Influence of cloud processing on CCN activation behaviour in the Thuringian Forest , Germany during HCCT-2010

Within the framework of the “Hill Cap Cloud Thuringia 2010” (HCCT-2010) international cloud experiment, the influence of cloud processing on the activation properties of ambient aerosol particles was investigated. Particles were probed upwind and downwind of an orographic cap cloud on Mt Schmücke, which is part of a large mountain ridge in Thuringia, Germany. The activation properties of the particles were investigated by means of size-segregated cloud condensation nuclei (CCN) measurements at 3 to 4 different supersaturations. The observed CCN spectra together with the total particle spectra were used to calculate the hygroscopicity parameter κ for the upwind and downwind stations. The upwind and downwind critical diameters and κ values were then compared for defined cloud events (FCE) and noncloud events (NCE). Cloud processing was found to increase the hygroscopicity of the aerosol particles significantly, with an average increase in κ of 50 %. Mass spectrometry analysis and isotopic analysis of the particles suggest that the observed increase in the hygroscopicity of the cloud-processed particles is due to an enrichment of sulfate and possibly also nitrate in the particle phase.


Introduction
Clouds are a key parameter in climate change prediction, due to their strong impact on the radiation processes in the atmosphere.However, the effect of aerosol particles on cloud formation, cloud glaciation and precipitation is still insufficiently quantified and remains therefore one of the largest uncertainties in climate change predictions (Lohmann and Feichter, 2005;IPCC, 2013).Studying the interaction of aerosol and clouds under natural conditions is challenging due to the height as well as the spatial and temporal variability of clouds.Aside from extensive airplane, balloon or helicopter investigations of natural clouds (e.g., Krämer et al., 2013, p. 337), a well-established method for cloud investigation is to take advantage of natural clouds reaching the ground, allowing for groundbased experiments.This experimental design allows e.g the investigation of fog formation (e.g. Po valley fog experiments, Svenningsson et al., 1992), advected stratiform clouds (e.g.Kleiner Feldberg, Wobrock et al., 1994) and also orographically triggered clouds (e.g.Schmücke, Herrmann et al., 2005).
One variation of the ground-based cloud experiments is a Lagrangian-type design.This involves locating several measurement stations in the predominant flow directions above a ridge, in order to probe the air masses before, during and after cloud passage (e.g., Bower et al., 1999;Herrmann et al., 2005).The use of hill cap clouds as a natural flow-through reactor was for example successfully realized on Great Dun Fell (UK) (Bower et al., 1999;Choularton et al., 1997;Svenningsson et al., 1997), on the German mountain Kleiner Feldberg (Wobrock et al., 1994), during the ACE-2 HILL-CLOUD experiment at Teneriffe, Spain (Bower et al., 2000) and during FEBUKO at Mt. Schmücke (Herrmann et al., 2005).
The findings here presented, were observed during the cloud experiment "Hill Cap Cloud Thuringia 2010" (HCCT-2010), which was also conducted at the low mountain ridge around Mt. Schmücke, where FEBUKO (Herrmann et al., 2005) took place.HCCT-2010 took place in September/October 2010 and dealt with several aspects of cloud microphysics and chemistry.An overview of HCCT-2010 is given in a companion paper in this special issue.
During the previous FEBUKO campaign, particle number size distribution measurements upwind (ambient) and on the summit (in-cloud: interstitial and residuals) were used to investigate the dependence of the scavenged aerosol fraction on the soluble volume fraction of the observed particles (Mertes et al., 2005a).In addition, at the upwind site the hygroscopic properties were investigated (Lehmann et al., 2005) using Hygroscopic Tandem Differential Mobility Analyzer (HTDMA) measurements.Promising results were achieved concerning the dependence of the scavenged aerosol fraction on the soluble volume fraction of the particles.However, directly comparable activation or hygroscopicity measurements before and after the cloud passage were not carried out during FEBUKO.Aerosol processing was investigated by model simulation (Tilgner et al., 2005) and by comparing number size distribution upwind and downwind (Mertes et al., 2005b).Both studies show an effect on the aerosol size distribution in the size range of the activation diameter and an increase in aerosol number and mass.
The focus of the work presented here was to investigate the influence of cloud processing on the activation properties of aerosol particles.The interaction of particles with water can be theoretically described via Köhler theory (Köhler, 1936), which gives the equilibrium vapor pressure over an aqueous solution droplet.The maximum of the Köhler curve gives the critical supersaturation necessary for droplet activation.Classical Köhler theory needs several input parameters (e.g., molar weight of the particle substance, surface tension of mixture) which are usually unknown for atmospheric particles.Therefore, one-parameter approximations were developed (e.g., Petters and Kreidenweis, 2007;Wex et al., 2007) which are applicable for the description of particle hygroscopic growth as well activation properties.One-parameter approaches are also well suited as a simple measure to implement particles' activation behaviour in modeling studies (e.g., Pringle et al., 2009).
The results presented in this study are based on size-segregated CCN measurements at the up-and downwind station during periods of connected flow.The hygroscopicity parameter κ was deduced from the derived activation diameters.For selected non-precipitating cloud events on Mt.Schmücke the droplet activation properties at the upwind and downwind valley station were compared, and the statistical significance of the findings was tested, in order to measure the influence of cloud passage on particle hygroscopicity.The same comparison between up-and downwind station was also done for defined cloud-free periods as a control experiment.

Experimental design and setup
The experiments were conducted as part of the HCCT-2010 campaign, a Lagrangian-style experiment in which air parcels were probed at several locations during passage through an orographic cloud, focussing on the influence of cloud presence on the physical and chemical properties of the air parcel of interest (see companion paper for details).Briefly, measurements of meteorological parameters and physical and chemical aerosol and gas properties were conducted at three sites along the mountain ridge of the Thuringian forrest, Germany: one upwind station, the in-cloud mountain peak station on Mt.Schmücke and one downwind station (Fig. 1).Time periods with optimal connected flow between the three stations were chosen by applying several different methods, e.g.shape of particle number size distribution, wind direction, wind speed and ozone concentration (Tilgner et al., 2014).These connected flow regimes were subdivided into periods with a cap cloud present on Mt.Schmücke and the valley sites cloud free, so-called Full Cloud Events (FCE, cf.Table 1), and cloud-free periods at all three stations, called Non Cloud Events (NCE).The FCE and NCE events with CCN measurements available at Goldlauter and Gehlberg station are listed in Table 1 together with liquid water content (LWC), wind direction (wd) and wind speed (ws) on Mt.Schmücke.LWC was measured applying an Particulate Volume Monitor (PVM-100, Gerber Scientific Inc., Reston VA Gerber, 1991)).By coincidence, the FCE time periods with CCN data available at both valley stations had an approaching flow from south-western direction while for the NCE cases the flow was approaching from north-eastern direction.This was taken into account in the data analysis and does not significantly affect the findings.

Measurement sites
The size-segregated CCN measurements took place at the up-and downwind sites either side of Mt.Schmücke, Gehlberg (GB) and Goldlauter (GL).
The valley station Goldlauter (GL, 50 • 38 ′ 14 ′′ N, 10 • 45 ′ 13 ′′ E), is situated on the southwestern slope of the Thuringian forest mountain ridge.For FCE (southwest wind) this was the upwind station, i.e. the air parcel was probed here before entering the cap cloud and therefore represents the "pre-cloud" status of the aerosol.For NE NCE (northeast wind) Goldlauter was the downwind station.All measurement equipment was placed inside an air-conditioned container.On top of the container a PM 10 inlet followed by a self-regenerating diffusion drier was placed (Tuch et al., 2009), maintaining the relative humidity of the aerosol flow below 20 %.Inside the containerbeside other instrumentation -a Mobility Particle Size Spectrometer (MPSS-type TROPOS; details of this instrument in Wiedensohler et al., 2012) was used to determine the particle number size distributions between 10 and 850 nm, and a Cloud Condensation Nucleus counter (CCNc, CCN-100, DMT Boulder, Roberts and Nenes, 2005) in combination with a Differential Mobility Analyser (DMA) was used to measure CCN distributions between 25 and 300 nm.
The measurement equipment at the FCE downwind station Gehlberg (GB, 50 was also placed inside an air conditioned container.Here the PM 10 inlet was followed by individual drier systems in front of the instruments instead of the self-regenerating diffusion drier applied in GL.This was the only difference in the size-segregated CCN measurement set-up between both stations.Nafion driers (30 cm, TROPOS-custom-made) have been placed in front of the TROPOS-type mobility particle size spectrometer and the DMA-CCNc, both of which kept the RH stable below 20 %.The data were corrected for the individual particle losses due to tubing and driers.
At these two sites and at Mt. Schmücke (SM, 50 • 39 ′ 17 ′′ N, 10 • 46 ′ 30 ′′ E, 916 m a.s.l.) sizeresolved (coarse and fine) particulate matter was collected for sulfur isotope analysis, with the in-cloud particulate separated into cloud droplet residual and interstitial fractions.In addition, at the upwind and downwind stations, SO 2 and H 2 SO 4 gas and ultrafine particulate matter were collected.Combined scanning electron microscopy (SEM) and NanoSIMS measurements were used to determine the isotopic composition of particulate sulfur samples (δ 34 S fractionation factors) of the samples.Stable sulfur isotopes fractionate during reactions, so the isotopic composition of a product is not equal to the isotopic composition of the reactant.Using previous measurements of sulfur isotope fractionation factors characteristic for different oxidation pathways (e.g.oxidation by OH, H 2 O 2 or transition metal ion catalysis, (Harris et al., 2012(Harris et al., , 2013a)), the isotopic analyses made during HCCT-2010 allow dominant sulfate production pathways to be determined and resolved for different particle types, as described in Harris et al. (2014).

Size-segregated CCN measurements
The set-up for the size segregated activation measurements was identical at the up-and downwind stations (Fig. 2), apart from the different drier types (cf.above).Downstream of the aerosol inlet and the drier unit, the 1 L min −1 aerosol flow passed through a neutralizer to achieve the bipolar charge equilibrium (Wiedensohler, 1988).The DMAs ran with an aerosol to sheath air flow of 1/10 to size-select aerosol particles based on electrical mobility, to achieve a quasi-monodisperse aerosol distribution.Multiply-charged particles with larger sizes were also selected, for which the size and activation scans had to be corrected, using the bipolar charge distribution.Downstream of the DMA a flow of 0.5 L min −1 particle-free air was added to the aerosol flow and the total flow was divided between to a particle counter (1 L min −1 working flow, CPC 3010, TSI Aachen Germany) and a cloud condensation nucleus counter (0.5 L min −1 working flow, CCNc, CCN-100, Boulder, USA).Measurements at Goldlauter station were taken from 11 September 2010 to 20 October 2010 and in Gehlberg from 12 September 2010 to 20 October 2010, which is slightly shorter than the duration of the whole HCCT-2010 campaign.
The CCNc, a stream-wise thermal gradient cloud condensation nucleus counter (Roberts and Nenes, 2005), was applied to investigate supersaturation-dependent activation of the particles.In this instrument the inlet flow is split into a particle-free sheath air flow, which is kept particle-free via a filter, and an aerosol flow.The sheath air is humidified before entering the flow tube and surrounds the aerosol at the centerline.The stream-wise temperature gradient applied in the flow tube determines the supersaturation to which the particles are exposed.The number of activated particles (N CCN ) is detected at the end of the flow tube with an optical counter.The ratio between the CCN number and the total particle number (N ) gives the activated fraction (AF) of the particles.
The CCNc was used to measure diameter scans for which the saturation is fixed and the dry particle diameter is varied.In this study we ran diameter scans for four fixed supersaturations (0.07, 0.1, 0.2, 0.4 %).The critical particle diameter D c , the diameter at which 50 % of the particles are activated at a particular supersaturation, is derived from such a diameter scan.
The supersaturation reached in the CCNc during the size-segregated CCN measurements was calibrated with ammonium sulfate particles.This was done by atomizing an ammonium sulfatewater solution (0.1 g (NH 4 ) 2 SO 4 (300 mL H 2 O) −1 ), passing the resulting aerosol through a diffusion dryer, and injecting the dried particles into the CCN measurement set-up (Fig. 2).The calibration procedure followed that described in Rose et al. (2008).In short, diameter scans were run at nominal supersaturations between 0.07 and 0.7 %, which relate to a certain temperature gradient in the flow tube of the CCNc.The AF were fitted applying a Gaussian error function to the data: where a and b are the upper and lower limit, to calculate critical diameters D c at the set nominal supersaturations.As (NH 4 ) 2 SO 4 particles were used, the activation diameter is known and the set temperature gradient in the instrument can be related to the effective supersaturation SS reached in the column.Repeated calibrations show an achievable accuracy in SS of 10 % (relative) at supersaturations above SS = 0.2 % and δSS ≤ 0.02 % (absolute) at lower supersaturations (Gysel and Stratmann, 2013).
In the work presented here, we apply the single parameter κ Köhler theory (Petters and Kreidenweis, 2007) to describe the hygroscopicity of the ambient particles.The hygroscopicity parameter κ is calculated in the following way (from Petters and Kreidenweis, 2007): with The critical diameter is determined by fitting the AF scans to the error function (Eq.1), analogously to the calibration procedure.The calibrated SS and measured D c are inserted into into Eq.( 2) to calculate κ at fixed supersaturations for ambient particles.The error in SS setpoint, especially at low supersaturations, results in quite a large level of uncertainty in κ values.A critical diameter D c of e.g.200 nm at a SS of 0.07 % ± 0.02 % corresponds to a κ range from 0.21 (SS = 0.09 %) < 0.35 < 0.94 (SS = 0.05 %).This is a critical point in working with κ and has to be considered in the interpretation.This will be discussed more in Sect.3.2.

Particle number size distribution measurements
In parallel to the CCN spectra, the particle number size distribution in the size range between 10 and 850 nm have been measured in Goldlauter and Gehlberg.The measurements were done with the above mentioned mobility particle size spectrometers, which were connected to the same PM 10 inlet as described above for the CCN measurements.Particle losses due to diffusion in the instrument and in the sampling lines have been corrected according to the method of "equivalent length" as described in Wiedensohler et al. (2012).
The CCN spectra have to be corrected for multiply-charged particles as the fitting of the AF with the error function is influenced by the appearance of a second step in the CCN spectra (Rose et al., 2008).This step is triggered by the fact that multiply-charged large particles have the same electrical mobility diameter as singly-charged smaller particles, and are thus falsely selected in the DMA.
In the CCNc, however, they activate at a lower supersaturation than the singly-charged particles, and appear in the activated fraction vs. particle diameter curve as a first activation step at smaller diameters.How pronounced this first step is depends on the particle number size distribution, especially on the number of larger particles.The performed multiply-charge correction is based on the measured number size distribution and is described in detail in Deng et al. (2011): in brief, starting at larger sizes the number of possible multiply-charged particles at one size is calculated based on the charge equilibrium (Wiedensohler, 1988) and subtracted from the particle number at the corresponding smaller sizes.This is done for the whole N and N CCN distribution from large to small particles.

Results and discussion
Ideally, a fixed time difference of 20 min would be applied to compare upwind and downwind measurements, i.e. the measurement from the upwind station would be paired with a measurement from the downwind station, which was taken 20 min later.However, the set supersaturation should be the same at both stations for comparable measurements.Therefore, downwind data within 60 min of the upwind time stamp was included in the analysis.The number of activation measurements (n) per supersaturation (SS) for the matching time periods are given in Table 2.

Activation diameter and hygroscopicity parameter κ
The averaged values of the critical diameter (D c ), its standard deviation (σD c ) and the hygroscopicity parameter κ for each SS across all Full Cloud Events (FCE) and Non Cloud Events (NCE) are given in Table 2.We merged all the FCE data and respectively all the NCE data to have a better statistical basis.During cloud events, D c at the upwind station was observed to be larger than at the downwind station, with upwind values between 194.3 (SS = 0.07 %) and 122.9 nm (SS = 0.2 %) compared to downwind D c between 173.9 (SS = 0.07 %) and 101.5 nm (SS = 0.2 %).Consequently, during cloud events, the calculated κ values at the upwind station (0.4, 0.42 and 0.20 for SS of 0.07, 0.1 and 0.2 %) were smaller than after cloud passage, where κ values of 0.54, 0.54 and 0.32 at SS of 0.07, 0.1 and 0.2 % were calculated.No significant change in D c and κ was observed for non-cloud events (n = 55, p > 0.01).
In Fig. 3a and b the results are illustrated.The error bars were calculated by assuming a maximum absolute error in SS of ±0.02 % for SS ≤ 0.2 % and assuming a 10 % relative uncertainty for SS > 0.2 % (Gysel and Stratmann, 2013), and applying Eq. ( 2) to calculate κ.Due to the nonlinear relation between SS and κ also the error bars are asymmetric and give the maximum uncertainty in κ.The increase in κ after the cloud passage in the FCE is obvious, whereas in the NCE the data fall together on the 1 : 1 line.However, the observed effect is within the measurement uncertainty -especially for the lower supersaturations.Therefore, we tested the statistical significance of the change in critical diameters (and thus κ values) between the stations during FCE and NCE, and re-estimated the uncertainty of κ by modeling the instrumental error in supersaturation by a Gaussian distribution.

Statistical analysis of the critical diameters and κ uncertainty estimation
We used statistical testing to determine if the change from upwind critical diameters D c,up to downwind critical diameters D c,down is significantly different between cloud and non-cloud events.This statistical testing scheme is known as "between-within" or "mixed" design, and it is analogous to the statistical experimental design in medicine called the "pre-post case control study" in which half of the patients are given medicine and the other half placebo, and the patients are tested before and after the treatment.The experimental design is illustrated in Fig. 4.While testing, it is essential to take into account that the pre-and post-(up-and downwind) measurement points during the same day are paired, which accounts for variability between days and thus reduces noise.The simplest statistical test for a mixed design is called Change Score Analysis (Oakes and Feldman, 2001), which is essentially a t test between ∆ FCE and ∆ NCE where ∆ = D c,up −D c,down .The null hypothesis was that there is no difference between FCE and NCE for D c,down with respect to D c,up .A confidence of p < 0.01 was needed to reject the null hypothesis.The statistical analysis showed that for every supersaturation (0.07 %, 0.1 %, 0.2 %) the downwind critical diameters with respect to upwind diameters were significantly smaller during FCE than during NCE, with p values of 2.676 × 10 −5 , 1.404 × 10 −3 and 3.137 × 10 −5 for 0.07 %, 0.1 % and 0.2 % supersaturations, respectively.The critical diameter data sets for each supersaturation were tested separately.0.4 % supersaturation was excluded from testing because there was no data for FCE periods.
We also checked with a t test that there is no significant difference between the FCE and NCE upwind critical diameters, to show that the differences in downwind critical diameters are caused by the cloud processes.As statisticians disagree on the correct statistical tests for mixed designs (Senn, 2006), we applied also the Analysis of Covariance (ANCOVA), assuming a linear model D c,down = α+cloudiness+D c,up +ϵ, where α is an intercept term, cloudiness is the parameter defining if the data point was measured during an FCE or NCE day and ϵ is a Gaussian noise term.This test is used to investigate the statistical significance of the term cloudiness, using D c,up values as covariates.
The results given by ANCOVA (not shown) are in good agreement with the p values obtained from the Change Score Analysis.
Next, we estimated the uncertainty distribution of κ with Monte Carlo simulations.We have previously observed that the instrumental supersaturation error of the CCNc is Gaussian, with standard deviations of 0.00714 for 0.07 %, 0.1 % and 0.2 % supersaturations and 0.01429 for 0.4 % supersaturation.These standard deviations are obtained from repeated calibration results showing that with 95 % confidence level the absolute uncertainty for supersaturations ≤ 0.2 % is +/− 0.014 % and for SS = 0.4 % the uncertainty is 0.027 %.The 95 % confidence level corresponds to 1.96σ, from which we can derive the aforementioned standard deviations.However, due to the nonlinear relationship between κ and the critical diameter, the uncertainty distribution of κ is non-Gaussian.The distribution of κ is simulated for each data point separately by drawing 100 000 random samples from a Gaussian supersaturation distribution (µ = 0.07, σ = 0.00714) and using Eq. ( 2).An example of a simulated κ distribution is presented in Fig. 5, showing the 2.5, 25, 50, 75, 97.5 and 100th percentiles.All the analyses were done using R statistical software (R version 2.15.3, 2013).
By applying this statistical approach to the data, it is possible to present more realistic error bars.
Using the maximum absolute error is a bad way of representing a Gaussian distribution, and since we know that the error in SS is Gaussian, the original error bars are a crude approximation.By assuming a Gaussian distributed SS error we are able to calculate the uncertainty distribution of κ (by Monte Carlo sampling), and from this distribution it is easy to calculate percentiles with which to represent error bars at desired confidence level.Percentiles, e.g.
95 % confidence intervals are a more correct way to represent the uncertainty in κ than the maximum absolute error.Figure 6a gives single κ values at the upwind station compared to the κ at the downwind station during FCE.The error bars presented in the figure are the 95 % confidence intervals calculated from Monte Carlo simulations as explained above.All κ values derived for the downwind station are higher than those at the upwind station.The same analysis was again done for the NCE periods (Fig. 6b).Here, within error bars, the data of all SS fall onto the 1 : 1 line.This leads to the conclusion, that we measured particles in NCE periods with the same hygroscopic properties at both up-and downwind stations.The statistical test results support this conclusion.There is still considerable error in the κ values, however the rigorous statistical analysis showed that the decrease in critical diameters due to cloud processing is significant.The results clearly demonstrate that the particle properties changed between up-and downwind stations only when a hill cap cloud was present, leading to more hygroscopic aerosol particles downwind of the cloud.

Chemical in-cloud processing of the particles
Our findings can be explained by the enrichment of hygroscopic material in the particles during cloud presence.Assuming a chemical composition similar to the one given in Wu et al. (2013) for the upwind station, with a mass fraction of 40 % organic material and 30 % each of ammonium nitrate and ammonium sulfate we can model the observed κ of e.g.0.40 (compare Table 2).The measured increase in κ would during FCE translate to an increase in the mass fraction of 20 % in ammonium nitrate and ammonium sulfate between the upwind and downwind stations.
This estimate is supported by measurement results from other groups during HCCT-2010, who focused on the chemical and isotopic signature of the particle population; for example, sulfur isotope analysis of the particulate material was used to investigate the in-cloud production of sulfate.Combined gas phase and single particle measurements allowed the dominating sulfate production sources to be identified (Harris et al., 2014).Direct sulfate uptake, through dissolution of H 2 SO 4 gas and scavenging of ultrafine particulate, as well as in-cloud aqueous SO 2 oxidation by H 2 O 2 , were found to be the most important sources for in-cloud addition of sulfate to mixed particles (the most common particle type at HCCT-2010).While in-cloud aqueous oxidation of SO 2 primarily catalyzed by transition metal ions (Harris et al., 2013b) was most important for coarse mineral dust.The isotopic analyses showed that the sulfate content of particles increased following cloud processing at HCCT-2010 by > 10-40% depending on particle type (cf.table 5 in (Harris et al., 2014)).
Consistent with our results of increased hygroscopicity, both offline (impactor) and online (AMS) measurements of the chemical aerosol composition during HCCT often indicate an increased mass fraction of sulfate in aerosol particles after their passage through a cloud (van Pinxteren, Poulain, D'Anna, personal communications, 2013, data yet to be published in forthcoming companion papers of this special issue).
Mass spectrometric analysis of cloud residuals at the in-cloud station Schmücke showed an enhancement of nitrate in the cloud residuals compared to particles sampled under cloud-free conditions (Schneider et al., 2014).Additionally, a change of mixing state was observed by single particle mass spectrometry (Roth et al., 2014).The cloud residuals showed a higher fraction of particles internally mixed with sulfate and nitrate compared to the particles sampled under cloud free conditions.These findings can be explained by uptake of HNO 3 and sulfate production in the cloud droplets, resulting in an increased hygroscopicity after the cloud passage.

Summary and conclusions
In the ground-based cloud experiment HCCT-2010 the activation diameter of aerosol particles were determined before and after passage across a hill.For cases, with a proved connected flow the activation properties of aerosol particle at the up-and downwind stations were compared.For cases with a cap cloud on Mt.Schmücke a decrease in the critical diameter and a consequent increase of about 50 % in the hygroscopicity parameter κ was observed.In the cases with a connected flow between the valley stations and no cloud on the hill top, no change in activation diameter was detected.The statistical significance of these findings was rigorously tested.All the κ values during cloud events were larger at the downwind than upwind station, and the critical diameters were significantly smaller than during non-cloud days.Therefore, we conclude that in-cloud processes significantly increased CCN activity during all observed cloud events at HCCT-2010.
A possible explanation for the increased κ is the enrichment of more hygroscopic material dur- Our results demonstrate the strong impact of in-cloud processing on the hygroscopic properties of potential CCN.Consideration of our findings in modeling studies will improve cloud representation substantially.
Fig. 1.Sketch of the terrain of the HCCT experiment.For an approaching flow from southwest, Goldlauter is the upwind station and Gehlberg the downwind station.This was the case for all defined full cloud events given in Table 1.

Fig. 2 .Fig. 3 .
Fig. 2. Set-up for the size segregated CCN measurements.The experimental set-up was identical for the upand downwind sites.

Fig. 4 .FrequencyFig. 5 .Fig. 6 .
Fig. 4. Schematic depiction of the experimental mixed design.The starting and end points of the black lines refer to the critical diameter data points.
ing cloud processing, such as nitrate and sulfate.Particulate isotope measurements support our observations: dissolution of H 2 SO 4 and scavenging of ultrafine particulate in the cloud as well as in-cloud aqueous SO 2 oxidation by H 2 O 2 were identified as being the most important in-cloud sulfate addition process for modifying CCN activity in the majority of particles.Mass spectrometric measurements also corroborate the enrichment of soluble material in the particles during cloud: increased nitrate and a change of mixing state was found in cloud residuals.Our measurements suggest that after cloud dissipation the added hygroscopic material remains in the cloud residual aerosol particles.

Table 1 .
Overview of all defined Full Cloud Events (FCE) and Non Cloud Events (NE NCE) for which CCN data are available at both valley stations.