the Creative Commons Attribution 3.0 License. Atmospheric Chemistry and Physics

Data from measurements of hygroscopic growth of submicrometer aerosol with a hygroscopicity tandem dif- ferential mobility analyzer (HTDMA) during four campaigns at the high alpine research station Jungfraujoch, Switzerland, are presented. The campaigns took place during the years 2000, 2002, 2004 and 2005, each lasting approximately one month. Hygroscopic growth factors (GF, i.e. the relative change in particle diameter from dry diameter, D0, to diam- eter measured at higher relative humidity, RH) are presented for three distinct air mass types, namely for: 1) free tropo- spheric winter conditions, 2) planetary boundary layer influ- enced air masses (during a summer period) and 3) Saharan dust events (SDE). The GF values at 85% RH (D0=100 nm) were 1.40±0.11 and 1.29±0.08 for the first two situations while for SDE a bimodal GF distribution was often found. No phase changes were observed when the RH was varied between 10-90%, and the continuous water uptake could be well described with a single-parameter empirical model. The frequency distributions of the average hygroscopic growth factors and the width of the retrieved growth factor distri- butions (indicating whether the aerosol is internally or exter- nally mixed) are presented, which can be used for modeling purposes. Measurements of size resolved chemical composition were performed with an aerosol mass spectrometer in par- allel to the GF measurements. This made it possible to esti- mate the apparent ensemble mean GF of the organics (GForg) using inverse ZSR (Zdanovskii-Stokes-Robinson) modeling. GForg was found to be 1.20 at aw=0.85, which is at the up- per end of previous laboratory and field data though still in agreement with the highly aged and oxidized nature of the Jungfraujoch aerosol.


Introduction
Aerosol particles in the atmosphere affect the earth's radiation balance in various ways (e.g. Solomon et al., 2007). Firstly, aerosol particles absorb and scatter radiation. This direct aerosol effect is influenced by the hygroscopicity of the aerosol particles, which turn influences the chemical composition. These processes are commonly referred to as the ageing of aerosols.
Aerosols and their properties, such as hygroscopicity, are currently modeled in global climate models (GCMs), mostly to better predict the scattering properties and size distribution under varying humidity conditions (Randall et al., 2007). Relatively few mea- 10 surements of background aerosol from the lower free troposphere exist (e.g. Kandler and Schütz, 2007). To increase available data and validation possibilities four measurement campaigns at the high alpine site Jungfraujoch (JFJ), with a duration of about one month each, are presented here. During 2000During , 2002During , 2004During and 2005 the CLACE (CLoud and Aerosol Characterization Experiment) field studies were performed within 15 international collaborations, including both summer and winter seasons. The general goals of the field campaigns were i) a physical, chemical, and optical characterization of the aerosol at the JFJ in order to better quantify the direct aerosol effect, and ii) an investigation of the interaction of aerosol with clouds, for a better quantification of the indirect effect. The cloud forming processes were studied under different meteorological 20 conditions, with a special focus on aerosol-cloud partitioning in mixed-phase clouds (Cozic et al., 2007a;Verheggen et al., 2007). Further topics were the physical and chemical characterization of ice nuclei (Cozic et al., 2007b 1 ; Mertes et al., 2007), and the processes responsible for the formation of new particles in the free troposphere. Instrumentation was deployed to characterize the aerosol size distribution (scanning 25 particle mobility sizer and optical particle counter), size segregated chemical compo-Introduction EGU sition (Aerodyne aerosol mass spectrometer, AMS) and hygroscopicity (hygroscopicity tandem differential mobility analyzer, HTDMA). In this study AMS and HTDMA results will be analyzed in greater detail. Atmospheric aerosol components can be classified into inorganic and organic fractions (e.g. Kanakidou et al., 2005). The hygroscopic properties of most inorganic salts 5 present in the atmospheric aerosol are known. Of the many organic species identified in the aerosol (e.g. Putaud et al., 2004), the hygroscopic properties of quite a few substances have been investigated. Inorganic salts (for instance ammonium sulfate (AS) and sodium chloride (NaCl) can show a hysteresis behavior during uptake and loss of water, i.e. by exhibiting a difference between the deliquescence and efflorescence rel- 10 ative humidities (DRH/ERH), and with a higher water content of the deliquesced than the effloresced particles in this relative humidity (RH) range. Conversely, organic constituents of the aerosol often do not show efflorescence which can contribute to an uptake of water at lower RH than the DRH of inorganic salts.
A method for characterizing water uptake is the HTDMA (Liu et al., 1978;Rader and 15 McMurry, 1986;Weingartner et al., 2002). The set-up used in three of the campaigns was a low-temperature HTDMA (−10 • C during the winter campaigns and 0.5 • C during the summer campaign), and in the winter campaign 2005 measurements were done at laboratory temperature (25)(26)(27)(28)(29)(30)(31)(32)(33) • C). Furthermore, measurements with an AMS supplied time-and mass-resolved chemical composition of sulfate, nitrate, ammonium and 20 organics during the campaigns 2002, 2004 and 2005. The hygroscopic growth was predicted with the Zdanovskii-Stokes-Robinson (ZSR) relation using the measured composition from the AMS Stokes and Robinson, 1966), and compared with the hygroscopicity measured by the HTDMA. 13702 1998). During winter it is predominantly in lower free tropospheric air masses. During summer the aerosol sampled is influenced by injections of air from the planetary boundary layer (PBL) (Baltensperger et al., 1997;Nyeki et al., 2000). The station is surrounded by glaciers and rocks, and no local vegetation is present. The JFJ boasts the highest European (electrical) railway station and is easily accessible throughout the year. Within the World Meteorological Organization (WMO) Global Atmosphere Watch (GAW) program continuous measurements of aerosol parameters have been performed at the JFJ site since 1995 (Collaud Coen et al., 2007). The research station is also part of the Swiss National Monitoring Network for Air Pollution (NABEL) and the Swiss Meteorological Institute (SMI).

15
The aerosol loading at the JFJ shows an annual cycle with highest concentrations in August to July and minimum concentrations in January to February (e.g. Cozic et al., 2007c;Nyeki et al., 1998;Weingartner et al., 1999). Based on comparison with the continuous aerosol measurements that are available for the JFJ since 1995, campaigns in 2000 and 2004 appear as typical winter conditions (with low aerosol con-20 centration present in the free troposphere), while the campaign in 2002 is typical of summer conditions, and the 2005 campaign can be considered as spring-like conditions, with features situated in between winter and summer. In the following, the data are separated accordingly to these cases: non-disturbed lower free tropospheric winter conditions (abbreviated FT) and PBL influenced summer conditions (abbreviated 25 PBL INF). Further, at times the JFJ is influenced by Saharan dust events (SDE). These events were detected according to the method described by Collaud Coen et al. (2004) EGU is negative. This method was also corroborated by size resolved chemical analysis by ion chromatography, where, during dust events, ∼6% of the total calcium concentration was found in the PM 1 samples. Thus, a third type of air mass (abbreviated SDE) is distinguished here. The criterion used for SDE was when the Angström exponent was less than −0.1 for more than three hours. Conversely non-disturbed FT air masses were 5 defined as the periods where the Angstrom exponent was positive, and furthermore 1 h around each SDE was removed from the data to avoid transition periods.

Measurements
Several different inlets were used during the experiments. In this study a heated total inlet (25 • C) was used, which was designed to evaporate the condensed water from 10 cloud hydro-meteors thus sampling the sum of all particles including both cloud droplet residual and interstitial particles. Calculations for this setup showed that cloud droplets smaller than 40 µm can be sampled at wind speeds up to 20 m s −1 (Weingartner et al., 1999). An interstitial inlet was operated with a PM 1 or PM 2.5 cyclone and sampled only the interstitial submicron-sized aerosol, with hydro-meteors being precipitated in 15 the cyclone. The difference in response downstream of the two different inlets provides insight into the fractionation of aerosol particles between the cloud phase and the interstitial phase. Table 1 lists the dates for the campaigns as well as details of instruments used and setup of the HTDMA. 20 An Aerodyne quadruple AMS (Jayne et al., 2000) was used to provide on-line, quantitative measurements of the total mass and size distributed non refractory chemical composition of the submicron ambient aerosol at a high temporal resolution. The instrument works by sampling air through an aerodynamic lens to form a particle beam in a vacuum and accelerating the focused beam of particles as a function of their mo-25 mentum towards a tungsten heater (550 • C) that flash vaporizes the particles. The by scanning the m/z spectrum with the quadruple mass spectrometer, without size resolved information, (ii) using the aerosol time-of-flight (ToF) mode selected m/z representative of key chemical components can be resolved as a function of the vacuum aerodynamic diameter of the particles. More detailed descriptions of the AMS measurement principles and various calibrations (Canagaratna et al., 2007;Jayne et al., 10 2000), its modes of operation  and data processing and analysis (Allan et al., 2003;Allan et al., 2004) are available. The AMS supplies the concentrations of inorganic ions, i.e. sulfate, nitrate and ammonium. These ions account for 96% of the composition of inorganic ions at the JFJ (Cozic et al., 2007c;Henning et al., 2003;Krivacsy et al., 2001). Furthermore the total concentration of the organic content 15 is supplied, although no detailed speciation is possible. Mass loadings at the JFJ site are generally low. Therefore 3-h averages were calculated as a compromise between counting statistics and time resolution.

Black carbon concentration
During the first two campaigns 2000 and 2002 the black carbon (BC) concentration 20 was measured with an AE31 Aethalometer (at wavelength λ=880 nm) (Weingartner et al., 2003). During the last two campaigns BC was measured with a multiple angle absorption photometer (MAAP, at λ=630 nm) as well as with an AE31 Aethalometer (at λ=880 nm). A mass absorption efficiency of 7.6 m 2 g −1 for winter and 11.1 m 2 g −1 for summer was used for the MAAP data (Cozic et al., 2007c). BC concentrations from the 25 aethalometer were determined accordingly, taking advantage of a high correlation between these two instruments during simultaneous measurements (Cozic et al., 2006 EGU fraction in each size range was independent of size and thus equal to the BC fraction in PM1 (defined as the sum of the AMS and the BC data). The choice of the mass absorption efficiency for the BC concentration and the assumption of size independence are not critical due to the low sensitivity of the hygroscopicity closure to these values.
2.5 HTDMA (Hygroscopicity Tandem Differential Mobility Analyzer) 5 Briefly, the HTDMA functions as follows: a differential mobility analyzer (DMA1) selects a monodisperse aerosol size cut with mobility diameter, D 0 , under dry conditions. The aerosol then passes through a humidifier with a controlled higher RH, and the mobility diameter D is measured with a second DMA (DMA2). The two DMAs are similar to the TSI 3071 type. The relevant RH in DMA2 was determined by measurement of the system temperature and the DMA2 excess sheath air dew point using a dew point mirror (model 2002 Dewprime, EdgeTech). The accuracy of the RH measurement at higher RH is for example 85±1.1%, assuming no temperature gradients in the DMA2. The residence time of the sampled aerosol at the set RH was >20 s before size measurement (Sjogren et al., 2007). The HTDMAs were employed in slightly different ways 15 during the different campaigns (see Table 1). In general the HTDMA measured at a constant RH which was set to 85%. On occasions the RH-dependence of the hygroscopic growth was investigated by both increasing and decreasing the RH in DMA2 between 10 and 85%. These are known as the dehydration and hydration modes of operation, respectively. This allows for detecting potential hysteresis effects in the hy-20 groscopic growth behavior with distinct efflorescence and deliquescence transitions. The hydration mode, where the mono-modal dry particles were exposed to a monotonically increasing RH in the HTDMA prior to the size measurement in DMA2 (allowing measurements of DRH), was applied during all campaigns, and is also the mode of operation used during measurements at constant RH. The dehydration mode, where the 25 dry particles are first exposed to RH>80% using a pre-humidifier before monotonically lowering the RH towards the RH in DMA2 (Gysel et al., 2004;Sjogren et al., 2007), was mainly applied during the 2004 winter campaign (allowing measurements of ERH

EGU
The hygroscopic growth factor (GF ) indicates the relative increase in mobility diameter of particles due to water absorption at a certain RH, and is defined as where D(RH) is the mobility diameter at a specific RH and D 0 is the particle mobility diameter measured under dry conditions (91% of the data were with DMA1 at RH<15%.

5
However to increase available data we included periods where the RH in DMA1 was up to 35%, where the GF is relatively low (<1.03) for the ambient aerosol (see below). Mobility diameter growth factors obtained with an HTDMA are only equal to volume equivalent growth factors if the particles do not change their shape during water uptake. This assumption is justified as the hygroscopicity was characterized by a continuous growth curve for the majority of the time (see below), thus the particles can be considered liquid and consequently roughly spherical at all measured RH.
During the first three campaigns both DMAs and the humidifier were inserted in a well-circulated water bath, ensuring constant temperature as indicated. The aerosol line was cooled and insulated from the outside of the building to the entry of the first 15 DMA. This ensured that no artifacts during the sampling occurred (i.e. volatilization of semi-volatile material), as the measurements were performed close to ambient temperatures (Gysel et al., 2002;Weingartner et al., 2002). During the campaign in 2005 the first DMA was maintained at the laboratory temperature (25)(26)(27)(28)(29)(30)(31)(32)(33) • C) and only DMA2 was kept at a constant temperature in a water bath (22.8 • C) (Sjogren et al., 2007). This was 20 done because of the functionality of the HTDMA used, and because it was deduced from the three first campaigns that temperature artifacts were negligible (i.e. compared to measurement uncertainties). The HTDMA data were averaged to 3 h, in order to match the AMS time series. The performance of the instruments was verified with extensive testing with AS and NaCl before the campaigns. During 2004 and 2005 25 these salts were also measured at the JFJ. The growth of AS and NaCl particles was compared with the theoretical prediction using the Aerosol Diameter-Dependent Equilibrium Model (ADDEM) (Topping et al., 2005a(Topping et al., , 2005b, and corresponded to within 13707 Introduction EGU less than 0.04 in GF at 85% RH.

Inversion algorithm
Atmospheric particles of a defined dry size typically exhibit a range of growth factors or even clearly separated growth modes, because of external mixing or variable relative fractions of different compounds in individual particles (hereinafter referred to as 5 quasi-internally mixed). Growth factor probability distributions c(GF)=dC/dGF are retrieved from each measurement, and normalized such that C = c(GF)d GF = 1. The inversion method applied to the raw data (Gysel et al., 2007 2 ) has similarities to the inversion algorithm described by Cubison et al. (2005). The distribution c(GF) is also inverted from the measurement distribution into contributions from fixed classes of nar-10 row growth factor ranges, but instead of using a linear inversion, c(GF ) is fitted to the actual measurements using a full TDMA transfer forward model. A bin resolution of ∆GF =0.15 was chosen for the inversion because of counting statistics.
The AMS provides chemical composition data for the entire submicron aerosol particle ensemble in the air sample, whereas no information on the mixing state of individual 15 particles is obtained. Inverted growth factor distributions c(GF) obtained with the HT-DMA provide some information on the mixing state. The ensemble mean growth factor GF * is defined as the 3rd-moment mean growth factor of c(GF). GF* represents the growth factor that would be observed if the absorbed water were equally distributed among all particles, even in the case of several distinct growth modes. Thus GF* is 20 the quantity to be compared with growth factor predictions based on composition data obtained by the AMS (see below). Thus even if the measured GF is broad or even clearly bimodal GF* would represent the hygroscopicity as predicted from the AMS data as long as the AMS can measure all the relevant chemical components in both modes. This is not the case if some of the material sampled is composed of a refrac-

AMS.
The standard deviation σ of the inverted growth factor distribution c(GF) is used as a measure for the spread of growth factors. With a bin resolution of ∆GF =0.15 as chosen here for the HTDMA data inversion, the σ that would be obtained for a perfectly internally mixed aerosol with a well defined growth factor (i.e. σ=0) is between 0.06 5 and 0.10 depending on the bin positions relative to the GF*. The σ obtained with pure ammonium sulfate at 85% RH is <0.05, when inverting the data with high resolution. Therefore any σ≤0.10 indicates absence of distinct growth modes, i.e. a quasi-internal mixture with limited spread of growth factors, while any σ≥0.15 shows that the aerosol is externally mixed or quasi-internally mixed with substantial spread of growth factors. 10 We use the σ not only to describe the spread of a single mode, but also in the sense of describing a broader distribution, or describing cases which are clearly bimodal. Two HTDMA measurement examples (red) and corresponding inverted growth factor distributions (green) as well as the inverted growth distribution reprocessed through the HTDMA forward model (blue) are shown in Fig. 1. Panel (A) shows an aerosol 15 (D o =100 nm) observed during undisturbed FT conditions with GF*=1.28 and σ=0.08, indicating that it was internally mixed. Panel (B) shows an aerosol observed during an SDE with σ=0.22. This aerosol is obviously externally mixed with two distinct modes at GF =1.05 and GF =1.45, whereof the former can be attributed to mineral dust. The GF* is 1.227 and is not representative of the hygroscopic behavior of the aerosol but would 20 be the value to compare with the predicted hygroscopicity from the ensemble chemical composition, which would need to take mineral dust into account. In this case the blue line does not exactly follow the measurement because of the limited resolution of the inversion. However, better results cannot be obtained even with increasing the resolution, because the measurement uncertainties are too big at such a low number The measurements were generally done at 85% RH. To obtain a more complete time series data set, data between 80 and 90% RH were corrected to 85% RH using the following equation: where k captures all solute properties. a w is the water activity. First the k-value was calculated from the measured GF and RH (left hand side of Eq. 2), and then the corresponding corrected GF at 85% was calculated using this k (right hand side of Eq. (2). Equation (2)  (2) are given in Kreidenweis et al. (2005). Using a constant k-value for RH corrections is equivalent to a constant van't Hoff factor. This 15 assumption is justified for differences of ±5% RH as chosen here.

ZSR relation
The hygroscopic growth factor of a mixture (GF mixed ) can be estimated from the growth factors of the individual components of the aerosol and their respective volume fractions, ε, with the ZSR relation (Gysel et al., 2004;Stokes et al., 1966): where the summation is performed over all compounds present in the particles. The model assumes that: the particles are spherical; ideal mixing behavior (i.e. no volume 13710 EGU change upon mixing); and independent water uptake of the organic and inorganic components. The volume fractions ε ι or the components in the particles were calculated as where w i is the measured mass fraction and ρ i the density of component i .

5
The AMS measured size-resolved component mass concentrations during the last three campaigns : 2002, 2004 and 2005. We compared the fraction of particles in the range 88-196 nm vacuum aerodynamic diameter (d va ) from the AMS with the measured 100 nm mobility diameter (d mob ) from the HTDMA. The range was chosen so that the volume difference was equal on each side of 100 nm, (i.e. 49-122 nm in d mob ). 10 The chosen width of the size range is a compromise between size resolution and signal statistics. The mobility diameter can be calculated from the aerodynamic diameter as follows (Zelenyuk et al., 2006), with the assumption that the particles are spherical (χ v =1, dynamic shape factor in the free molecular regime):

15
Where ρ 0 is standard density (1000 kg m −3 ) and ρ p is the particle density. The particle density was calculated from the bulk composition averaged over the campaigns and was 1565 kg m −3 . In general the hygroscopicity, predicted with ZSR relation, is more sensitive to substances with higher GF than the ones closer to 1.0 due to the cubic weighting. Thus it is more important to accurately predict the hygroscopicity of the 20 pure inorganic salts, than the organic or the soot components. A second factor of importance is the correct determination of the volume fractions of the more hygroscopic substances. All individual growth factors and densities used were taken from literature (see Table 2).  (Cozic et al., 2007c). However, occasionally the measured ammonium concentration was insufficient to fully neutralize the sulfuric acid, in-5 dicating an acidic aerosol. Then it was assumed that an equilibrium with first NH 4 HSO 4 and subsequently H 2 SO 4 was formed. As expected during cases with incomplete neutralization the NO − 3 values were low. Note that this choice of sulfate salts is crucial for a correct application of the ZSR mixing rule, i.e. choosing only a combination of H 2 SO 4 and (NH 4 ) 2 SO 4 to match the measured sulfate and ammonium concentrations 10 would result in significant prediction errors . Splitting the sulfate salts based on the measured ammonium may not be highly accurate, but it indicates at least whether the aerosol is neutralized or acidic. The mass fractions of the different components are shown in the time series in Fig. 4.

Hygroscopicity at the JFJ
The RH-dependence of GF was measured by variation of the RH in the HTDMA between 10 and 85%. Figure 2 shows three typical humidograms examples showing the features observed at the JFJ. Generally a continuous growth without differences between hydration and dehydration operating mode was found, thus indicating absence 20 of phase changes. This does not exclude existence of efflorescence at RH<10%, because our measurements were technically limited to RH>∼10%. Such continuous growth is expected and has been reported for complex mixtures with an increasing number of organic components (Marcolli et al., 2004;Marcolli and Krieger, 2006). The aerosol at the JFJ seems to exist predominantly as dissolved liquid or amorphous par-ticles. Furthermore, the growth curves can be well described with the single-parameter (k) semi-empirical model given in Eq. (2), as can be seen from the solid lines in Fig. 2. The growth curves were also fitted with an empirical power law fit GF =(1-a w ) γ (Swietlicki et al., 2000), dashed lines in Fig. 2, but for this model we found consistently larger χ 2 -residuals than with the former model. As can be seen from Figs. 2, 3 and 4 5 the magnitude of the hygroscopic growth at the JFJ varies substantially over time, but the RH-dependence at any time can be captured with a single parameter (k). Figure 3 shows the temporal evolution of GF distributions for a period of the campaign in 2000. Several SDE were observed, as indicated with the shaded areas in Panel (A). During undisturbed FT conditions, a size dependence of the growth factor can be seen, with larger growth for larger particles. This feature, which has previously been shown by Weingartner et al. (2002), was also observed during the other campaigns and is attributed to a size dependent chemical composition. This is confirmed by larger k-values at larger dry diameters (top panel of Fig. 3), which are a measure of the hygroscopicity without the influence of the Kelvin effect. The ensem-15 ble k-values have been calculated from the ensemble mean growth factor GF * using Eq. (2), whereas the water activity corresponding to the measured RH has been calculated assuming a surface tension of pure water. It is hypothesized that smaller particles contain a larger fraction of organic compounds from secondary organic aerosol (SOA) formation. Such a dependence was confirmed by AMS measurements during summer 20 2002 (Alfarra, 2004). During major SDE two distinct growth modes can be seen for the 250 nm particles, while no clear change in hygroscopic behavior is seen for the 50 nm particles. This is also reflected in strongly decreasing k-values at D 0 =250 nm, while little or no changes occur at D 0 =100 and 50 nm. Thus mineral dust particles are only found at larger sizes. As during SDE larger sized particles are more externally 25 mixed, the reduction of the k-value in these periods only reflects the influence of the increasing amount of particles, presumably mineral dust, with lower hygroscopicity, not the k-value of each mode.
In summer, a strong diurnal variation is typically found in most aerosol variables (Lu-  , 2000). During the 2002 campaign this diurnal variation was also present in the observed mass loadings, though to a slightly lesser extent (Alfarra, 2004). However, the mass fractions of different components did not vary to a large degree during this diurnal variation, resulting in a fairly constant hygroscopicity on a timescale of hours seen both in the HTDMA results and in the closure (Fig. 4). However, more data with 5 days showing a strong diurnal variation are required for a conclusive description of this influence. As can be seen from the GF * values in Table 3, the summer campaign is characterized by lower hygroscopicity, due to a higher organic loading (68% in summer compared to 42% in winter). This is most probably due to higher emission rates of SOA precursors and higher photo-oxidation activity, which can also be transported upwards 10 through valley venting. Figure 4 shows the predicted growth and GF* for the campaigns 2002, 2004 and 2005 (no high time resolution composition data were available for 2000). A growth factor for organic compounds of 1.2 at a w =0.85, corresponding to 1.182 for a 100 nm particle at 85% RH was used in this work, which is in agreement with smog chamber 15 experiments of organic aerosols (Duplissy et al., 2007 3 ;Baltensperger et al., 2005). The sensitivity to this value was tested by comparing the model fit to the data over several values of GF org and 1.2 at a w =0.85 gave the best fit with a slope of 1, though the fit is relatively insensitive as the influence of GF org is relatively low compared to the influence of the inorganic fraction. The mass spectrum delivered by the AMS at 20 the Jungfraujoch is characterized by a relatively low m/z 57 signal indicating that little unprocessed primary organic material is present and that the majority of the organics are composed of oxidized organic mater (Alfarra et al., 2006;Zhang et al., 2007). During SDE there will be an increased fraction of mostly insoluble mineral dust material (Vlasenko et al., 2006) which is not detected in the AMS. This will lead to an increase Recently, Gysel et al. (2006) have found significant discrepancies between measured and predicted GF s if substantial mass fractions of ammonium nitrate were present. They concluded that the most likely cause for the discrepancies was an evaporation artifact of ammonium nitrate in the HTDMA, which was operated at ∼25 • C and with a residence time of ∼60 s. No systematic prediction bias for data points with high am-10 monium nitrate mass fraction was found in the data set presented here. An important difference is that the HTDMA measurements of this study were mostly done at low temperatures (T =−10 to 0.5 • C) and the residence time was kept short (in the order of 20 s), thus minimizing potential evaporation artifacts. EGU with the inversion considering the instrument limits and the low mass loading. For the 50 nm particles this shoulder at low GF s is less pronounced, but this is to some extent a consequence of the smaller hygroscopicity of the main mode, which is slightly overlapping with this shoulder. Kandler and Schütz (2007) also reported GF values for March 2000 at the JFJ (measured at 90% RH at ∼20 • C), which are in agreement with 5 the GF s shown here. Kandler and Schütz (2007) indicated a bimodal distribution for all sizes, however they do not show the relative number fractions in each mode, which, if small for the lower GF mode, would be in agreement with our data. The PBL INF measurements show a more homogeneous GF distribution, but the hygroscopicity is also lower. SDE only occurred during the two winter campaigns, and for these cases  Table 3 has been calculated. The hygroscopicity for summer indicates a similar chemical composition for different 15 sizes, while during winter the hygroscopicity increases with size. Panels C, F and I of Fig. 5 show the frequency distribution of σ averaged for each of the relevant periods. The σ of individual scans can be used to distinguish between quasi-internally mixed aerosols with limited growth factor spread (σ≤0.1) and externally or quasi-internally mixed aerosols with substantial spread (σ≥0.15). The frequency dis-20 tribution of σ thus indicates the fraction of time of each period that a certain mixing state (σ) is encountered. The most frequent spread observed in summer is σ=∼0.125 which is internally mixed, whereas larger spread is seen in winter FT conditions. This can be attributed to a larger separation of the main growth mode from the minor fraction of particles with growth factors <1.2. For the same reason the spread also increases 25 with particle size in winter with σ=∼0.1, 0.125 and 0.15 for D 0 =50, 100 and 250 nm particles, respectively. This indicates that even under FT conditions observed during winter the aerosol contains a fraction of particles which appear to remain less processed and thus less hygroscopic also at a remote location. The mineral dust during 13716 EGU SDE mostly influences the larger particles with D 0 =250 nm, as already exemplified in Figs. 3 and 5A. Here it has to be stressed that different scenarios can end up with a bimodal shape of the mean GF distributions as shown in Fig. 5A. Either the GF distribution is always bimodal with similar number fractions of particles in both modes, or only monomodal but the GF distribution are observed with the mode centered at 5 GF=∼1.0 or GF=∼1.45 during 50% of the time each. Frequent occurrence of σ≥0.2 and GF*=1.3-1.5 for D 0 =250 nm during SDE indicates that the former alternative with simultaneous presence of non-hygroscopic mineral dust and more hygroscopic background particles, both in comparable number fractions, dominates. No clear influence of SDE on σ and GF* is seen at 50 and 100 nm confirming the finding from Panels (D) 10 and (G).

Frequency distributions of GF
The frequency distributions of the GF * and σ can be used to simulate internally or quasi-internally mixed hygroscopic behavior of particles in different air masses encountered at the JFJ. Additionally to the frequency distributions it has to be known whether GF * and σ are dependent on each other. We explored the relationship between the 15 two distributions, but no dependence between GF* and σ was found. This is different from results found by Aklilu and Mozurkewich (2004) in the Lower Fraser Valley, British Colombia, who reported a horseshoe-shaped relationship with maximum σ values at intermediate GF.

20
A statistical analysis of measurements from four field campaigns of about one month each at the high alpine site Jungfraujoch is presented. During the winter season when the station was in the undisturbed free troposphere, the average GF measured with an HTDMA was 1.40±0.11 at 85% RH for D 0 =100 nm particles. During the summer season, due to higher SOA formation, the GF was 1.29±0.08 at 85% RH. During 25 mineral dust events GF distributions were partly bimodal for D 0 =250 nm particles. The frequency distributions of the width of the retrieved growth factor (internally/externally EGU mixed) distributions are presented, which can be used for comparison with simulations of the hygroscopic behavior of the aerosol encountered at the JFJ. The hygroscopicity was also predicted using the ZSR mixing rule along with chemical composition data. The ZSR mixing model can be used to predict the variability of measured hygroscopicity of submicrometer particles. However, due to low loadings at the JFJ (apart from times 5 when influenced by PBL), the spread in error of the predicted GF from the chemical composition as well as the error for the HTDMA measurement is on average ±0.1, which makes it difficult to verify the absolute hygroscopicity values. The most important factor for the modeling is the accuracy in GF s of the inorganics and their composition. It is also important to consider the separation of SO 2− 4 into ammonium sulfate, ammonium tion by wavelength dependence of the single scattering albedo and first climatology analysis, Atmos. Chem. Phys., 4, 2465Phys., 4, -2480Phys., 4, , 2004