The e ﬀ ect of local sources on particle size and chemical composition and their role in aerosol-cloud interactions

The e ﬀ ects of local particle chemical cloud interactions were investigated by measuring cloud interstitial and total aerosol size distributions, particle chemical composition and hygroscopic growth factors and cloud droplet size distributions on an observation tower, with a special focus on com- 5 paring clean air masses with those a ﬀ ected by local sources. The polluted air masses contained more particles than the clean air masses in all size classes, excluding the accumulation mode. This was caused by cloud processing, which was also observed for the polluted air but to a lesser extent. Some, mostly minor, di ﬀ erences in the particle chemical composition between the air masses were observed. The average size 10 and number concentration of activating particles were quite similar for both air masses, producing average droplet populations with only minor distinctions. As a case study, a long cloud event was analyzed in detail regarding emissions from local sources, including a paper mill and a heating plant. Clear di ﬀ erences in the total and accumulation mode particle concentrations, particle hygroscopicity and chemical composition during 15 the cloud event were observed. Particularly, larger particles, higher hygroscopicities and elevated amounts of inorganic constituents, especially SO 4 , were linked with the pollutant plumes. In the air masses a ﬀ ected by tra ﬃ c and domestic wood combustion, a bimodal particle hygroscopicity distribution was observed, indicating externally mixed aerosol. The variable conditions during the event had a clear impact on cloud droplet 20 formation.

paring clean air masses with those affected by local sources. The polluted air masses contained more particles than the clean air masses in all size classes, excluding the accumulation mode. This was caused by cloud processing, which was also observed for the polluted air but to a lesser extent. Some, mostly minor, differences in the particle chemical composition between the air masses were observed. The average size 10 and number concentration of activating particles were quite similar for both air masses, producing average droplet populations with only minor distinctions. As a case study, a long cloud event was analyzed in detail regarding emissions from local sources, including a paper mill and a heating plant. Clear differences in the total and accumulation mode particle concentrations, particle hygroscopicity and chemical composition during

Introduction
Anthropogenic aerosol particles such as sulphates and carbonaceous aerosols have significantly increased the global mean burden of atmospheric aerosol compared to the pre-industrial times. Prediction of the current and future behaviour of the Earth's Particle size, number concentration and chemical composition are the key aerosol properties in the cloud droplet activation process (Dusek et al., 2006;Hudson, 2007), which has been confirmed in studies based on satellite observations (Brenguier et al., 2003;Sekiguchi et al., 2003), model calculations (Menon et al., 2002;Rotstayn and Liu, 2005) and in-situ measurements (Coakley and Walsh, 2002;Wang et al., 2008). 15 The effect of size and number concentration is well known (e.g. Vong and Covert, 1998;Henning et al., 2002;Komppula et al., 2005;Anttila et al., 2009), whereas the role of chemical composition is still under more investigation (e.g. Drewnick et al., 2007;Hao et al., 2013;Wu et al., 2013).
Using the ratio of the inorganic mass concentration to the total mass concentra-20 tion (inorganic fraction, IO) as a measure of particle composition, Dusek et al. (2006) showed that ∼ 80 % of the particle activation is explained by the particle size distribution and only 20 % by particle chemical composition. Kivekäs et al. (2009) found a positive correlation between activation efficiency and IO but IO was also correlated with accumulation mode particle concentration, making the separation of the effect of cles are often less hygroscopic than larger particles, which are aged and possibly cloud processed. So far, long-term in-situ observations on aerosol-cloud interactions are available only from a few measurement stations, e.g. the Global Atmospheric Watch stations at Pallas, Finland (e.g. Komppula et al., 2005) and Jungfraujoch, Switzerland (e.g. Hen- 10 ning et al., 2002) as well as the SMEAR (Station for Measuring Forest Ecosystem-Atmosphere Relations) IV station at Puijo, Finland (Leskinen et al., 2009;Portin et al., 2009;Hao et al., 2013;Ahmad et al., 2013). Puijo is located in a semiurban environment, which makes it easier to investigate the effects of local pollutant sources and therefore the effect of aerosols with different chemical composition on aerosol-cloud 15 interactions. In this paper we present the results from two intensive measurement campaigns (Puijo Aerosol Cloud Experiment, PuCE 2010 and 2011) and provide new, detailed information about the effect of aerosols with different origins and chemical composition on the particle activation process. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | sources: a paper mill in the north, the city center in the southeast, a heating plant in the south and a highway in the east in north-south direction. Also, residential areas of different sizes surround the tower, with the biggest in the east and south and smaller in the southwest, west and northwest. All local sources are located within 10 km from the tower at approximately 200 m lower altitude than the measurement level (Table 1). 5 A more detailed overview of the station and the surrounding area can be found in Leskinen et al. (2009).

Cloud events
A cloud event is considered to take place at Puijo when the visibility at the top of the tower drops below 200 m. Below this limit the cloud and particle activation properties 10 have been observed to be stable, providing data with best possible quality (Portin et al., 2009). The clouds with a visibility above 200 m may already be non-uniform and the time resolution of the twin-inlet system is not enough to distinguish quickly varying particle properties. Furthermore, cloud events (or cloud event hours, see Sect. 2.4) are classified as rainy if the average rain intensity exceeds 0.2 mm h −1 . This classification 15 is necessary, since rain drops remove both unactivated aerosol particles and cloud droplets, thus affecting the data.

Twin inlet system
At the Puijo station the aerosol sample is collected with two separate inlets located 20 on the top of the tower. The sample is drawn through the roof of the tower to the measurement devices which are located in a room on the top floor. The total air inlet has a cutoff size of approximately 40 µm. The inlet and the upper part of the sampling line are heated to 40 • C. Thus, when the tower is in a cloud, the total air inlet will sample both the cloud droplets and the unactivated, interstitial aerosol Introduction particles. The water from the droplets evaporates because of the heating, leaving only the residual particles. This way it is possible to observe the aerosol size distribution as it would be outside of the cloud. The interstitial inlet has a PM 1 impactor (Digitel DPM 10 for 1 m 3 h −1 flow rate) to prevent the cloud droplets from entering the sample line. When a cloud is present, this 5 inlet samples only the interstitial aerosol since the cloud droplets are too large to enter the sampling line. During clear weather, both sampling lines measure the same aerosol distribution if the aerosol is not changing within 12 min measurement cycle. Between the main sampling lines and the measurement equipment is a valve system consisting of four controllable valves (Comparato, model Diamant, 2000) which are used to switch 10 the measurement devices between the sampling lines in six-minute intervals.

Particle size distribution and number concentration
Particles in the size range of 7 to 800 nm were measured with a twin differential mobility particle sizer (twin-DMPS) (Winklmayr et al., 1991;Jokinen and Mäkelä, 1997). One DMPS measured between 7 to 49 nm with sheath and sample flows of 13.4 and 15 2 L min −1 , and the other from 27 to 800 nm with sheath and sample flows of 5.5 and 1 L min −1 , respectively. Flow checks were made periodically. The instrument was connected to the twin inlet system all the time and a full size distribution for both sampling lines was provided with a 12 min time resolution. The times of the measured size distributions from interstitial and total lines differ by six minutes, which has to be considered 20 in the data analysis, normally by averaging over some time period. By comparing the size distributions between the sampling lines, it is possible to observe the size dependent cloud droplet activation of the particles. In this study, we defined the nucleation mode particle concentration (N nuc ) as the concentration of particles with a diameter D p < 25 nm. For Aitken mode particle con-Introduction

Cloud droplets
Cloud droplets were observed with a cloud droplet probe (CDP, Droplet Measurement Technologies) with a 10 s time resolution. The CDP measures the cloud droplet size 5 distribution in the size range of 3 to 50 µm by classifying the droplets into 30 size bins according to the scattered light of a laser beam at a wavelength of 658 nm. The cloud droplet number concentration in each size bin is calculated by dividing the raw droplet counts with the volume of air passing through the sampling area of the laser beam. The instrument has a custom-built tubular inlet with an external pump to provide 10 a constant sample flow (13 m s −1 , checked in the beginning of both campaigns). It is mounted on a swivel, which keeps the inlet facing the wind. The accuracy of the CDP is estimated to be 20-30 %, typical for other devices with the same detection principle (e.g. forward scattering spectrometer probe, FSSP) (Brenguier and Bourrianne, 1998). The size detection of the probe was proven with glass beads of 5 to 40 µm in diameter 15 in the beginning of both campaigns. The CDP data was also used to estimate the cloud liquid water content (LWC) by calculating the total volume of the droplet population.

Particle chemical composition
The particle chemical composition was studied with an Aerodyne high resolution aerosol time-of-flight mass spectrometer (HR-ToF-AMS, DeCarlo et al., 2006). An aero-20 dynamic lens focuses the particles into a narrow beam, which enters a vacuum, where the particles are flash-vaporized and ionizated. The ion fragments are detected by a time-of-flight mass spectrometer. The aerosol mass spectrometer (AMS) provides the mass concentrations of organics, sulphate, nitrate, ammonium and chloride in the size range 40 nm < D p < 1 µm. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | were negligible. The inorganic fraction (IO) is defined as the ratio of the inorganic mass concentration to the total mass concentration. Twin inlet data for the AMS is available for the whole 2011 campaign. In the 2010 campaign, the AMS was connected to the twin inlet system only for a period of 28 h for a case study (Hao et al., 2013), otherwise to the total line. To get uniform data from both campaigns, AMS data collected from the 5 total line is used when discussing the whole 2010-2011 data set and twin-inlet data in the case study from the 2011 campaign.

Particle hygroscopicity
A hygroscopicity tandem differential mobility analyzer (H-TDMA, Joutsensaari et al., 2001) was used to observe the hygroscopic growth of aerosol particles during PuCE 10 2011. It was connected to the total air inlet in order to measure dried aerosol. The setup has a humidifier between the two DMAs. The first DMA selects particles with a certain dry size from the original polydisperse aerosol. In this study, the selected dry sizes were 80, 100 and 150 nm. The monodisperse aerosol enters the humidifier, which is set at 90 % relative humidity. The size distribution of the humid aerosol is measured with the 15 second DMA. From this size distribution the average hygroscopic growth factor (GF H , the ratio of wet to dry particle diameter) for a certain dry diameter is calculated. The instrument measures one dry size for five minutes, so a full cycle takes about 15 min. As the H-TDMA was operated only for a few days during the 2011 campaign, the data will be presented only for the case study. 20 Typical values of GF H for 100 nm ambient aerosol particles (GF 100 ) vary from 1.0 to 1.5 (Sjogren et al., 2008). Black carbon is hydrophobic (GF H = 1.0), organics are less hygroscopic (GF H ≈ 1.2) and anthropogenic particles with higher IO are more hygroscopic (GF H > 1.3). The ratio between the number concentrations of more and less hygroscopic particles is defined as R GF = N GF>1.25 /N GF≤1.25 , where N GF>1.25 and 25 N GF≤1.25 are the number concentrations of particles with GF H more than and less than or equal to 1.25, respectively. The limit 1.25 was chosen as it represented in most cases ACPD 13,2013 The effect of local sources on aerosol-cloud interactions the midpoint between the low GF H and high GF H modes of the hygroscopicity distributions of this study and the same limit was also used in Kammermann et al. (2010).

Data evaluation
As the first step of the data analysis, one-hour averages were calculated for the whole data set from both 2010 (20)(21)(22) and 2011 (26 September-31 Oc-5 tober) campaigns, except for the CDP, for which the 10 s data were used. The averaging was done in order to even out discrepancies in the twin-DMPS size distributions between the sampling lines. The hours with average visibility below 200 m were classified as cloud event hours. For the case study (Sect. 3.3), instead of hourly averages, the data were averaged over 10 the different subperiods.
An hour was classified as a clear hour if the average relative humidity was below 80 % or the average height of the lowest cloud layer, measured by a ceilometer (Vaisala CT25K) located in a nearby weather station, was over 500 m (∼ 300 m above the top of the tower). The choice to use these criteria instead of some high value for visibility was 15 made because even at visibilities > 40 km, relative humidities higher than 90 % were sometimes observed, which is enough to have a noticeable effect on the twin inlet data.
To study the possible effects of the different local sources, the area surrounding the tower was divided into five sectors according to the local sources described in Sect. 2.1 ( Table 1). The same sectors are also used in Leskinen et al. (2012). It must be noted ACPD 13,2013 The effect of local sources on aerosol-cloud interactions

Overview of cloud events
During PuCE 2010 and 2011, 39 cloud events were observed, ranging from short periods of 15 min to events lasting up to 31 h. In total, these events provided 156 cloud event hours (visibility < 200 m). The majority of the cloud event hours took place when 5 the wind direction was from sector 3 (69 h) or sector 5 (50 h). It is very likely that the air masses coming from sector 5 are cleaner and possess some marine characteristics (Portin et al., 2009), whereas the air masses from sector 3 are affected by local sources. Thus, from now on, the results and discussion presented here will focus on the comparison of these two sectors, which will be referred as polluted (3) and clean (5) sectors, respectively.

Particle size distribution
A summary of the aerosol properties for the sectors with and without local pollutant 15 sources is shown in Table 2 along with the average values calculated from the whole data set. All the particle data discussed are from the total air inlet, if not mentioned otherwise. Also the standard error of the mean was calculated for the observations for the times corresponding to the one-hour averages. The values were calculated for clear (RH < 80 % or height of the lowest cloud > 500 m) (943 h in total) and cloudy 20 conditions (156 h) during the campaigns. The corresponding average size distributions are shown in Fig. 3. The average particle number concentration (N tot ) in the air mass coming from the polluted sector in clear conditions (2930 cm −3 ) was higher than that of the clean sector (2000 cm −3 ) for all particle sizes. The mean total particle volume concentrations (V tot ) were 3.0 µm 3 cm −3 and 0.80 µm 3 cm −3 for the polluted and clean sec-ACPD 13,2013 The effect of local sources on aerosol-cloud interactions tors, respectively. Furthermore, the size distribution for the polluted sector was much broader, suggesting that the particles had originated from multiple sources. In cloudy conditions, the mean N tot decreased by 43 % for the polluted and by 51 % for the clean sector due to particles impacting into cloud droplets and wet removal. Scavenging was most significant for nucleation mode particles, leading to an increase 5 in the geometric mean particle diameters (GMD) of the total aerosol (Fig. 3, Table 2). For the clean sector the GMD increased by 120 %, which is considerably more than the 16 % increase for the polluted sector. The V tot was equal (2.5 µm 3 cm −3 ) for both sectors in cloudy conditions. For the clean sector the V tot in cloudy conditions was three times that in clear conditions. The differences in the particle populations of the two 10 sectors can be explained by cloud processing: some of the smaller particles diffuse to droplets and trace gases convert to particulate matter within the droplets. This increases the size of activated particles and produces bimodal size distributions when the cloud droplets evaporate. The cloud processing is often most evident in clean, marine aerosol (e.g. Hoppel al., 1986;Frick and Hoppel, 1993;Mochida et al., 2011). At Puijo, cloud processing has been observed in the air masses from both sectors but it was more distinguishable in the air masses with possible marine characteristics arriving from the clean sector. For the polluted sector the effect of cloud processing was partly masked by the higher N tot . A clear hump can be seen in the clean sector size distribution at around 200 nm 20 ( Fig. 3b), indicating cloud processing. The hump is also seen in the size distribution measured in clear conditions (Fig. 3a), meaning that the air masses have gone through cloud formation and processing on their way to Puijo. For the polluted sector, the hump can also be observed in both clear and cloudy conditions but it overlaps more with the Aitken mode.

Particle activation and cloud droplet size distribution
The average activated fractions as a function of particle diameter for the two sectors are shown in Fig. 4. For the polluted sector, even smaller particles activate and the activa-32143 ACPD 13,2013 The effect of local sources on aerosol-cloud interactions tion curve is less steep than for the clean sector. The steepness of the activation curve gives information about the aerosol mixing state (Asmi et al., 2012). A steeper curve, like the one observed for the clean sector, is an indication of more internally mixed and more aged particles. A less steep curve means that aerosol from several sources with variable chemical composition and hygroscopic properties have been present, as is the case for the polluted sector. The number concentration of activated particles (N act , calculated as the concentration difference between the total and interstitial sampling lines) differed by 21 % between the two sectors, being 210 cm −3 and 165 cm −3 for polluted and clean, respectively (Table 3). However, the size distributions of the activated particles were very similar for both sectors (Fig. 3b). The only difference was that the size distribution for the polluted sector was tilted towards smaller particle sizes, which also explains the difference in N act . The average droplet concentrations (N d ) were 293 and 266 cm −3 for the polluted and clean sectors, respectively. These numbers are comparable to N act within the instrumental uncertainties of 20-30 %. The arithmetic mean droplet diameters (D d ) were 15 8.3 and 8.9 µm for the polluted and clean sectors, respectively (Table 3). Although these differences were small, this is just what one would expect based on the particle population properties of the two sectors. Higher N tot , especially N acc , of the polluted sector favors more and smaller droplets. The liquid water contents (LWC) were equal, 0.14 g m −3 , for both sectors.

Particle chemical composition
The mass concentrations of the chemical components for the two sectors are shown in the concentration even exceeded that of the polluted sector and for NH 4 the concentrations were equal. The IO was 0.42-0.44 for both sectors and for both clear and cloudy conditions.

A case study on the effect of local sources on aerosol-cloud interactions
During PuCE 2011, a long cloud event took place between 22 October, 09:00 and 5 24 October, 05:15 LT, lasting in total 44 h. The wind direction, temperature and rain intensity varied considerably during the event. Also, different air masses and pollutant plumes from local sources were observed. Thus, it was possible to perform a detailed analysis on the effects of these variable conditions on aerosol-cloud interactions. The event could be divided into eight "sub-events" ( Table 5). The time series of the most 10 important weather and other parameters are shown in Figs. 5 and 6.

Rainy period
The rainy period, with a southerly wind from the polluted sector, a temperature of slightly over 0 • C and some rain (on average 0.8 mm h −1 ), was characterized by the highest N tot of all cloud periods (Table 6). This is mainly explained by a high N ait 15 (Fig. 7a), probably from fresh, anthropogenic emissions. N acc was relatively high compared to the other periods, leading to a high droplet number concentration N d with the smallest D d of all the periods. Normally, the droplet size distribution was bimodal, with the first mode around 10 µm and second mode at ∼ 16 µm (Fig. 7d). For this period, however, only the first mode was observed with a high amount of small droplets. LWC 20 during this period was the lowest during the whole event, so it is possible that the droplet growth was limited by the availability of water. The activated fraction of particles for this period remained low, even for the larger particles, reaching only 0.8 (Fig. 7c). This may also have been caused by the removal of droplets by rain, which affects the particle measurements. Unfortunately, as can be Introduction surement range provides poor statistics, wrongly suggesting very low activated fractions for particles larger than 600 nm in diameter. Furthermore, the particle activation data in Fig. 7b and c suggests that also the smallest particles contributed to droplet formation. This inaccuracy was likely caused by the large variation in the concentrations of small particles, the 6 min time difference between the interstitial and total sampling 5 lines and for some of the periods, the short averaging time. This has to be kept in mind when interpreting Fig. 7b and c and hence the data for particles smaller than 80 nm in diameter is illustrated with dashed lines. The chemical composition of particles was dominated by organics, with the concentrations of other components remaining low ( Table 7). The activated fraction of organics 10 was the lowest for all periods. Also, the particles during this period had a low average hygroscopicity (Table 8) with very low R GF indicating a strong contribution from the low GF H particles. The growth factor distribution was clearly bimodal, especially for the 80 nm particles (Fig. 7e and f). It is likely that the nonhygroscopic mode consisted of particles containing organics or black carbon, some of which remained unactivated.

15
The largest residential areas and a majority of the traffic in Kuopio are concentrated to the south from the tower. Both biomass burning and traffic related combustion aerosols are known to be less hygroscopic (Herich et al., 2009). This could also partly explain the low activated fraction. 20 During the clean period, air masses were coming from the clean sector, there was no rain and the temperature dropped below 0 • C. The air was very clean, likely of marine origin and contained aged aerosol with low N tot and N acc . (Table 6). Also, there were no nucleation mode particles, which was already shown to be typical for this wind sector (Fig. 3). A low N acc led to the lowest N d of all the periods and a large D d .

25
The mass concentrations of inorganic components were somewhat higher during this period compared to the rainy period (Table 7). Also, their activated fraction was higher, meaning that a larger fraction of them was found in the accumulation mode particles. This suggests that the air mass was aged and had gone through some cloud processing, producing internally mixed aerosol before arriving to Puijo. This is also supported by high values for the hygroscopic growth factors (Table 8). The hygroscopicity distribution was dominated by the more hygroscopic mode, especially for the 100 and 150 nm particles as indicated by the high R GF values. The R GF of 100 and 150 nm particles was 5 also strongly dependent on the concentration of SO 4 . In the beginning of the period, SO 4 was almost absent but throughout the period, its mass fraction increased to 0.45. R GF was around 2 and 6 in the beginning of the period but towards the end increased to 7 and 40 for 100 and 150 nm particles, respectively ( Fig. 6c and 8). The average GF 100 was 1.42, comparable with the Jungfraujoch free tropospheric aerosol, which is 10 also aged and internally mixed (Sjogren et al., 2008;Kammermann et al., 2010).

Paper mill period
This short 30 min period was characterized by a heavy pollution plume from the nearby paper mill. There was no rain and the temperature was below 0 • C. The particle population properties differed greatly from those observed during the other periods. N tot was 15 very low but N acc was elevated (Table 6). Due to the pronounced accumulation mode, a very high D 50 , 202 nm, was observed, compared to the normal D 50 at Puijo of around 120 nm, as was the case during the first two periods. A time series of the cloud droplet data for the whole cloud event is shown in Fig. 6b. During the paper mill plume N d increased momentarily, coinciding with a quick de-20 crease in the average droplet size. This sharp change in the droplet population properties is mainly explained caused by the high N acc but the possibility that the different chemical composition of particles also played a role cannot be excluded. The inorganic components all experienced a drastic increase, with SO 4 dominating the composition (Table 7). Growth factor distributions also showed elevated hygroscopicity for the high-  (Table 8). For example, the average GF 100 was 1.37, whereas for the high-hygroscopicity mode it was around 1.6.

Clean 2 period
In the beginning of this period, the wind direction shifted back to the north, fluctuating between the clean sector and sector 1, with no rain and a temperature of just below 5 0 • C. This period shared many similarities with the clean period. N tot was even lower, nucleation mode particles were absent and a pronounced accumulation mode was observed (Fig. 7a, Table 6), indicating that strong cloud processing had taken place before the air mass arrived to Puijo. The main difference compared with the clean period was an elevated inorganic mass 10 concentration, largely due to 2-3 times higher SO 4 and NH 4 concentrations ( Table 7). This indicates that the air mass has probably encountered some anthropogenic influence on its way to Puijo, but not from nearby sources. This had a clear influence on particle hygroscopicity, as seen from the high GF H and R GF values (Table 8) but not on cloud droplet activation. N d was 10 % higher than during the clean period (Table 6) 15 which is of similar magnitude as the difference in N acc between the periods.

Heating plant period
The period started with a rapid shift in the wind direction from north to south and the polluted sector. The temperature was still slightly below 0 • C. At the same time, a heavy pollutant plume from the heating plant reached the tower. The particle population con-20 sisted of pronounced nucleation and accumulation modes (Fig. 7a). N acc was the highest of all the periods, and the particles consisted mainly of SO 4 (Fig. 6c). Also, as the aerosol was highly acidic, the NO 3 concentration was very low. The plume also contained an exceptionally high amount of SO 2 (Fig. 6d), so it is likely that the majority of the SO 4 observed in the activated particles was formed from SO 2 as a result of cloud 25 processing. However, SO 4 also dominated the composition of cloud interstitial parti- cles. Since the smaller, inactivated particles are also liquid at high RH, it is possible that cloud processing from SO 2 to SO 4 took place also in the interstitial aerosol. Another explanation is that a part of the SO 4 particles was formed already at the heating plant.
The conclusions about particle activation parameters for this period have to be made 5 carefully as the time resolution of the distribution scan is not good enough to capture the observed rapid changes in the aerosol properties. For example, D 50 was very high, 273 nm, (Table 6) but this does not necessarily represent the actual size of the activating particles. The activated fraction of particles as a function of size (Fig. 7c) showed a bimodal behavior. Particles with a diameter of around 100 nm already started activating, similar to some other periods, but reached only an activation fraction of 0.4 at 150 nm. After this there is a dip in the curve as the heating plant particles started to affect the activation curve and produced a seemingly high D 50 . The GF H distributions showed a clear bimodal behavior, with the high-GF H mode slightly elevated by the heating plant particles. The low mode was also pronounced, 15 and for 100 and 150 nm particles, similar to the observations during the rainy period, indicating emissions from traffic and residential areas. For 80 nm particles there was a clear increase in the low GF H mode, indicating that the plume contained significant amounts of small particles with low hygroscopicity, likely soot. Since both hygroscopicity modes were affected by the plume, R GF remained moderate. Unfortunately the 20 CDP was frozen shortly after the beginning of the period, making analysis of the cloud properties impossible.

Southern 1 period
During this period, the conditions returned back to normal as the heating plant plume passed the tower. The temperature rose above 0 • C during the period and the wind 25 direction was still from the south and the polluted sector. Similar nucleation and Aitken modes were present as during the heating plant plume (Fig. 7a). For the chemical components the concentrations were quite normal (  (Table 8). As the heating plant had no effect on data during the rainy period, it is likely that this was the case also here. Thus, southern 1 can be considered to represent normal "semipolluted" conditions for this sector when the effects of the heating plant and weather are minor. The CDP was still frozen part of the time, 5 so no reliable droplet data is available.

Southern 2 period
After a short clear period, the tower was again covered in cloud with southerly wind and a temperature of above 0 • C. The aerosol during this period was moderately affected by the heating plant, indicated by the elevated SO 4 and SO 2 concentrations (Fig. 6). Also the concentration of organics was higher than during the earlier periods, which might already be related to the transportation of organic aerosol which was more pronounced during the next period, southern 3. The presence of two different kinds of aerosols had some effect on the activation of particles. The activated fraction curve was less steep than for most of the other periods and the size distribution of activated particles was 15 broader ( Fig. 6b and c). Also bimodal GF H distributions and low R GF indicated the presence of externally mixed aerosol. R GF also (Fig. 8) correlated with SO 4 and SO 2 concentrations, with higher values in the middle of the period. The cloud droplet size distribution was unimodal, similar to the rainy period. This suggests that the unimodality is an occasional feature for southerly clouds and not 20 related to removal of droplets by rain as suggested for the rainy period. However, this does not exclude the possibility that rain removal of droplets was taking place during the rainy period. N d during this period was higher despite a lower N acc compared to the rainy period (Table 6). 13,2013 The effect of local sources on aerosol-cloud interactions

Southern 3 period
During southern 3 period, wind was still blowing from the south. The period started with a drop in the mass concentration of SO 4 and in the concentration of SO 2 (Fig. 6). At the same time, the organic mass concentration increased to the highest value during the whole event (Table 7). The rise in the amount of organics was explained by an 5 increase in N acc , although N tot remained lower than during other southern periods. As the chemical composition and N tot showed little variance during the period (Fig. 6), this would suggest that the effect of local sources was minor. It is likely that these large organic particles were transported to Puijo from somewhere else. These organic particles were also characterized by low hygroscopicity (Table 8). For 80 nm particles the hygroscopicity distribution was unimodal with one broad peak centered at GF H = 1.1. Also for the 150 nm particles the low and high hygroscopicity peaks were broader than for the other periods. (Fig. 7e and f). It is possible that some of the larger particles remained unactivated because of this, as suggested by the unusually high D 50 (Table 6). The availability of water was not a limiting factor. Although N d was 15 quite normal, the droplets were the largest and LWC the highest of all the periods.

Ratio of inorganics to total mass
Also shown in Table 7 is the IO from the AMS measurements for each of the periods. IO was the lowest for the rainy and southern 3 periods, 0.34 and 0.38, respectively, and higher for the clean, clean 2, southern 1 and southern 2 periods, 0.59, 0.64, 0.51 and 20 0.54, respectively. The highest IOs were observed during the paper mill and heating plant plumes, 0.85 and 0.87, respectively. The differences between the periods were considerably larger than those between the different air masses analyzed in

Summary and conclusions
Aerosol-cloud interactions were investigated during two intensive measurement campaigns at Puijo measurement site during autumns 2010-2011. The object was to find out the possible effects of local pollutant sources and particle chemical composition on aerosol-cloud interactions. The first approach was to compare data from two different 5 wind direction sectors for the whole data set. One sector was considered to be clean, with air masses of mostly marine origin. The other sector was affected by local pollutant sources, including residential areas, traffic and a heating plant. In clear conditions, the total particle number concentration and the accumulation mode concentration were higher for the polluted than for the clean sector. In cloudy 10 conditions cloud processing took place, leading to lower particle concentrations. However, unlike for the polluted sector, the accumulation mode concentration increased for the clean sector, indicating stronger cloud processing. The In-cloud particle chemical composition was quite similar for both sectors. The main difference was a higher amount of sulfates for the polluted sector. Despite of some differences in the particle 15 properties, the droplet activation behavior was surprisingly similar for the two sectors. The particles that activated were roughly of same size. For the polluted sector the average droplet concentration was higher and the average diameter smaller than for the clean sector but also these differences were minor.
The second approach was a case study of a cloud event with variable conditions.

20
The wind was blowing from both the clean and polluted sectors and plumes from the local heating plant and paper mill were observed. The total and accumulation mode particle concentrations were clearly elevated for the polluted sector compared to the clean sector. This also created large differences in the droplet properties, with higher concentrations and smaller particle diameters for the polluted sector. Also the particle 25 chemical composition, especially the ratio of inorganic to total mass concentration, varied considerably during the event.
ACPD 13,2013 The effect of local sources on aerosol-cloud interactions Aged, cloud processed air masses from the clean sector typically resulted in an internally mixed, more hygroscopic aerosol with an inorganic fraction of ca. 0.6. With southerly winds, the particle hygroscopicity distributions were clearly bimodal, suggesting externally mixed aerosols. Likely sources for the less hygroscopic particles include local domestic wood combustion and traffic. The concentration of organics was higher, 5 as indicated by the lower inorganic fraction, 0.3-0.5.
The paper mill plume was short in duration but a high accumulation mode particle concentration was observed, leading to a momentary increase in droplet concentration and a decrease in droplet size. The heating plant plume caused an even bigger increase in the accumulation mode concentration. In both plumes, elevated amounts 10 of SO 4 and NH 4 were observed, leading to inorganic fractions of over 0.8. Unlike the paper mill plume, the heating plant plume also contained a large amount of SO 2 . Thus, the SO 4 from the heating plant was formed from SO 2 as a result of cloud processing. For the paper mill plume, the SO 4 particles were either generated at the mill or then SO 2 was present to a lesser extent and was completely transformed into particulate 15 SO 4 before arriving to Puijo. Another difference was the NO 3 concentration, which was elevated in the paper mill plume but very low in the heating plant plume due to highly acidic aerosol. In both plumes, elevated amounts of more hygroscopic particles were observed in addition to smaller, hydrophobic soot particles.
As a conclusion, the case study presented here supported and complemented the 20 results from the sector comparison and the main results from these two methods can be summarized as follows: (1) the particle concentration in aged, cloud-processed, internally mixed and more hygroscopic air masses is low but a pronounced accumulation mode is present, leading to fewer cloud droplets with larger size.
(2) Air masses affected by local sources contain more nucleation and Aitken mode particles with lower hygro-25 scopicity. The aerosol is externally mixed with a higher inorganic content. The cloud droplets are smaller but more numerous.
(3) Local point sources have the potential to affect aerosol-cloud interactions both through an increased particle concentration and through their effect on chemistry.