Characterization of aerosol properties at Cyprus, focusing on cloud condensation nuclei and ice-nucleating particles

As part of the A-LIFE (Absorbing aerosol layers in a changing climate: aging, LIFEtime and dynamics) campaign, ground-based measurements were carried out in Paphos, Cyprus, to characterize the abundance, properties, and sources of aerosol particles in general and cloud condensation nuclei (CCN) and ice-nucleating particles (INP) in particular. New particle formation (NPF) events with subsequent growth of the particles into the CCN size range were observed. Aitken mode particles featured κ values of 0.21 to 0.29, indicating the presence of organic materials. Accumulation mode particles featured a higher hygroscopicity parameter, with a median κ value of 0.57, suggesting the presence of sulfate and maybe sea salt particles mixed with organic carbon. A clear downward trend of κ with increasing supersaturation and decreasing dcrit was found. Super-micron particles originated mainly from sea-spray aerosol (SSA) and partly from mineral dust. INP concentrations (NINP) were measured in the temperature range from −6.5 to −26.5 C, using two freezing array-type instruments.NINP at a particular temperature span around 1 order of magnitude below −20 C and about 2 orders of magnitude at warmer temperatures (T >−18 C). Few samples showed elevated concentrations at temperatures >−15 C, which suggests a significant contribution of biological particles to the INP population, which possibly could originate from Cyprus. Both measured temperature spectra and NINP probability density functions (PDFs) indicate that the observed INP (ice active in the temperature range between −15 and −20 C) mainly originate from long-range transport. There was no correlation betweenNINP and particle number concentration in the size range> 500 nm (N>500 nm). Parameterizations based onN>500 nm were found to overestimate NINP by about 1 to 2 orders of magnitude. There was also no correlation between NINP and particle surface area concentration. The ice active surface site density (ns) for the polluted aerosol encountered in the eastern Mediterranean in this study is about 1 to 3 orders of magnitude lower than the ns found for dust aerosol particles in previous studies. This suggests that observed NINP PDFs such as those derived here could be a better choice for modeling NINP if the aerosol particle composition is unknown or uncertain.

mass in the western Mediterranean. Salameh et al. (2007) reported AOD around 0. 15-0.20 (at 865 nm) within a SSA plume during strong wind events with wind speeds up to 18 m s −1 .
Clouds in the atmosphere form when water vapor condenses on aerosol particles that serve as CCN. Clouds in the atmosphere glaciate at temperatures above −38 • C if droplet freezing in initiated by aerosol particles called ice nucleating particles (INP), or at temperatures below −38 • C also through homogeneous freezing (without INP) (Pruppacher and Klett, 2010). Therefore, a change in atmospheric aerosol particles, especially CCN and INP, is bound to impact cloud properties, precipitation, and cloud radiative effects (Fan et al., 2016). Even though clouds are omnipresent in the Earth's atmosphere, and constitute an important role in regulating the radiative budget of the planet, the response of clouds to climate change remains highly uncertain, in particular with regard to aerosol-cloud interactions and feedback mechanisms.
In-situ observations of CCN on Crete were reported by Kalivitis et al. (2015), highlighting new particle formation (NPF) as 10 a source of CCN. At Finokalia, Crete, Bougiatioti et al. (2011) found that air masses originating from central eastern Europe tend to be associated with higher N CCN , and slightly lower hygroscopicity (κ = 0.18), than other air masses.
Seldomly, measurements of INP have been carried out in the Mediterranean. Excluding situations characterized by high altitude transport of dust plumes, Rinaldi et al. (2017) found that at a measurement station in the Po valley basin, INP number concentration (N INP ) was roughly double that of what they observed at the top of an Apeninne mountain. Schrod et al. (2017) 15 found that mineral dust, or a constituent related to dust, was a major contributor to INP on Cyprus. However, due to Sahara dust plumes travelling at several kilometer altitude, N INP at higher altitudes was 10 times higher than at ground level (height ∼700 m).
As outlined above, the aerosol in the Mediterranean region represents a complex mixture of primary and secondary aerosol particles from both natural and anthropogenic sources, with these sources being non-uniformly distributed across the greater 20 Mediterranean region. Most regional and global climate simulations have investigated impacts of global warming on the Mediterranean climate without detailed considerations of possible radiative influences and climatic feedback from different types of Mediterranean aerosols (Mallet et al., 2016). Besides, to the best of our knowledge, seldom studies paid attention to the CCN and INP simultaneously, which both have an effect on climate. The aim of this study is to provide a quantitative understanding concerning the abundance, properties and source of CCN and INP in the eastern Mediterranean. where different and complex aerosol mixtures occur. On one hand, the Sahara Desert in the southwest, and the desert of the Arabian Peninsula in the southeast, favor a regular occurrence of mineral-dust-rich air masses. One the other hand, Cyprus is influenced by anthropogenic emissions from southeastern Europe, as well as the Middle East, and of course, local pollution. This exposure to diverse air masses makes Cyprus an ideal place for investigating the abundance and properties of climate relevant aerosol particles in general, and CCN and INP, in particular. As shown in Fig. 1, the measurement site was located in Paphos,Cyprus (34 • 43 N,32 • 29 E). The measurements took place at the side of a fairly calm coastal highway, facing the 5 Mediterranean Sea. On the northeastern side of the measurement site, 1 km away, is the Paphos International Airport.
The instrumental setup used for these investigations is shown in Fig. S1. An aerosol PM 10 inlet, employed to remove particles larger than 10 µm in aerodynamic diameter, was installed on top of a measurement container. Downstream of the PM 10 inlet, a vertical tube (inner diameter of 1.65 cm), and a diffusion dryer (130 cm), were arranged before the aerosol was lead into the measurement container. The diffusion dryer was installed vertically to avoid gravitational losses of larger particles. Downstream 10 of the dryer and inside the container, the sampled aerosol was split to supply the aerosol to various instruments, measuring particle number size distribution (PNSD), number concentration, as well as hygroscopic and optical (not discussed in this paper) properties.
For the measurement of N INP , two different filter-based sampling systems were utilized. For one set of samples, total suspended particles were collected with a flow rate of ∼10 L min −1 . For a second set of samples, a separate PM 10 inlet was used 15 as inlet, and an air flow of ∼15 L min −1 was sampled onto the filters. No dryer was arranged in the filter sampling system.
The CCN hygroscopicity was derived from N CCN combined with the PNSD. INP freezing behavior and N INP were determined by filter sampling and off-line analysis using freezing array type instruments. In the following, we will briefly introduce the different measurement techniques applied in this study, including calibrations, measurements and data processing.
And lastly, to get additional information on the presence of super-micron particles, depositing aerosol particle were collected 20 at ambient conditions outside of the measurement container.

Particle number size distribution
PNSDs were measured in the size range from 10 nm to 10 µm using a TROPOS-type MPSS (Mobility Particle Size Spectrometer) (Wiedensohler et al., 2012), and an APS (Aerodynamic Particle Sizer, model 3321, TSI Inc., St. Paul, MN, USA). For the multiple charge correction (Wiedensohler, 1988) of the MPSS data, the APS data was accounted for in the inversion of the 25 measured PNSD (Pfeifer et al., 2016). The combined PNSD is then given on the base of the volume equivalent particle diameter, where a dynamic shape factor of 1.17 was used for particles >1 µm, based on Schladitz et al. (2011). More details about the combined MPSS and APS PNSD can be found in Schladitz et al. (2011). Size-dependent particle losses due to diffusion, deposition and sedimentation within the inlet were corrected for utilizing the empirical particle loss calculator (von der Weiden et al., 2009), as shown in Fig. S2. Total particle number concentrations (N total ) were calculated from the measured PNSDs 30 and the size-dependent particle losses. The calibration of the MPSS before, during and after the intensive field study was done following the recommendations given in Wiedensohler et al. (2018).

Cloud condensation nuclei
N CCN was measured using a Cloud Condensation Nuclei counter (CCNc, Droplet Measurement Technologies (DMT), Boulder, USA). The CCNc is a cylindrical continuous-flow thermal-gradient diffusion chamber, establishing a constant streamwise temperature gradient to adjust a quasi constant centerline supersaturation. The sampled aerosol particles are guided within a sheath flow through this chamber and can become activated into droplets, depending on the supersaturation and the particles' 5 ability to act as CCN. The details of the CCNc are described in Roberts and Nenes (2005).
During our study, the supersaturation was varied from ∼0.08 % to ∼0.77 % at a constant total flow rate of 0.5 L min −1 .
To assure stable column temperatures, the first 5 minutes and the last 30 seconds of the 12-minute long measurement at each supersaturation, were excluded from the data analysis. The remaining data points were averaged. A supersaturation calibration (following the protocol by Gysel and Stratmann, 2013) was done at the cloud laboratory of the Leibniz Institute for 10 Tropospheric Research (TROPOS) prior to and after the measurement campaign, to determine the relationship between the temperature gradient along the column and the effective supersaturation. Calibrated supersaturation set-points were 0.08 %, 0.19 %, 0.31 %, 0.54 % and 0.77 %. These calibrated values were used for further calculations.
According to Köhler theory (Köhler, 1936), whether or not a particle can act as a CCN depends on its dry size, chemical composition and the maximum supersaturation it encounters. Petters and Kreidenweis (2007) presented a method to describe 15 the water activity term in the Köhler equation by utilizing the hygroscopicity parameter κ. The κ values reported in this study were calculated as follows, assuming the surface tension of the examined solution droplets σ s/α to be that of pure water: where d crit is the critical diameter above which all particles activate into cloud droplets for a given supersaturation. M w and ρ w are the molar mass and density of water, while R and T are the ideal gas constant and the absolute temperature, respectively. To derive d crit , simultaneously measured N CCN and PNSD are used. Thereto, it is assumed that all particles in the neighborhood of a given particle diameter have a similar κ, meaning that the aerosol particles are internally mixed. At a given supersaturation, a particle can be activated to a droplet once its dry size is equal to or larger than d crit . Therefore, d crit is the diameter at which 25 N CCN is equal to the value of cumulative particle number concentration, determined via integration from the upper towards the lower end of the PNSD. Hygroscopicity κ can be calculated with d crit and the corresponding supersaturation, based on Eq.(1).
Note that the particle losses inside the CCNc (discussed in Rose et al., 2008) are also considered before κ is calculated. More details about the correction method and data processing can be found in previous literature (Kristensen et al., 2016;Herenz et al., 2018).

Ice nucleating particles
We used two setups to sample airborne particles for further analysis. With the first setup, particles were collected on 200 nm pore size polycarbonate filters (Nuclepore Track-Etch Membrane, Whatman) with ∼20 hours time resolution and a flow rate of ∼10 L min −1 . As shown in Fig. S1, we used a computer-based system to switch between filters based on wind directions.
Two sectors were distinguished, i.e., the ocean sector comprising wind directions from 120 to 240 degree, and the land sector, 5 covering the remaining directions. During the campaign, we collected 4 filters with air from the ocean sector, 17 from the land sector, and 2 blind filter samples in total. All of the filters were stored at −18 • C on Cyprus and cooled below 0 • C during transportation. The start and end times of sampling, flow rates, duration and total sample volumes, are shown in Tab. S1. These filters were transported to TROPOS for analysis. At TROPOS, all filters were stored at −18 • C until they were prepared for the measurement. Each filter was immersed into 1 mL ultrapure water (Type 1, Millipore) and shaken for 25 minutes to wash 10 off the particles. The resulting water samples were characterized with the Leipzig Ice Nucleation Array (LINA). It should be mentioned that results from separate tests using 1 mL and 10 mL of washing water were well in agreement (see Fig. S3). LINA is based on the freezing array technique and follows the design described in Budke and Koop (2015). Briefly, 90 droplets with a volume of 1 µL are pipetted from the water samples onto a thin hydrophobic glass slide, with the droplets being separated from each other inside individual compartments. The compartments are sealed at the top with another glass slide, to minimize 15 evaporation of the droplets and to prevent ice seeding from neighbouring droplets. The bottom glass slide is cooled with a Peltier element with a cooling rate of 1 K min −1 . A camera takes pictures every 6 seconds, corresponding to a temperature resolution of 0.1 K. The number of frozen versus unforzen droplets was derived automatically. More details concerning the experimental parameters and temperature calibration of LINA can be found in Chen et al. (2018).
For the second filter-based sampling system, 200 nm pore size polycarbonate filters (Nuclepore Track-Etch Membrane, 20 Whatman) were pre-treated with 10% H 2 O 2 solution, washed with particle free ultrapure water and dried prior to insertion into the filter holder. Daily filter samples with an air flow rate of ∼15 L min −1 for ∼8 hours were taken. In total 25 day time and 2 blind filter samples were collected. All of the filters were stored at −18 • C in Cyprus and cooled below 0 • C during transportation. The start and end times of sampling, flow rates and duration are shown in Tab 24 wells were filled with pure and particle free water, to be able to account for impurities resulting from the washing water and PCR tray surfaces. The PCR trays were then placed into aluminum cooling blocks. Those blocks have been customized by drilling channels into the bulk aluminum, through which the cooling agent thermostated by means of an external chiller (LAUDA PROLINE RP 855) is directed. The temperature of the cooling agent is then lowered by 0.33 K min −1 and monitored by eight calibrated temperature sensors inserted into the aluminum blocks. The number of frozen versus unfrozen wells was derived visually in 0.5 K steps.
For both measurement systems, the cumulative concentration of INP per air volume as a function of temperature can be calculated based on Vali (1971): where N t is the number of droplets/wells and N (θ) is the number of unfrozen droplets/wells at temperature θ. V means the volume (at 0 • C and 1013 hPa) of air distributed into each droplet/well.
The background freezing signal of ultrapure water and water samples resulting from washing of blind filters is determined for the two sampling systems as well. Measured N INP is corrected by subtracting the background concentrations determined for the blind filters and the ultrapure water. 10 Due to the usually small number (order of tens and lower per examined droplet/well) of INP present in the washing water, and the limited number of droplets/wells considered in our measurements, statical errors need to be considered in the data evaluation. Therefore, confidence intervals for the frozen fraction (f ice ) were determined using the method suggested by Agresti and Coull (1998). More details about the background subtraction and measurement uncertainties can be found in the supplement.

Chemical composition
Aerosol particle dry deposition was collection with a flat plate type sampler (Ott and Peters, 2008) on carbon adhesive mounted to standard electron microscopy stubs. Sample substrates were exposed for approximately 24 hours, collecting particles approximately between 1 and 100 µm particle diameter at ambient conditions. Samples were subject to automated electron microscopy single particle analysis, yielding particles size (projected area diameter) and average elemental composition for each particle. 20 Particles were classified according to the composition in group based on a static rules set. For more information on sampling, analysis and data processing refer to Kandler et al. (2018). In this study, we calculated the particle mass deposition rate in the size range from 1 to 8 µm.
3 Results and discussion 3.1 Overview of the meteorology and air quality 25 Time series of meteorological and air quality parameters as measured from 2 to 30 April are shown in Fig. 2. The relative humidity (RH), temperature, wind speed, wind direction, NO x and N total (retrieved from MPSS and APS measured PNSD) were determined at the measurement site. Note that all times presented here are in UTC (corresponding to local time−3).
RH exhibited large variability throughout the campaign, varying from 22.6% to 89.2%, with a mean of 68.4%. Temperature varied from 10.0 to 26.5 • C, with a mean of 17.5 • C. The local wind speeds ranged from 0.1 to 10.1 m s −1 , with a mean of 30 2.8 m s −1 . Fig. S4 shows the wind rose plot based on 10 minutes mean of wind speed and wind direction. It is clear that winds are mainly from northwest, west and northeast. The winds from northwest and west featured higher wind speeds while winds from northeast featured lower wind speeds. NO x varied from 0.0027 to 25 ppbv, with a median of 0.67 ppbv. N total varied from 658 to 61308 cm −3 , with a median of 3954 cm −3 . The NO x and N total were relatively low during most of the campaign. However, sharp increases in NO x and N total were observed frequently and extremely high concentrations (NO x >1.6 ppbv, N total >8000 cm −3 ) only occurred for few hours.

5
A good correlation (R 2 =0.62) was found between such extremely high concentrations of NO x and N total (Fig. S5), indicating a nearby pollution source. The extremely high concentrations of NO x and N total together with the wind direction typically connected to their occurrences, suggests the nearby airport as the source for these pollutions, as will be discussed in more detail in Sec. 3.2.
To get indications concerning possible particle sources, we studied the air mass origin and transport by means of backward 10 trajectory analysis. The calculations were performed with the HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) Model (Stein et al., 2015;Rolph, 2003). Fig. 3(a) shows the 6-day backward trajectories with 1 hour time resolution ending at 500 m above the measurement site. Fig. 3

Particle number size distribution and sources
Particles of different sizes have different formation pathways, sources and behaviors. Fig. 4(a) presents measured super-micron PNSDs as contour plot, together with wind speed information. The super-micron particle concentration varied from 0 to 11 20 cm −3 , with a mean of 2 cm −3 . Fig. 5 shows the time series of particle mass deposition rate for different compounds at Cyprus, for particles between 1 and 8 µm dry diameter. Overall, sea salt accounted for more than 60% of the super-micron particle mass throughout the whole campaign.
Higher super-micron particle number concentrations were mainly observed from 6 to 7, 12 to 14 and 21 to 22 April, with the corresponding air masses originating from dust areas, as shown in Fig. 4(a) by brown dots. As shown in Fig. 5, high dust 25 deposition rates of ∼1 mg m −2 d −1 were also observed during these periods. Therefore, mineral dust was another important constituent of super-micron particle mass during these periods. However, the observed super-micron particle concentrations were relatively low compared to those reported in previous studies (Mamouri et al., 2016;Schrod et al., 2017) for Cyprus during dust plumes. Low concentrations of super-micron particles were observed on 15 April although the respective backward trajectories featured paths over the Sahara dust region. In summary, the super-micron particles observed during the campaign, 30 were a mixture of ∼60% sea salt, ∼32% mineral dust and ∼8% others (mainly sodium sulfate), with the relative contributions being dependent on the actual meteorological conditions and source regions. of NO x were also observed. An exemplary case is shown in Fig. S6. Such kind of behavior usually appeared from 03:00 to 06:00 UTC and 17:00 to 22:00 UTC. A wind rose plot shown in the supplement indicates that during these periods, winds were from the northeast (Fig. S7), i.e., the direction where the Paphos International Airport is located. This is highly suggestive for the airport being the origin of the observed ultrafine particles and NO x . Fig. 6 shows the comparison of medians of PNSDs 5 observed during airport affected (PNSDa) and non-affected time periods. The error bars indicate the range between the 25% and 75% percentiles. It is clearly seen that airport affected PNSDa exhibit a very pronounced ultrafine particle mode with a maximum at diameters of about 15 nm. Such a mode is indicative for a nearby particles source, such as the combustion of fuel at the airport. Previous studies found that airport emitted particles featured similar PNSDs (Hudda and Fruin, 2016;Jasinski and Przylebska, 2018). Therefore, in the following, time periods affected by pollution from the airport were excluded from The criteria are, first of all, the appearance of a distinct new mode (in the nucleation mode size range) in the size distribution.
Secondly, the mode must prevail over a time span of hours. Lastly, the new mode must show signs of growth. For example, 15 newly formed particles occurred at 07:00 UTC 5 April, 08:00 UTC 12 April and 07:00 UTC 22 April, with subsequent particle growth in the next few hours up to days. All observed NPF started during daytime, suggesting that photochemistry products were likely to contribute to the formation of the new particles. The NPF events, which occurred at 07:00 UTC 5 April and 07:00 UTC 22 April, featured continuous particle growth up to several tens of nanometers. The NPF event occurring at 08:00 UTC 12 April exhibits a more complicated time evolution. Around 15:30 UCT 12 April, the PNSDs were affected by pollution from 20 the airport due to the wind direction shifting to the northeast. Around 00:00 UTC 13 April, the wind speed increased and wind directions were from the clean ocean, i.e., clean air mass weakened the particle growth process. Later on, i.e., at 01:00 UTC 14 April, precipitation occurred. This influenced the evolution of the NPF and growth event, but the growing trend in particle size is still to be seen. The observed particle growth events show that newly formed particles can grow up to sizes where they can act as CCN. However, there are several more NPF and growth events which we do not discuss here, because particles did not 25 grow up to sizes making them potential CCN. The low N CCN around 03:00 UTC 14 April was associated with precipitation as can be seen in Fig. 2. Most of the time, high N CCN are associated with NPF and growth events. For example, around 09:00 UTC 5 April, N CCN at higher supersaturations (0.54% and 0.77%) started to increase. The N CCN at lower supersaturations (0.19% and 0.31%) followed at 04:00 UTC 6 April.

CCN and particle hygroscopicity
However, N CCN at the lowest supersaturation (corresponding the d crit around 163 nm) did not increase in connection with the NPF and growth event. Newly formed particles did not grow into that size range, i.e., N CCN at the lowest supersaturation was not affected. The same behavior was observed from 08:00 UTC 22 to 00:00 UTC 23 April. From 13 to 14 April, the NPF and growth were affected by changing wind directions and precipitation. N CCN also shows respective influences, but the overall 5 trend still can be seen.
The probability density functions (PDFs) of N CCN at different supersaturations are shown in the upper panel of Fig. 8. As discussed, N CCN at lowest supersaturation was not affected by the NPF and growth events, so a unimodal PDF was observed.
However, the PDFs of N CCN at other supersaturations are bimodal, with the larger mode (higher concentrations) representing the NPF and growth events. Kalivitis et al. (2015) also found that CCN production resulted from NPF in the eastern Mediter-  However, for the lowest supersaturation of 0.08%, d crit is located in the accumulation mode. Consequently, hygroscopicities derived at these supersaturations, can be assumed to be representative for the Aitken (at supersaturations of 0.77% and 0.54%) and the accumulation mode (at a supersaturation of 0.08%), respectively.
The particle hygroscopicity, expressed as κ, can be seen as a which has also been observed in previous studies (Kalivitis et al., 2015;Kristensen et al., 2016). At the lowest supersaturation of 0.08%, corresponding to the d crit of 163±10 nm, the median of κ is 0.57±0.09. Particles in this size range are members 25 of the accumulation mode, and have undergone cloud processing and aging. This results in higher amounts of sulfates being present, and consequently higher hygroscopicities. A few sea salt particles mixed with organic carbon might also be present in the accumulation mode, according to a previous study (Prather et al., 2013). But the absolute number concentration of sea salt mixed with organic carbon particle in the size range <200 nm is likely limited. A clear downward trend of κ is observed with increasing supersaturations and decreasing d crit (Fig. 9). The κ values in the Aitken and accumulation modes are clearly 30 different, with the error bars considered, indicating significant differences in particle chemical composition for the two modes.
The PDFs of κ change from unimodal to bimodal to unimodal with decreasing supersaturation. As mentioned above, the Considering that Cyprus is only a small island surrounded by ocean, its effect might be limited. Besides, for a location such as Cyprus, it is difficult to determine sources for different air masses only based on wind direction, alone. it should suffice to express this hypothesis here.
Overall, N INP of the land samples are not clearly different from those of the ocean samples, besides for some samples at > −15 • C for which a biogenic contribution is expected. Therefore, a contribution of INP from pollution from the airport is not expected. This would be in line with Chen et al. (2018), who found that aerosol in Beijing did not contain higher N INP during strong pollution events, compared with clean phases.  (Welti et al., 2018), the performed tests yield prove for the INP (ice active at −20 ≤T≤−15 • C) sampled during our measurements originating from long-range transport rather than local sources, as the proximity of sources would cause a more strongly skewed frequency distribution (Ott, 1990;Welti et al., 2018). At −20 • C the data from Welti et al. (2018) is omitted, because more than half of all samples were fully frozen (f ice =1).
As can be seen from Fig. 11(b), our results are comparable to those given in Welti et al. (2018)  To the best of our knowledge, the only in-situ observations at −20 • C for supersaturated conditions (101%) in the eastern Mediterranean was reported by Schrod et al. (2017) during a heavy dust plume at high altitude with 0.03 to 3 std L −1 .

Correlation of N INP with particle number/surface area concentration and parameterization
Scatter plots of LINA-and INSEKT-measured N INP at temperatures of −15, −18 and −20 • C against particle number concentration in the size range >500 nm (N >500nm ) are shown in Fig. 12(a) and Fig. 12(b). The averaged N >500nm during each filter 5 sample varied from 2 to 14 cm −3 . The N >500nm in this study is much lower than that observed during the dust plume period in Cyprus (maximum 75 cm −3 Schrod et al., 2017). The R 2 between N >500nm and N INP are shown in Tab of parameterizations in connection with measured particle number concentrations has to be done with extreme caution, as the encountered particle populations may significantly differ from those considered when developing the parameterizations.
Fig. S12 shows the median particle surface area size distribution (PSSD) for the whole campaign (excluding the airport pollution events). Two different modes were observed, i.e., a small mode (20-500 nm) and a larger mode (500-7000 nm).

20
Based on the PSSD, the concentrations for the total surface area of the small mode (S <500nm ), the large mode (S >500nm ) and for both modes combined (S all ) were calculated. The S <500nm is about 4 times higher than S >500nm . Scatter plots of LINA and INSEKT measured N INP against S <500nm , S >500nm and S all are shown in Fig. S13(a) and Fig. S13(b). The R 2 between N INP and particle surface area concentration are shown in Tab. S5. The R 2 are all below 0.20, indicating no correlation between N INP and particle surface area concentration. 25 The ice nucleating properties of aerosol particles may be characterized by its ice active surface site density (n s ). The n s is a measure of how well an aerosol acts as a seed surface for ice nucleation. The n s can be calculated as: Where S is the particle surface area concentration.
Depending on which particle size range was investigated, previous studies calculated n s based on either the total surface area 30 concentration (S all ) or on the surface area concentration of particles larger than 500 nm (S >500nm ). Here, both approaches were used, resulting in n s_all and n s>500 nm , respectively. Fig. 13 shows the n s>500 nm as black box plot and the n s_all as red box plot at −15, −18 and −20 • C. As can be seen, n s increases towards lower temperature, which is expected. The n s results, calculated using LINA and INSEKT measured N INP , are shown in Fig. 13(a) and Fig. 13(b), respectively. The n s values determined from LINA measurements are consistent with those from INSEKT measurements.
To the best of our knowledge, many studies dealt with the n s for dust aerosol particles, while no study investigated the n s for the type of polluted aerosol we encountered in the eastern Mediterranean. In the following, we compare our n s_all for the polluted aerosol on Cyprus, with n s_all based on existing parameterizations (Niemand et al., 2012;Ullrich et al., 2017) for dust 5 aerosols (Fig. 13). However, the n s_all values from the parameterizations are more than 2 orders of magnitude larger than the n s_all found in this study. Price et al. (2018) carried out an airborne measurement in dust laden air over the tropical Atlantic.
The n s_all reported in Price et al. (2018) (shown in Fig.13 as yellow shadow) is about 1 to 2 orders of magnitude higher than our results. Based on airborne measurement, Schrod et al. (2017) found that the n s>500nm at Cyprus ranged between 10 5 to 10 8 m −2 at T=−20 • C, RH water =101 %, shown as green shadow in Fig. 13. 10 In short summary, parameterizations purely based on N >500 nm or particle surface area concentration in mineral dust dominated model systems overestimate the N INP of the polluted aerosol we encountered on Cyprus. Although we cannot clearly say to which extent the aerosol we observed was influenced by anthropogenic pollution, our results here fit to what was found in a different context, anthropogenically polluted air masses in Being (Chen et al., 2018), and is based on the fact that more strongly polluted air masses have larger numbers of particles in the size range above 500 nm than naturally ones.

Conclusions
The A-LIFE campaign took place in April 2017 on the island of Cyprus to investigate the aerosols prevailing in the eastern Mediterranean region. As part of the A-LIFE campaign, ground-based measurements were carried out in Paphos, Cyprus, to characterize the abundance, properties (size distribution, hygroscopicity, ice activity), and sources of aerosol particles in general, CCN and INP in particular.

20
During these activities, NPF and growth events were observed. Following NPF, during some events, on time scales of few hours to days, particles grew into the CCN size range. In fact, the highest observed N CCN were connected with NPF and growth events, which confirms the importance of NPF as source of CCN in the eastern Mediterranean.
Usually, trimodal (Aitken, accumulation, coarse mode) PNSDs were observed. Aitken mode particles featured low hygroscopicities (κ values about 0.21 to 0.29), indicating the presence of organic materials. Accumulation mode particles featured 25 higher κ values of about 0.57, indicating that particles in the accumulation mode underwent cloud processing and aging, resulting in higher amounts of sulfate being present. A few sea salt particles mixed with organic carbon might also be present in the accumulation mode. The super-micron particles were mainly from SSA and partly mineral dust.
PDFs of κ in both, the Aitken and the accumulation mode, exhibit a unimodal structure, while the κ-PDFs for particles sizes close to the Hoppel minimum, feature a bimodal shape. This indicates the presence of both, non-cloud-processed (Aitken mode) 30 and cloud-processed (accumulation mode particles), in the size range around the Hoppel minimum. The average observed κ of 0.39 confirms values found in previous field measurements ( Kalivitis et al., 2015) and in model results (Pringle et al., 2010) for the Mediterranean region.
Atmospheric N INP where determined in the temperature range from −6.5 to −26.5 • C, using two freezing array type instruments (LINA, TROPOS, and INSEKT, KIT). N INP at a particular temperature span around 1 order of magnitude below  Welti et al. (2018). This indicates, that these sampled INP which are ice active below −15 • C originate from long-range transport rather than local sources.
No correlations were found between N INP and N >500nm . Parameterizations (DeMott et al., 2010;Tobo et al., 2013) based 10 on N >500nm were found to overestimate the N INP by about 1 to 2 orders of magnitude. There was also no correlation between N INP and particle surface area concentration. The n s for the polluted aerosol we encountered on Cyprus was found to be 1 to 3 orders of magnitude lower than the n s for dust aerosol particles resulting from previous studies (Niemand et al., 2012;Ullrich et al., 2017;Price et al., 2018). This clearly highlights, that usage of such parameterizations, just based on measured         PNSDa median PNSD median Aitken mode Accumulation mode Coarse mode Figure 6. Comparison of the median PNSD during airport affected (red line) and non-affected (black line) time periods . The error bar indicates the range between the 25 % and 75 % percentiles. Aitken, accumulation and coarse modes are fitted with log-normal distribution, displayed in blue, green and brown lines, respectively.    . ns>500 nm (black box plot) and ns_all (red box plot) as a function of temperature. The results were determined based on LINAmeasured NINP in Fig.(a) and INSEKT-measured NINP in Fig.(b). The boxes represent the interquartile range. Data not included between the whiskers are plotted as an outlier with a star. Two ns parameterizations (Niemand et al., 2012;Ullrich et al., 2017) for desert dust are shown in dashed and solid line. We also compare to recent data from airborne measurement by Schrod et al. (2017) and Price et al. (2018), as shown in green and yellow shadow, respectively.