensation of precursor gases onto the existing particles, coagulation, or chemical transformation, thus resulting in increasing number concentration of accumulation particles, increasing fraction of CCN-active particles and increasing particle hygroscopicity in general.

Summary and conclusions
As part of the CLACE-6 campaign we performed size-resolved CCN measurements at the high-alpine research station Jungfraujoch in March 2007.The dry activation diameters measured at S = 0.079 % to S = 0.66 % were in the range of ∼ 200 to 40 nm and the effective hygroscopicity parameters κ a varied in the range of ∼ 0.1 to 0.6.The mean value of κ a characterizing the hygroscopicity of all aerosol particles averaged over the whole campaign and investigated size range was 0.3, which is slightly higher than the average particle hygroscopicity observed during a 17 month period at the JFJ by Jurányi et al. (2011).It is similar to the average hygroscopicity inferred for other continental locations (e.g. Rose et al., 2010Rose et al., , 2011;;Chang et al., 2010;Shinozuka et al., 2009;Andreae and Rosenfeld, 2008;Pöschl et al., 2009;Pringle et al., 2010).Particles in the Aitken size range were on average less hygroscopic (0.23-0.30) than particles in the accumulation size range (0.32-0.43).During a two-day period, when air masses came from easterly directions crossing the Po Valley in Italy, the particles were much more hygroscopic (κ a ≈ 0.34 to 0.52 in size range from ∼ 50 to ∼ 160 nm).This increase in κ a could be explained by large inorganic mass fractions probably due to sulfate or Introduction

Conclusions References
Tables Figures

Back Close
Full SO 2 emissions from the densely populated and highly industrialized area of the Po Valley.
The observed total CN and CCN concentrations (N CN,tot ≈ 400 cm −3 ; N CCN,tot,m ≈ 50-250 cm −3 for S = 0.079-0.66%) are similar to other measurements at the JFJ (Jurányi et al., 2011).They are about the same as in other remote clean regions (e.g., Amazon Basin), and about one to two orders of magnitude smaller than in polluted regions (e.g., European and Chinese megacities, respectively).Diurnal cycles could not be observed for any CCN parameter due to the lack of periodical injections from the PBL, which is typical for this high altitude site at winter time.
Particle mass concentrations measured with an AMS in the size range of D va = 40-1000 nm were on average ∼ 0.5 µg m −3 .The time series of the derived organic mass fraction (f org ) correlated well with the κ a values obtained at S = 0.079 %: κ a decreased with increasing f org .Extrapolation of the linear correlation between κ a and f org yielded κ org = 0.18 and κ inorg = 0.63 for the hygroscopicity of the organic and inorganic components, respectively.The value for κ inorg is in accordance with earlier studies relating κ a to AMS derived mass fractions (e.g. Rose et al., 2011;Gunthe et al., 2011Gunthe et al., , 2009)), while κ org is slightly higher.This is likely due to longer aging and higher oxygenation of the particulate organic matter sampled in the free tropospheric air sampled at the Jungfraujoch compared to the boundary layer air sampled in the preceding studies in rainforest (Gunthe et al., 2009) and megacity environments (Rose et al., 2011;Gunthe et al., 2011).The average size distribution of organic and inorganic mass fractions as obtained from AMS data showed higher proportions of inorganic matter with increasing particle size, which correlated well with the observed increased hygroscopicity of the aerosol particles.
In the night of 9/10 March we observed a precipitation event at the JFJ site.Our aerosol and CCN measurements showed that almost all accumulation mode particles and a part of the Aitken mode particles were activated as cloud droplets and then removed from the air by precipitation.The number concentration of externally mixed weakly CCN-active soot particles remained constant before and after precipitation, Introduction

Conclusions References
Tables Figures

Back Close
Full which confirms that neither they were activated to cloud droplets nor they were washed out by the precipitation.
During the CLACE-6 campaign we had the opportunity to directly compare two DMT CCN instruments for a certain time.When they both measured in polydisperse mode, i.e., directly determining the total CCN number concentration, the results agreed very well (∼ 6 % deviation) as long as the effective supersaturation was similar.Big differences in the supersaturation of the two instruments (e.g., 20 %), however, led to enormous differences in N CCN,tot,p (e.g., 50 %), which emphasizes that it is extremely important to carefully calibrate the supersaturation of the instrument, especially at low S. When one CCNC was operated in polydisperse mode and the other one measuring size-resolved activated fractions from which the total CCN concentration can be calculated, the differences in the total CCN concentration were on average 20 to 25 %.This inconsistency can be most likely attributed to a long sampling line and the inappropriate correction for diffusional losses.
Therefore we conclude that monodisperse CCN measurements are better for determining the hygroscopicity parameter accurately (additionally providing information on the mixing state), whereas they are less suitable for the measurement of total CCN concentrations, as the uncertainty is increased by the uncertainty of the SMPS measurement (Jurányi et al., 2013).Polydisperse CCN measurements, in contrast, provide more accurate total CCN concentrations, whereas inferred hygroscopicity parameters are tainted by the additional uncertainty of the SMPS.Figures

Conclusions References
Tables Figures

Back Close
Full  Full  Full

Longitude [°]
03.03.00:00 03.03.12:00 04.03.00:00 04.03.12:00 05.03.00:00 05.03.12:00 06.03.00:00 06.03.12:00 07.03.00:00 07.03.12:00 08.03.00:00 08.03.12:00 09.03.00:00 09.03.12:00 10.03.00:00 10.03.12:00 11.03.00:00 11.03.12:00 12.03.00:00 12.03.12:00 13.03.00:00 13.03.12:00 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | were performed over the period of 3-13 March 2007 at the Global Atmosphere Watch (GAW) laboratory of the high-alpine research station Jungfraujoch (JFJ hereafter, 3580 m a.s.l., 46.548 • N, 7.984 • E, see Fig. 1) in Switzerland and were part of the "CLoud and Aerosol Characterization Experiment" in 2007 (CLACE-6).Because of the station elevation the JFJ is considered to be prevalently in the free troposphere and therefore suited for the measurement of background aerosol (Nyeki et al., Discussion Paper | Discussion Paper | Discussion Paper | From 8-11 March air masses came mostly from northerly directions (over Germany and the Netherlands), and in the end (11-13 March) air masses came from easterly directions crossing northern Italy (Po Valley) and the northern Balkan region.Discussion Paper | Discussion Paper | Discussion Paper | operated by the Max Planck Institute for Chemistry and the Paul Scherrer Institute (MPI CCNC and PSI CCNC, respectively) were measuring the activation of aerosol particles to cloud droplets at different water vapor supersaturations.During the Discussion Paper | Discussion Paper | Discussion Paper | 3 K) corresponding to S values in the range of 0.079-0.66% (MPI CCNC).For each ∆T and S, respectively, the diameter of the dry aerosol particles (D) selected by the DMA was set to 6 different values in the range of 20-250 nm depending on the 32582 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | impact, and the ions are analyzed by a high resolution time-of-flight mass spectrometer.Calibration of peak intensities with test aerosol (ammonium nitrate) allows for conversion of the measured peak height into quantitative mass loadings.Detected aerosol species are particulate organic matter (Org), as well as particulate sulfate ( Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | pendent of S. The high bias can be most likely explained by the high uncertainty that is involved in the calculation of N CCN,tot,m by integrating the CCN number size distribution.It is known that already the determination of the total particle number concentration by integrating the particle number size distribution measured by SMPS (dN CN /d log D) has an uncertainty of 10-20 % compared to direct measurements by a CPC (Wiedensohler et al., 2012).Another reason for the discrepancy between N CCN,tot,m (MPI) and N CCN,tot,p (PSI) might be that the diffusion correction of the particle number size distribution along the long sampling line was not sufficient.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

(
European megacity, Paris: Jurányi et al., 2013); and about two orders of magnitude smaller than in very polluted regions (Chinese megacities: Rose et al., 2010; Gunthe et al., 2011).Discussion Paper | Discussion Paper | Discussion Paper | 3.4.The comparison with data of size-resolved CCN measurements from an earlier campaign at the Jungfraujoch (CLACE-5; February/March 2006; unpublished data) showed similar ranges for the hygroscopicity parameter κ a (yellow data points in Fig.4).In the investigated diameter range of ∼ 40-80 nm κ a increased from 0.26 to 0.37.This is ∼ 20 % higher than the average values, but well within the range of temporal fluctuations observed during the CLACE-6 campaign.It is still smaller than the average κ a values observed during a polluted air flow from easterly directions during the measurements in 2007 (red data points in Fig. 4; more details in Sect.3.3).Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | The service charges for this open access publication have been covered by the Max Planck Society.Discussion Paper | Discussion Paper | Discussion Paper | Wu, Z. J., Poulain, L., Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Table 1 .
Comparison between the total CCN concentration measured with the MPI CCNC and the PSI CCNC for different operation modes (Fig.2).The abbreviation "poly" means the operation mode when the total CCN concentration was recorded directly by the CCNC (polydisperse mode) and "mono" means the operation mode when the total CCN concentration was calculated from the size-resolved CCN efficiency spectrum (monodisperse mode).The table lists the average total CCN concentrations measured by the two CCNCs for each comparison, the average relative deviation between the concentrations of the MPI and the PSI CCNC, and the number of data points, n.The first two lines display the supersaturation levels of the PSI and the MPI CCNC and the third line is the relative deviation of the supersaturation of the MPI CCNC from the one of the PSI CCNC.

Table 2 .
Characteristic CCN parameters (arithmetic mean value ± one standard deviation) averaged over the entire CCN measurement period of the CLACE-6 campaign (3-13 March 2007), and over other indicated periods: midpoint activation diameter (D a ), maximum activated fraction (MAF), CDF standard deviation (σ a ), heterogeneity parameter (σ a /D a ), hygroscopicity parameter (κ a ), total CN concentration (N CN,tot ), total CCN concentration (N CCN,tot,m ), total CCN efficiency (N CCN,tot,m /N CN,tot ), and available number of data points of CCN efficiency spectra (n es ) and of size distributions (n sd ).