New Particle Formation in the South Aegean Sea during the Etesians: importance for CCN production and cloud droplet number

We examine the concentration levels and size distribution of submicron aerosol particles along with the concentration of trace gases and meteorological variables over the central (Santorini) and south Aegean Sea (Crete) from 15 to 28 July 2013, a period that includes Etesian events and moderate northern winds. Particle nucleation bursts were recorded during the Etesian flow at both stations, with those observed at Santorini reaching up to 1.5 × 10 particles cm. On Crete (at 20 Finokalia station), the fraction of nucleation-mode particles was diminished, but a higher number of Aitken-mode was observed as a result of the downward mixing and photochemistry. Aerosol and photochemical pollutants covaried throughout the measurement period: lower concentrations were observed during the period of strong Etesian flow (e.g. 43 70 ppbv for ozone, 1.5 -5.7 μg m for sulfate), but were substantially enhanced during the period of moderate winds (i.e., increase of up to 32% for ozone, and 140% for sulfate). To understand how new particle formation (NPF) affects cloud formation, we 25 quantify its impact on the CCN levels and cloud droplet number concentration. We find that NPF can double CCN number (at 0.1% supersaturation) but the resulting strong competition for water vapor in cloudy updrafts decreases maximum supersaturation by 14% and augments the potential droplet number only by 12%. Therefore, although NPF events may strongly elevate CCN numbers, the relative impacts on cloud droplet number (compared to pre-event levels) is eventually limited by water vapor availability and depends on the prevailing cloud formation dynamics and the aerosol levels associated 30 with the background in the region. Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 31 May 2016 c © Author(s) 2016. CC-BY 3.0 License.

been frequently observed at Finokalia (Manninen et al., 2010;Ždímal et al., 2011;Pikridas et al., 2012;Kalivitis et al., 2014Kalivitis et al., , 2015 and Akrotiri (Kopanakis et al., 2013) stations during different periods of the year, but more frequently during winter than summer. According to the literature, the NPF events are favored when airmasses are enriched by a reactant (probably NH 3 ), prior reaching the site of Finokalia (Pikridas et al., 2012;Kalivitis et al., 2014). During the Etesians in particular, the particle size distributions were centered on the lower end of the accumulation-mode size range (Kalivitis et al., 2014). This 5 was partly attributed to the fast oxidation of SO 2 that resulted mainly in acidic particles and to the increase of the condensation sink, which is the main limiting factor suppressing the events during the summer (Pikridas et al., 2012). It has been only recently shown that NPF events could occur at Finokalia during Etesians (Kalivitis et al., 2015). A large number of PM 1 particles (of the order of 10 4 cm −3 ) were also observed at the northeastern AS during an Etesian outbreak (Tombrou et al., 2015), whereas high number concentrations of nucleation-mode particles observed in the north AS by Triantafyllou et 10 al. (2016), have been associated with polluted air masses transported from Istanbul.
A natural question therefore, is to understand the track record of the air masses passing over the Aegean before arriving at Finokalia. In particular, we need to elucidate the atmospheric and chemical processes that affect ageing of the air masses passing over the AS maritime area between the Cyclades and Crete, and furthermore, examine whether NPF events observed at Finokalia would be stronger over the central Aegean during the northern Etesian flow. What is more, evidence 15 that the NPF events are associated with an increase in the concentration of cloud condensation nuclei (CCN) production in the EM atmosphere, has been recently demonstrated based on simultaneous measurement of particle number size distributions, CCN properties and aerosol chemical composition (Kalivitis et al., 2015). Bougatioti et al. (2009Bougatioti et al. ( , 2011 have shown high CCN concentrations at Finokalia, from air masses coming from the Balkans, during a period representative of an Etesian regime. However, no study to date has actually focused on understanding the increase in cloud droplet number that 20 results from NPF, which is the true microphysical link between NPF and the aerosol indirect effect. Driven by the above arguments, we chose to perform measurements at a remote site on Santorini, which is located within the same path of airmasses reaching the station of Finokalia, during the Etesians. Our aim is to elucidate both atmospheric and chemical processes that affect ageing of the air masses passing over the AS before reaching its southern edge, the island of Crete. Continuous ground measurements of particle properties, concentration of gaseous species, and 25 meteorological variables were simultaneously collected on Santorini and Crete. During this short-term campaign (15-28 July 2013) intense bursts of nucleation-mode particles were observed at both sites. The synoptic wind flow and boundary layer dynamics as well as the atmospheric chemical composition that favor the enhanced NPF events during the Etesian flow are examined in this study. To understand how NPF could affect cloud formation throughout its evolution, we quantify its impact on CCN levels, cloud droplet number concentration (CDCN) and supersaturation formed in clouds that develop 30 before, during and after NPF events at both ground sites. Complementary to this analysis, wind patterns and atmospheric chemical composition based on WRF-Chem mesoscale model simulations, are presented. Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 31 May 2016 c Author(s) 2016. CC-BY 3.0 License.

Experimental Observations
Ground level measurements were conducted simultaneously at two remote coastal areas (cf. Fig. 1), from 15 to 28 July 2013: on the island of Santorini (at Ag. Artemios;36° 26' N,25° 26' E) and at the monitoring station of Finokalia on the island of Crete (35° 20' N, 25° 40' E;http://finokalia.chemistry.uoc.gr;Mihalopoulos et al., 1997). Ag. Artemios (hereafter referred to 5 as Santorini) is located at an elevation of 153 m above sea level (asl), while Finokalia on the top of a hill at 260 m asl. Both measuring sites are far from any large city or anthropogenic activity, and are close to the sea; Finokalia is facing the sea within a sector of 270º to 90º, whereas the station on Santorini within a sector of 340º to 120º.
The Finokalia monitoring station houses a suite of instruments for measuring the meteorological parameters, the concentrations of gaseous species, as well as the physical properties and chemical composition of atmospheric particles.  (Draxler and Rolph, 2015) to reveal the origin of air masses reaching the study area. The back-trajectories, initialized with meteorological conditions from GDAS (0.5 o ), were computed at several heights. All three-dimensional trajectories were computed with an end point either at Santorini or Finokalia station. 15

Prevailing atmospheric and air quality conditions
Northern winds prevailed over the AS throughout the entire campaign. Based on the simulated wind patterns at 100 m above ground level (agl) throughout Greece (cf. Fig. S1 in the supplementary material) and on the sea level pressure fields (NCEP/NCAR; Fig. S2) the 17 -18 July and 22 -24 July are periods of strong Etesian winds (Brody and Nestor, 1985;20 Kotroni et al., 2001;Anagnostopoulou et al., 2014). Hereafter, we refer only to the second period, as higher aerosol number concentrations were measured at both stations, but also because there were no O 3 measurements at Santorini, during the first period. Immediately after the second period, another characteristic period followed (25)(26)(27), having a similar pressure pattern with the previous two; the pressure gradient over the Dardanelles was weaker. Back trajectory analysis of the air masses sampled at both stations indicates almost the same source regions, for both periods (Figs. 2, S3). However, different 25 conditions prevailed during these two periods altering mainly the last part of the journey of the airmasses, over the AS. From 22 to 24 July, strong northern wind speeds prevailed (> 10 m s -1 ) with the wind direction forming the characteristic 'ringshape' (Fig. S1) of the Etesian flow around Turkey (Tyrlis and Lelieveld, 2013). From 25 to 27 July, moderate wind speeds (up to ca. 8 m s -1 ) with northeasterly surface flow displayed over the central and southern AS, while stagnant conditions prevailed at the north (Fig. S1). 30 The measured wind speeds at Finokalia station exceeded 9 m s -1 , and the wind direction was mainly from westsouthwestern during the daytime hours (Fig. 3) owing to topographic features that steer the prevailing direction towards the Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016  west/southwest direction. Capturing this local feature is a known challenge for regional models (e.g., Gauss et al., 2011;Im et al., 2011;Hodneborg et al., 2012). At the same time, the simulated wind direction at Santorini station exhibited a northern direction, with wind speeds exceeding 8 m s -1 during the daytime hours (Fig. 3).
The number concentrations for the three particle modes (nucleation, Aitken and accumulation) together with the O 3 concentrations are shown for both periods at Santorini and Finokalia stations in Fig. 4. Simultaneous routine meteorological 5 measurements, such as surface temperature and relative humidity, are also provided for each station. Apart from the regionwide differences, intense bursts in the concentration of nucleation-mode particles having diameters smaller than 25 nm were observed at both stations during the period of Etesians; it should be noted that these events were not observed at any of the stations during the period of moderate winds (Fig. 4). In the subsequent sections the different characteristics and processes prevailing under strong or moderate northerly wind flow are explored aiming to elucidate the interconnection between the 10 two stations.

Ozone concentrations
During the Etesian flow period, O 3 levels at Santorini and Finokalia stations ranged between 38 and 66 ppbv and 43 to 70 ppbv, respectively (Figs. 4, S4, Table 2); these levels are consistent with previous measurements (57 ± 4 ppbv) inside the MABL for Etesian flow carried out during the Aegean-GAME campaign. The values also agree with the climatological 15 values recorded over the greater area during summer (Gerasopoulos et al., 2005;Kalabokas et al., 2007Kalabokas et al., , 2013. The less pronounced diurnal cycle at Finokalia station, compared to Santorini (Fig. S4, is attributed to a shallower and more stable MABL over Santorini compared to Crete during Etesian flow conditions (Tombrou et al., 2015) that favors higher primary concentrations and thus O 3 scavenging at Santorini, especially when the MABL collapses during nighttime. In the vicinity of Crete, the MABL becomes less stable, due to larger sea surface temperatures (SST) existing southern of Santorini. This fact, 20 together with the topography (i.e., Crete forms a mass of land that is located perpendicular to the Etesian flow), enhances the mixing and the downward transport from the above rich in O 3 concentrations layer. During the moderate winds, high O 3 levels (the highest concentrations of the summer in 2013) were measured at both stations, ranging between 50 and 99 ppbv (Figs. 4, S4, Table 2). At both stations the highest values were observed on 26 July. The lower winds over the northern AS contributed to O 3 accumulation at this area, explaining the high O 3 concentrations at both stations. The maximum O 3 25 concentration observed (but simulated as well) at Finokalia had a 4-h delay compared to that observed at Santorini.
Simulations confirm that the air masses received at both stations during the strong wind period are of the same origin, and representative of Etesian flow conditions (Fig. S3) albeit with a small Ο 3 under-prediction (average bias during afternoon hours up to -21% on 23 and -15% on 24 July, Fig. 5) while they also indicate an O 3 increase during the moderate winds, especially in the southern AS, but also underpredicted (average bias during afternoon hours up to -24% on 26 July, Fig. 5). 30 The model underestimation mainly arises from inaccuracies of the emissions inventory as well as to fixed chemical boundary conditions, representative of clean environment conditions (McKeen et al., 1991). Bossioli et al. (2016) have shown that when realistic representation of the stratosphere-troposphere exchange processes are implemented in WRF-Chem Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016 Manuscript under review for journal Atmos. Chem. Phys.  Figure 6 shows the non-refractive submicron aerosol concentrations measured at Finokalia during the whole experimental period. In general the inorganic and organic mass concentrations had a similar behavior with O 3 during most of the 5 experimental period (Fig. 4). During Etesian conditions, the PM 1 mass concentrations were reduced roughly by a factor of two compared to those during the moderate wind period (Table 2), and were in the range of concentrations measured in the framework of the Aegean-GAME campaign. However, despite that the concentrations of all four species (SO 4 2-, NO 3 -, NH 4 + and organics), were substantially decreased during 23-24 July, the organic fraction exhibited a relative increase, especially at the beginning of this period. Due to lack of data at Santorini, simulated PM 2.5 mass concentrations are used for the analysis. 10

Aerosol mass and number concentrations
The modeled concentrations for sulfate are about 3 μg m -3 , at both stations at 09:00 LST (Fig. 7) quite close to the measured values at Finokalia (Fig. 6). Similar to the case of O 3 , the two stations are located along the less polluted airflow over the AS, on the boundary of the heavily polluted air masses over the eastern AS (Fig. 7) and downward of a thin plume, that starts over the northwestern Asian Turkish coast.
During the moderate wind period (Fig. 3), the aerosol mass concentrations at Finokalia were substantially higher 15 (Table 2; Fig. 6). The increased concentrations were retained until noon of 27 July for sulfate and ammonium, while those of organics continued to increase further until the end of the campaign. The modeled spatial distribution of sulfate concentrations was nearly uniform over the AS, while as for ozone their concentrations increased offshore of the northeastern coast of Crete due to the ageing of air masses in combination with the strong impact of the topography (Fig. 7).
The simulated mass concentrations of secondary inorganic fine aerosols are in agreement with the measured values at 20 Finokalia station.
In contrast to the fine aerosol mass concentrations, their total number concentrations were substantially increased, reaching continental levels during Etesian flow conditions (from 1.5 x 10 3 to 1.5 x 10 4 cm -3 at Santorini and from 2.4 x 10 3 to 7.5 10 3 cm -3 at Finokalia; Table 2, Fig. S5). The Aitken-mode particles followed a similar diurnal variation at both stations, ranging from 4.4 × 10 2 to 7.7 × 10 3 cm -3 and peaking around noon. Accumulation-mode particles were higher at Finokalia. 25 The total particle number concentration measured within the MABL of eastern AS during Aegean-GAME campaign under similar atmospheric conditions were on average 8 × 10 3 cm -3 , with almost 20% (1.4 ± 1.2 × 10 3 cm -3 ) being in the 20-50 nm size range (Tombrou et al., 2015). Greater differences were observed for the nucleation-mode particles (i.e. particles having diameters smaller than 25 nm), with sudden concentration bursts observed at both stations (Fig. 4). On 23 July, a nucleationmode burst was recorded, reaching number concentrations up to 1.3 × 10 4 cm -3 at Santorini and almost 1.4 × 10 3 cm -3 at 30 Finokalia. A second event, but of lower intensity, was recorded on 24 July. It is worth mentioning that apart from the strong winds and lower temperatures, this period is considered humid (relative humidity values reaching up to 80% at Finokalia Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 31 May 2016 c Author(s) 2016. CC-BY 3.0 License. station) in comparison to the period of moderate winds (Fig. 4). The nucleation-mode particles shift gradually towards larger sizes in a banana-shape pattern at both stations, as shown in Fig. 8. The number of particles remained high for several hours at Santorini (cf. Fig. 8), indicating regional NPF (Kulmala et al., 2012).
The associated growth rates (GR) for particles that increased in size from 10 to 25 nm were estimated to be 3.06 nm h -1 at Santorini and 2.05 nm h -1 at Finokalia on 23 July, and 2.08 nm h -1 and 1.76 nm h -1 , respectively, on 24 July. The average 5 GR for particles increasing in size from 7 to 20 nm at Finokalia was reported to be substantially higher (7.5 ± 5.8 nm h -1 ) by Pikridas et al. (2012), with the highest daily GRs observed during the hottest months of the year (May to July 2008). It should be mentioned, however, that the nucleation events reported in that study were mainly related to air masses spending most of the time over the island of Crete, which is not the case for the observations reported here. Neglecting any coagulation losses, formation rates, J D , can be computed as J D = −1 , where ΔN is the number increase of nucleated particles (for a defined 10 size range) during the NPF event. For the two consequent events at Santorini, J D for particles having diameters from 10 to 25 nm ranged between 2.12 cm -3 sec -1 (23 July) and 1.16 cm -3 sec -1 (24 July; Fig. 8). At the station of Finokalia, J D was lower for particles between 9and 25 nm, ranging between 0.27 (23 July) and 1.01 cm -3 sec -1 (24 July; Fig. 8). The similarity between the J D rates at the two sites on 24 July indicate a region-wide NPF event has occurred, yet the rates taken a day earlier are markedly different and thus, indicating a local event. However, we will show later (section 3.4) that this is not the case and 15 more information needs to be taken into account.
Under moderate winds, the total fine aerosol number concentrations were considerably lower than those during the Etesians (from 1.4 × 10 3 to 2.9 × 10 3 cm -3 at Santorini and from 2.6 × 10 3 to 5.1 × 10 3 cm -3 at Finokalia; Fig. S5). Particles in the nucleation mode were absent, while the concentrations in the Aitken mode were substantially lower at both stations, varying from 3.2 × 10 2 to 4.1 × 10 3 cm -3 (Fig. S5). The particle concentrations in the accumulation mode at Santorini had a 20 comparable variation to that of the Aitken-mode , while they were apparently always higher at Finokalia.

Spatial extent of NPF event
The synoptic wind flow and boundary layer dynamics as well as the chemical atmospheric background conditions that favor the enhanced NPF events during the Etesian flow are further examined here. This type of event could be characterized as "type A" according to Boy and Kulmala (2002), owing to the sudden appearance of nucleation-mode particles and their 25 consistent growth for at least 1 hour. The horizontal scale of this event was estimated based on air mass back-trajectory analysis (Hussein et al., 2009), taking into account the time during which measurements at the site indicate a distinct nucleation mode. Following Birmili et al. (2003), HYSPLIT4 back-trajectory calculations started at the time when a nucleation mode was first distinguishable from the Aitken mode at Santorini and were performed for each subsequent hour until the two modes merged (nucleation duration). Following Crippa and Pryor (2013), the duration of NPF was based on the 30 geometric mean diameter of particles with sizes between 10 and 100 nm and from 30 to 100 nm; an event is said to initiate when the difference between the two geometric mean diameters becomes maximum and ends when this difference is less Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 31 May 2016 c Author(s) 2016. CC-BY 3.0 License. than 15% (Fig. S6). Assuming a linear GR (Lehtinen and Kulmala, 2003), this approach showed that the ca. 10-nm particles (the smallest particles we could detect with our instrumentation) were able to grow up to 60 nm within 4.5 h of initial detection. This GR was then used to calculate the minimum spatial scale. On 23 July, the distance covered by the backtrajectories within 4.5 h (starting when the nucleation-mode burst was first recorded at Santorini) spans at least over 250 km to the northeast of Santorini in the center of AS, upwind of the Cyclades complex (at 39° N; cf. Fig. 2). A couple of hours 5 before the sunrise these back-trajectories (both below and above 500 m agl) are observed over the northwestern Asian forest peninsula of Turkey ( Fig. 2; area marked as A in Fig. 1), having previously passed (at higher altitudes ≥ 1000 m agl) from the greater area of Istanbul and the west coast of the Black Sea. WRF-Chem simulations provide additional evidence of this flow pattern on 23 July and show that the prevailing strong wind favors the LRT of air masses, including those originating from higher levels. A similar spatial extent also occurs during the less intense event on 24 July, although this starts with two 10 hours delay (Fig. 2). Air masses are better mixed throughout the boundary layer covering a broader area over Asian Turkey on 24 July.
The plume along the back-trajectories characterized by high concentrations of ultra-fine particles, after crossing the Turkish mainland over night, is transported over the AS, with most of its air masses above the stable MABL. The plume is moving fast with rather negligible mixing, especially above the MABL, thereby affecting areas located further away, such as 15 the central AS, within a couple of hours after sunrise on 23 July (around 9:00 LST) and around noon on 24 July. The rapid advection above MABL, leaves almost intact the majority of the newly formed particles, formed upwind due to favorable chemical background conditions and solar irradiance. Thereafter, while the part of the plume above the MABL passes over the Cyclades complex, the wakes on the lee side of the islands enhances vertical mixing, enabling its entrainment into the MABL. The freshly nucleated particles that remained constantly inside the well-mixed MABL, suffered an early ageing (i.e. 20 growth by condensation and coagulation). The concentrations at both nucleation and Aitken modes jump almost simultaneously accompanied by concurrent increases in O 3 , NO 2 and SO 2 concentrations, at Santorini station, during these two consequent events (Fig. 4). This could be an indication that this station receives masses simultaneously from different layers (inside and above the MABL), in line with a number of cases where maximum rate of change of ultra-fine particle concentrations close to the surface was always preceded by breakdown of the nocturnal inversion and enhancement of 25 vertical mixing (Crippa et al., 2012). Simulation results are in qualitative agreement with these findings, as the concentrations of Aitken mode (nucleation mode is assigned to the Aitken mode) are substantial at these two distinct layers.
In particular, a plume with large particulate load in the Aitken mode (> 1 × 10 4 particles cm -3 ) is spread over northwestern Turkey and subsequently advances over the AS, both inside and above the MABL (Fig. S7 B and C). Our results agree with previous observations during an Etesian event, where number concentrations up to ~1.2 × 10 4 cm −3 were observed at the 30 northeastern AS with the Aitken-mode particles dominating by up to 70% (Tombrou et al., 2015). According to the simulations, the air masses passing over northwestern Turkey (area A in Fig. 1), are enriched with sufficient H 2 SO 4 (Fig. 9) as well as with biogenic isoprene from the forested area (not shown) both necessary to initiate NPF. Due to limited concentrations of NH 3 in this area, according to the simulations (not shown), the air masses over northwestern Turkey are Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016 Manuscript under review for journal Atmos. Chem. Phys. less neutralized (the molar ratio of ammonium over inorganic ions concentrations is < 0.5; Fig. S7 A) compared to the more neutral aerosol reaching the western part of AS. This chemical background over northwestern Turkey seems to favor the intensity of the NPF events; the less intense event on 24 July is consistent with the much lower simulated concentrations in the Aitken mode over this area (3.5 × 10 3 particles cm -3 ) as well as with more neutral aerosols (molar ratio of ammonium over inorganic ions concentrations ~ 0.8-0.9). 5 The air masses arriving 3-h later (after 13:00 LST) at Finokalia, on 23 July (Fig. 4) have trace a lower number of nucleated particles, but higher number of Aitken-mode particles likely from the microphysical evolution of the NPF particles during their transport towards Crete. The 3-h transit timescale is in agreement with the prevailing wind speed (about 10 m s-1; Fig. S1) and the 120 km distance between Santorini and Finokalia. The simulated Aitken mode ranges from 4 to 5 × 10 3 cm -3 ) peaking around 15:00 LST (data not shown). The differences between the two stations corroborate the importance of 10 site-to-site variability even in cases of regional NPF events. Based on the current simultaneous measurements along the same flow stream, it seems that both stations are under the influence of regional NPF events, during the Etesians. Nevertheless, the nucleation mode particles are significantly reduced as they have shifted gradually towards larger sizes, before reaching Finokalia.
During the period of moderate winds on 26 July, the air masses arriving at lower levels (below 500 m agl) at 15 Santorini station (Fig. 2) have mainly passed over continental areas, and have been substantially enriched by anthropogenic emissions, while those at higher levels have covered longer distances over Eastern Europe at the same time. These atmospheric conditions stimulate the mixing of air masses with local anthropogenic and natural emissions favoring photochemical production of secondary pollutants such as O 3 (Figs. 4, S4) and secondary aerosols (Fig. 6) and probably limiting the NPF events. WRF-Chem simulations also provide additional evidence that high concentrations of gases (e.g. O 3 20 shown in Fig. 5) and secondary aerosols (e.g. SO 4 shown in Fig. 7) have been previously well mixed and neutralized (molar ratio of ammonium over inorganic ions concentrations equal to ~ 1) either over the eastern Balkans and/or the western part of Turkey before starting their journey over the AS.

Impact of NPF events to CCN production and droplet number
Understanding how NPF affects cloud formation requires quantification of its impact on the CCN levels that develop for 25 cloud-relevant supersaturations. Although CCN concentrations were not measured, they can still be calculated using the observations of size distribution and chemical composition as follows. First, Köhler theory (Κöhler, 1936;Seinfeld and Pandis, 2006;Petters and Kreidenweis, 2007) is applied to determine, based on knowledge of aerosol composition, the minimum dry size of particles, d c , that can activate at a given level of supersaturation, s. Then, the CCN concentration is determined from the observed size distributions, by calculating the concentration of particles with sizes above d c (Seinfeld and Pandis, 2006). s is 30 either prescribed or determined from a cloud parameterization, both of which are used here. Chemical composition is expressed in terms of the hygroscopicity parameter, κ, (Petters and Kreidenweis, 2007). The presentation of the results and the relevant discussion will be based on the periods before and after the NPF events. Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016 Manuscript under review for journal Atmos. Chem. Phys.  Bezantakos et al. (2013). The aerosol hygroscopicity follows a diurnal cycle being minimum just before noon and becoming maximum late in the afternoon, owing to a higher sulfate-to-10 organic mass ratios (Fig. 6). Consequently, average κ values were estimated to be higher after the NPF events compared to the period before (increase by ~35% on 23 July and up to 15% on 24 July). Given a lack of PM 1 chemical composition measurements at Santorini, the chemical composition at Finokalia is applied instead to the Santorini size-distribution observations. The WRF simulations support this assumption, as a similar chemical behavior is simulated for both stations (Figs. 5,7). The model systematically underestimates the organic fraction at both stations (organic volume fraction does not 15 exceed 0.2), but with minimal impact on resulting κ values, as they do not differ from measurements for more than ±6% throughout the simulation period. From long-term measurements in the study area, the relative contribution of the main PM 1 constituents, including ammonium, is quite consistent over the years (Sciare et al., 2003;Koulouri et al., 2008, Bougiatioti et al., 2013. Thus, a sensitivity test of CCN concentration at Santorini to shifts in κ by ±20%, is also carried out. The resulting CCN timeseries during Etesian flow are shown in Figure 10. Average values of κ, d c , and CCN 20 concentrations at both stations, before and after the NPF events are provided in Table 3. The calculated CCN number concentrations follow a diurnal cycle and tend to be maximum during the afternoon, after the NPF events, following the  (Figs. 4, 10). The higher CCN number concentrations at Finokalia, compared to those observed at Santorini (Table 3), is due to the higher number of accumulation-mode particles passing previously from Santorini (that are too small to activate at Santorini, but have grown to CCN-relevant sizes by the time they arrive at Finokalia, section 3.3). Accordingly, the higher activation fractions (CCN/CN) are observed at the station of Finokalia, with larger and more aged aerosol particles, while at Santorini this is observed at the end of the events, when the smaller particles drop in concentration because they grow to larger 30 sizes. On 23 July, the NPF event increases the CCN concentrations by 157% at Santorini and 106% at Finokalia, compared to their pre-event values; while in some moments the increase can reach up to a factor of 6. During the second less intense event, on 24 July, the CCN increase is lower at both stations (31% at Santorini and 53% at Finokalia). The lower increase is also due to the pre-event background, characterized by higher CCN concentrations. Changes in chemical composition, as described above exhibit a relative low variation in CCN concentrations (at s = 0.6%) up to 10%, following the same diurnal behavior. As expected, lowering the supersaturation at 0.2% leads to the activation of larger particles with d c ranging from 91 to 106 nm that is consistent with the observations reported by Kalivitis et al. (2015). At s = 0.2%, both NPF events contribute up to 50% to the increase of the CCN concentrations at both stations. However, the higher CCN production at Finokalia on 24 July is 5 associated with the accumulation-mode particles at the end of both events.
Studying the impact of NPF on CCN concentrations for prescribed levels of supersaturation (the usual approach for observational studies of NPF) provides an incomplete picture, as it does not consider the feedback of CCN on cloud supersaturation that develops in cloudy updrafts. Mechanistic cloud droplet formation parameterizations (Ghan et al., 2011;Morales Betancourt and Nenes, 2014) can capture this complexity by efficiently calculating the maximum supersaturation 10 (s max ) that forms in a cloud given knowledge of the aerosol size distribution, composition and updraft velocity; the droplet number (N d ) that forms is then given by the CCN concentration at s max . Using this approach, we calculate the droplet number and supersaturation for clouds forming in the vicinity of both sites during all NPF events. The droplet parameterization used is based on the "population splitting concept" of Nenes and Seinfeld (2003), later improved by Barahona et al. (2010) and Morales Betancourt and Nenes (2014). In the calculations of droplet number, the size distribution is represented by the 15 sectional approach, derived directly from the SMPS distribution files. The updraft velocity has been calculated from highresolution airborne measurements performed over this region of AS, under similar atmospheric conditions (Tombrou et al., 2015). The observations suggest that the distribution of vertical velocities in the boundary layer display a spectral dispersion of  w = 0.2-0.3 m s -1 around a zero average value, which is consistent with vertical velocities observed in marine boundary layers (e.g., Meskhidze et al., 2005;Ghan et al., 2011). When applying the droplet parameterization, we employ the "characteristic 20 velocity" approach of Morales and Nenes (2010) to obtain velocity PDF-averaged values of cloud droplet number concentration (CDNC) and s max . As a sensitivity test, we also consider calculations for a convective boundary layer ( w = 0.6 m s -1 ).
The calculation of PDF-averaged values of N d and s max is carried out for every distribution of aerosol number and composition measured for all NPF events. The calculated timeseries are shown in Figure 11 for Santorini (top panel) and 25 Finokalia (bottom panel). s max is negatively correlated with N d at both stations, owing to the increased competition for water vapor by the growing droplets when CCN increase. As a result, N d responds sublinearly to CCN increasesthe degree to which this depends on the level of aerosol concentrations before and during the NPF event. At Santorini, the CCN levels are much lower than at Finokalia (Table 3), so we expect the relative increase in N d from NPF to be higher there. Assuming  w = 0.3 m s -1 , the NPF events are associated with s max decreases at both stations, compared to the period before the events. On 23 30 July, the decrease is on average 12% at Santorini and 9% at Finokalia. As a result, N d concentrations during the NPF event increases by 13% to 124±8 cm -3 at Santorini, compared to the period before the event. At Finokalia, however, aerosol levels are much higher and N d remains virtually the same before and after the NPF event (Table 3). The effect of the less intense NPF Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 31 May 2016 c Author(s) 2016. CC-BY 3.0 License. event on 24 July is higher; N d concentration increases by 36% at Santorini and 4% at Finokalia compared to pre-event values.
The decrease of s max is also higher on this day, 17% at Santorini and even higher 36.4% (at 0.06-0.08%) at Finokalia (Table 3) owing to the higher accumulation particle concentrations compared to the previous events. The variance of N d during the event period, for  w equal to 0.3 m s -1 , is 475 cm -3 at Santorini and 37 cm -3 at Finokalia, while for  w equal to 0.6 m s -1 the variance is 865 cm -3 and 20 cm -3 , respectively. Altogether, this clearly shows that when NPF particles age (e.g., arrive at Finokalia) 5 their competition for water vapor can reduce cloud supersaturation to very low levels.
The larger updraft velocity ends in larger values of s max , which allow smaller particles to activate into cloud droplets.
In particular, N d exhibits a substantial increase for  w = 0.6 m s -1 , but with a similar pattern to that with the lower velocity, especially at Santorini indicating that the impact of mean vertical velocity on the CDNC is higher at this station. In this case, the average N d concentration is 217±15 cm -3 at Santorini and 619±109 cm -3 at Finokalia (increase relative to  w = 0.3 m s -1 by 10 75% and 52%), after the event on 23 July and 286±15 and 786±11, respectively (increase relative to  w = 0.3 m s -1 by ~76%, for both stations), on 24 July. It is interesting to note that for  w = 0.6 m s -1 two N d peaks are observed at Finokalia, from which the first is attributed to local processes as it is observed much earlier than the NPF event at Santorini. The stronger variation of N d at Finokalia, under the higher vertical wind, compared to Santorini, indicates that vertical velocity variations likely dominate the variance of droplet number for clouds in the region of Finokalia. Furthermore, from the partial sensitivity of N d 15 to the total aerosol number, and to κ, the relative contribution of chemical composition and total aerosol number to the variance of N d is attributed. We find that in most cases the predicted N d variability is almost exclusively governed by the aerosol number variation (> 98%, Table 3) and to a lesser extent by the chemical composition (< 2%). The relative contribution of chemical composition becomes more significant at Finokalia only after the intense NPF event on 23 July (10% for  w = 0.3 m s -1 and 19% for  w = 0.6 m s -1 ). This can be attributed to the more "aged" nature of the sampled aerosol at 20 Finokalia, compared to the one at Santorini. This is consistent with the lower s max observed at Finokalia, leading to the activation of larger particle sizes that have been subject to longer atmospheric processing, during their transition to more unstable conditions after Santorini. Altogether, although NPF events may strongly elevate CCN numbers, the relative impacts on cloud droplet number (compared to pre-event levels) is eventually limited by water vapor availability and depends on the aerosol levels associated with the background. 25

Conclusions
Concentrations of chemically-and size-resolved submicron aerosol particles along with concentrations of trace gases and meteorological variables have been simultaneously measured at Santorini (central AS) and Finokalia on Crete (southern AS) from 15 to 28 July 2013. Two well-distinguished periods are identified: the first with strong wind speeds and wind directions forming the characteristic 'ring-shape' of the Etesian flow around Turkey, and the second with moderate wind speeds and 30 northerly direction over the AS. The two periods exhibited intense differences on air quality levels. Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-330, 2016 Manuscript under review for journal Atmos. Chem. Phys. In general the inorganic and organic aerosol mass concentrations have a similar to ozone behavior during most of the experimental period. During Etesian conditions, the mass concentrations were reduced roughly by a factor of two compared to those during the moderate wind period. The total number concentration of aerosol particles was increased during the Etesian flow, varying from 1.5 × 10 3 to 1.5 × 10 4 particles cm -3 at Santorini and from 2.4 × 10 3 to 7.5 × 10 3 particles cm -3 at Finokalia.
Furthermore, intense burst of nucleation-mode particles have been recorded at both stations, with more intense those observed 5 at Santorini. At Finokalia, the fragment of nucleated particles is diminished, and a higher number concentration of the Aitkenmode particles is observed, attributed to downward mixing and photochemistry. The nucleation mode particles are gradually shifting towards larger sizes at both stations, however, at Santorini the number of particles remains high for several hours, indicating regional NPF. During the period of moderate winds, the total number concentration of the particles reaches lower values, while nucleation-mode particles are not detected at any of the stations. 10 The observed NPF events have been initiated at least 250 km (covered within 4.5 hours) to the northeast of Santorini in the center of AS, upwind of Cyclades complex, under favorable meteorological conditions, under a strong-channeled northeastern wind flow received by both stations. The plume, after crossing Turkish mainland during the night, is transported over the AS, with most of its air masses remaining above the stable MABL. The fast advection above MABL leaves intact most of the newly formed particles formed upwind, despite that the wakes on the lee side of the islands enhance vertical 15 mixing, enabling its subsequent entrainment into the MABL. The freshly nucleated particles that remained constantly inside the well-mixed MABL, suffered an early ageing (i.e. growth by condensation and coagulation).
To understand the impact of NPF on CCN levels, using the κ of particles and in conjunction with a typical supersaturation for the area, we calculated the number concentration of particles which act as CCN at both stations. It occurred that due to NPF, CCN concentrations augment considerably during early afternoon (87% on average for both stations and both 20 events), with concentration levels at Finokalia being higher due to particle growth and atmospheric processing. Calculations of droplet number that would form in clouds influenced by the observed airmasses indicate that NPF augments droplet number, but to a much lesser extent (12%) than implied by CCN variation. This behavior demonstrates there is a limit in the amount of droplets that NPF can contribute because the supersaturation in cloud depresses (here, by roughly 14%) as additional CCN are added from NPF. The pre-NPF aerosol levels and prevailing dynamics of the clouds determine the degree of water vapor 25 competition and precondition clouds to be sensitive -or not -to further CCN increases from NPF.