Sources and processes that control the submicron organic aerosol in an urban Mediterranean environment ( Athens ) using high temporal resolution chemical composition measurements

Submicron aerosol chemical composition has been studied during a year-long period 15 (26/07/2016-31/07/2017) and two winter-time intensive campaigns (18/12/2013 – 21/02/2014 and 23/12/2015 – 17/02/2016), at a central site in Athens, Greece, using an Aerosol Chemical Speciation Monitor (ACSM). Concurrent measurements include a Particle-Into-Liquid Sampler (PILS-IC), a Scanning Mobility Particle Sizer (SMPS), an AE-33 Aethalometer and Ion Chromatography analysis on 24 or 12 hour filter samples. Quality of the ACSM data was assured 20 by comparison versus the above mentioned measurements. The aim of the study was to characterize the seasonal variability of the main fine aerosol constituents and decipher the sources of organic aerosol (OA). Organics were found to contribute almost half of the submicron mass, with concentrations during wintertime reaching up to 200 μg m-3, on occasions. During this season, the primary sources contribute about 34% of the organic fraction, comprising of biomass burning 25 (10%), fossil fuel combustion (16%) and cooking (8%), while the remaining 66% is attributed to secondary aerosol. The semi-volatile component of the oxidized organic aerosol (SV-OOA; 31%) was found to be clearly linked to combustion sources and in particular biomass burning, and even a part of the very oxidized, low-volatility component (LV-OOA; 35%) could also be attributed to the oxidation of emissions from these primary combustion sources. These results highlight the 30 1 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-356 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 16 May 2018 c © Author(s) 2018. CC BY 4.0 License.


Introduction 40
The Greater Athens Area (GAA) has over 4 million inhabitants, gathering more than 40% of the total population of Greece in a basin on the west coast of the Attica Peninsula (450 km 2 ) (Kanakidou et al., 2011;Grivas et al., 2012).Being surrounded by mountains on three sides, the topography of the area is quite unfavorable to dispersion of air pollutants and ventilation takes place mostly under northeasterly flow (Kanakidou et al., 2011;Grivas et al., 2012;Pateraki et al., 45 2014).Under low wind speed conditions and the absence of removal via wet deposition, high levels of atmospheric pollutants are observed, significantly deteriorating the air quality of the city (Fourtziou et al., 2017).
The GAA is influenced by long-range transport of atmospheric particulate matter, as well as by significant local sources, namely traffic, industrial activities and combustion of fossil fuel and 50 biomass, the latter especially during wintertime.Particulate matter is considered as the highest priority pollutant to mitigate with respect to population exposure risks, as severe exceedances of the PM10 air quality standards and elevated concentrations are frequently encountered, especially during the cold period (Chaloulakou et al., 2005).
The economic recession in Greece, making its trace notable on population during the past 55 decade, has initially caused deceleration of industrial activity and reduction of vehicular circulation (Vrekoussis et al., 2013;Paraskevopoulou et al., 2014).At a later stage, the crisis compelled the residents to switch from fossil fuel combustion for domestic heating to sometimes uncontrolled burning of wood and biomass, leading to significant air quality deterioration of an episodic nature (Saffari et al., 2013, Fourtziou et al., 2017, Gratsea et al., 2017).During winter, wood combustion are expected to differ considerably.A detailed description on the source apportionment method applied can be found in the following section.

Comparison of ACSM data with ancillary measurements
As a first quality control/quality assurance of the ACSM data, the ammonium concentrations are compared to the respective ones derived from the PILS, on an hourly basis for winter 2016-17. 190 A very good agreement is found (squared Pearson correlation coefficient R 2 =0.83, slope of 0.98).
The sulfate and nitrate concentrations for the winter 2016-2017 period are compared to the respective ones from the ion chromatography analysis (PM2.5 filters), on a daily basis (R 2 =0.69, slope of 0.95 and R 2 =0.81, slope of 0.95, respectively).The organics concentrations are compared to the organic carbon concentrations of the PM2.5 filters determined by the thermal-optical method 195 (SUNSET Laboratory Inc.) using the EUSAAR-2 protocol (Cavalli et al., 2010).A very good agreement is found (squared Pearson correlation coefficient R 2 =0.94, slope of 1.77) with the slope being close to values reported for urban areas (Petit et al., 2015) and OM:OC calculations from AMS measurements in polluted environments (Saarikoski et al., 2012).The results from the aforementioned comparisons are provided in the Supplementary material (SI.1.1).

Chemical composition and characteristics 220
The time series of the main submicron aerosol components measured by the ACSM and the black carbon concentrations are presented in the upper panel of Figure 2 (one complete year period).The average cumulative concentration of the ACSM components and BC was 15.1±16.7 μg m -3 .The highest concentrations were measured during winter (average 20.9±26.4μg m -3 ) and the lowest during summer (average 9.1.±6.2 μg m -3 ).On an annual basis, the most abundant 225 component was organic aerosol, followed by sulfate, contributing 46.1 and 29.1% to the total submicron mass, respectively, while BC contribution was calculated at 12%, ammonium 8.5% and nitrate 4.3%.In the middle and bottom panels of Figure 2 the respective time series of the main submicron aerosol components during the two intensive 2-month winter campaigns are presented.
During winter 2013-14 the average mass concentration of the ACSM components (plus BC 230 concentrations) was 25.1±29.9μg m -3 , with organics and sulfate contributing 53.5 and 15.9% to the total submicron mass, respectively, followed by BC (12.9%).During winter 2015-16 the average concentration was 19.5±25.3μg m -3 , with organics and BC contributing 51.2 and 17.9% to the total submicron mass, respectively, followed by sulfate (17.6%), nitrate (6.7%) and ammonium (6.6%).It is clearly deduced that during the last winters, organics constitute half or 235 even more of the total PM1 mass, sulfate around 20% and BC around 14%.
What is also striking is the fact that during wintertime, PM1 concentration spikes can reach up to 240 μg m -3 hourly values, with organics taking up most of the mass.Maxima are recorded during night-time and mostly during meteorological conditions favoring pollutants emission and accumulation, such as low wind speed and low temperature (Fourtziou et al., 2017).There are on average 8 such incidents occurring during each winter (10 in 2013-14, 7 in 2015-16 and 8 in 2016-17), with organic levels being higher than 100 μg m -3 .Such levels are the highest reported for Europe during wintertime and highlight the impact of local emissions and especially those related to heating/wood burning (see below), to the levels of organics and consequently PM1.These observations are in accordance with Florou et al. (2017, same site from 10 January until 9 February 245 2013), whereas organics concentration alone reached up to 125 μg m -3 and maxima of 8 μg m -3 for BC and up to 5 μg m -3 for nitrate, were recorded.Similarly, wintertime pollution events with increased local character and elevated organics concentrations (around 100 μg m -3 , average of 22.6 μg m -3 ) have been reported at a regional background site, just outside of Paris, during February 2012 (Petit et al., 2015).

Seasonal variability
The seasonal variability of the main measured species, along with the average PM1 concentration (μg m -3 ), as calculated from the ACSM+BC measurements is shown in Figure 3 and the basic statistics are included in Table 1.Organics contribute 43.9% to the total submicron 255 aerosol mass in summer, followed by sulfate (33.1%), ammonium (12.5%) and BC (7.6%), while in winter, organics and sulfate contribute 49.1 and 24.1%, respectively, followed by BC (12.1%), ammonium (6.9%) and nitrate (6.7%).
The mass concentrations of organics, nitrate, chloride and BC exhibit a clear annual cycle, with minimum during summer and maximum in winter.This pattern seems to be due to a combination also contribute to the processes leading to the observed pattern.In support of the above, larger standard deviation is found in winter, demonstrating the frequency and magnitude of the observed pollution events due to the increased need for heating purposes (Fourtziou et al., 2017).Independently of the year, it can be seen that winter concentrations of organics, nitrate, chloride and BC are very similar and more than twice the respective ones during the rest of the seasons (Table 1).
Organics concentration are consistently high during all studied winters (from December to February), while the higher nitrate values can be also attributed to the increased local sources combined with the overall lower temperatures, which favor the stability of ammonium nitrate (Park et al., 2005;Mariani and de Mello, 2007).Ammonium and sulfate exhibit the opposite seasonal summer sulfate levels are the result of enhanced photochemistry associated with more intense insolation, combined with less precipitation, favoring the regional transport of polluted air masses (Cusack et al., 2012).The seasonal variation of concentrations is in agreement with that observed in Athens, during prior long-term measurement campaigns based on analysis of daily filter samples 280 (Theodosi et al. 2011, Paraskevopoulou et al., 2014;2015).

Diurnal variability
When investigating the diurnal patterns of the measured species (Figure 4), it is observed that during wintertime, ammonium and sulfate do not exhibit any significant variability, which is due 285 to the more regional character of ammonium sulfate (Seinfeld and Pandis, 2016).In order to quantify the extent of this variability we calculated the normalized diurnal pattern by dividing each hourly value with the respective mean concentration.More specifically, sulfate varies by 12.5% around the mean value while ammonium varies by 33%.On the other hand, organics, BC and nitrate vary significantly during the day (83.4%,79.8% and 74.3% respectively).These species 290 clearly double their concentrations during night-time, caused by the additional primary emissions.
Furthermore, BC and nitrate also exhibit a second maximum during early morning hours, which should be attributed to the primary emissions during the morning traffic rush-hour.
During summer, all concentrations are significantly lower, especially organics (note scale change) which exhibit a 5-fold decrease of their mean maximum concentration during night-time.

295
Normalizing the diurnals, as mentioned above, reveals a much less pronounced variability for organics (46.9%), implying a more regional character, while BC and nitrate exhibit the highest variability (67.7% and 57.6% respectively) in accordance to their local nature.The night-time maxima of BC vanishes, nitrate shows much lower concentrations, due to nitrate partitioning between gas and aerosol phase, favoring the vaporization of ammonium nitrate.BC still exhibits 300 only one maximum during early morning hours owing to traffic emissions.Ammonium and sulfate diurnal profile follows expected photochemistry patterns, with peaking concentrations around 14:00 LT (UTC+2), consistent with secondary aerosol formation and increased vertical mixing with regional aerosol from aloft due to the evolution of the convective boundary layer which exhibits a bell shaped diurnal structure ranging from a few hundred meters to above one kilometer, 305 with maximum heights during early afternoon (Asimakopoulos et al., 2004;Tombrou et al., 2007).
Sulfate concentrations exhibit lower night-time background, a concurrent (to winter) primary maximum (8:00-10:00 LT), and a secondary significant increase later in the afternoon (twice the night-time background).Finally, organics concentrations are somewhat higher during early night which could possibly be associated with biogenic/vegetation sources either local or regional that 310 produce volatile compounds and condense on the particulate phase during night when temperatures are lower.Furthermore organic variation also follows the morning peak related to traffic and the late afternoon peak also observed for ammonium and sulfate.Condensation of the particulate phase could also apply for nitrate, which also exhibits higher concentrations during night-time (almost double).315

Source apportionment of organic aerosol
A source apportionment analysis of the organic aerosol data was carried out, separately for the cold (1 November to 31 March) and the warm period (July-August-September 2016 and May-July 2017).The separation served to better characterize the different profiles of the sources, as it is 320 expected that the different m/z intensities vary throughout the year.Several runs were performed, following the technical guidelines and methodology proposed by Crippa et al. (2014).Initially, an unconstrained run with the number of factors ranging from 2 to 8 was performed, allowing for a first estimate of the number of potential deconvolved primary and secondary OA spectra.After identifying the presence of primary OA, namely hydrocarbon-like organic aerosol (HOA), cooking 325 organic aerosol (COA) for both periods and BBOA for the winter period, several constrained runs were performed using the α value approach.
In this context, the HOA mass spectrum was first constrained with a low α value, ranging from 0.05 to 0.1, using as reference the average HOA mass spectrum from Ng et al. (2011b).COA was also constrained with an α value ranging from 0.1 to 0.2, using as reference the mass spectrum 330 from Crippa et al. ( 2013) reported for Paris.Finally, for the three wintertime periods, BBOA was also constrained with higher α values, (between 0.3 and 0.5) using as reference the average BBOA mass spectrum from Ng et al. (2011b).Only m/z ≤ 120 were used in order to avoid interferences from the naphthalene signal (m/z 127, 128 and 129).Weak signals, with signal-to-noise ratio (S/N) below 0.2 were downweighted by a factor of 10, and those with S/N between 0.2 and 1 were 335 downweighted by a factor of 2 (Ulbrich et al., 2009) spectra with spectra found in the AMS mass spectral database (http://cires.colorado.edu/jimenezgroup/AMSsd/)and also the correlation of the time series with external time series (such as BC, 340 NO, CO and nss-K + ).The profiles of all solutions were inter-compared as means for sensitivity analysis and the results are provided in the Supplementary material (SI.2).The affinity between spectra is expressed using the θ angle approach (Kostenidou et al., 2009).In this context each deconvolved mass spectrum represents a vector.The cosine of the θ angle between two such vectors (e.g.MSa and MSb) is equal to the correlation coefficient R between them.The θ angle is 345 calculated using the dot product formula: Calculated angles less than 15 o correspond to R > 0.96 thus indicate spectra which are similar to each other, angles between 15° and 30 o correspond to 0.96 > R > 0.86 and indicate some similarity but also some important differences between the compared spectra, while angles larger than 30 o 350 correspond to correlation coefficient R < 0.86 and thus are considered to indicate spectra that do not compare well.
Warm period: In this period, the selected solution is a two factor constrained run (HOA using α =0.05 and COA using α =0.1) and consists of four factors: HOA (hydrocarbon-like OA), COAlike (cooking-like OA,), SV-OOA (semi-volatile oxygenated OA) and LV-OOA (low-volatility 355 oxygenated OA).The two summer periods have also been treated separately, but the derived spectra were almost identical (θ angles between 1.6 and 11.9 for all derived factors, see SI).The time series of the four identified sources during summer 2017 is shown in Figure 5 along with their diurnal variability and the respective average daily contribution.The mass spectra of the selected solution are also provided in the supplementary material.No primary biomass burning aerosol 360 could be identified, which is justified by the absence of fresh emissions over the city center during the warm period.In the summer periods HOA makes up 5.5% of the total organic fraction, COA around 12% on average (10.9 and 13% for 2016 and 17, respectively).In summer 2016 SV-OOA made up 30.5% and the rest 53% is LV-OOA.In summer 2017, SV-OOA contributes 46% to the total organic fraction while LV-OOA 35%.The dominance of secondary influence (SV-OOA &

365
LV-OOA) is apparent, and accounts for the majority of the organic aerosol.This finding is in accordance with Kostenidou et al. (2015), who reported that 65% of the sampled aerosol during summer can attributed to SOA (SV-OOA & LV-OOA), at a suburban site in Athens.we find a ratio of 0.19, which is comparable to the value of 0.24 obtained for COA during summer at a suburban site in Athens (Kostenidou et al., 2015).Similarly, due to the constraint (α =0.05)
This type of aerosol is the oxidation product of isoprene, denoting a possible link of SV-OOA with biogenic aerosol.This association is further strengthened by considering its similarities with SOA from biogenic precursors, such as a-and b-pinene (θ=20 o and 18 o , respectively) (Bahreini et al., 2005), which are maximum during night similarly to SV-OOA.On the other hand, the derived SV-

385
OOA shares some similarities with SOA from diesel exhaust after 4 h of photochemical ageing (Sage et al. 2008).Finally, SV-OOA does not compare well with the mass spectrum from aged organic aerosol emissions from meat charbroiling (Kaltsonoudis et al., 2017) (32 o <θ<37 o ).This indicates that during summer, it is not linked to the oxidation of primary COA, but rather to SOA formation from the oxidation of VOCs from both biogenic and traffic sources.Finally, LV-OOA 390 is identical (θ=3.6 o ) with the very oxidized regional OOA found in the area (Finokalia, Crete) (Bougiatioti et al., 2014) and has many similarities with the LV-OOA reported by Crippa et al.

395
HOA correlates significantly with nitrate (R 2 =0.63) as well as with BC (R 2 =0.52) while COA shows moderate correlation with CO (R 2 =0.32) and nitrate (R 2 =0.36).SV-OOA is highly correlated with nitrate (R 2 =0.89), implying common mechanisms in their variability, possibly linked with the partitioning between the gas and particulate phases.from a combustion source.LV-OOA shows moderate correlation with sulfate (R 2 =0.47) and ammonium (R 2 =0.44), consistent with the regional character of this factor.Results from the trajectory cluster analysis (Figure 8) show that enhanced LV-OOA levels are related to air masses originating from Eastern Europe and the Black Sea region, which have both been identified as the main areas of influence for secondary aerosols that are regionally processed and transported to 405 Athens (Gerasopoulos et al., 2011;Grivas et al., 2018).
Primary fossil fuel emissions (HOA) are very low during summer, as in July and August most of the Athenians leave for their summer vacations, thus reducing local traffic.Concentrations peak around 7:00 and after 19:00 LT that corresponds to the early morning and evening rush hours in downtown Athens.COA exhibits a slight hump during lunch hours (12-15:00 LT), also seen in the 410 relative contribution of the factor, while a large night-time peak is present at around 22:00 LT.
This late peak is consistent with the late dinner hours and operation of tavernas (typical restaurants) in Athens during the touristic season.SV-OOA exhibits two-fold higher concentrations during night-time, which apart from boundary layer dynamics may also be attributed to the condensation of semi-volatile compounds, as also implied by the excellent correlation of SV-OOA with nitrates.
Finally, LV-OOA exhibits a peak during mid-day, consistent with increased photochemical processes that lead to further organic aerosol oxidation.In coincidence to that a moderate hump is also observed for SV-OOA.
In summary, during the warm period, the vast majority (more than 80%) of organic aerosol in the area is linked to secondary organic aerosol formation.The semi-volatile product is of mixed 420 origin, linked to quick atmospheric processes, within a few hours, such as photochemistry of primary sources, like biogenic emissions from vegetation, traffic emissions, or to a lesser extent regional biomass burning.On the contrary, the low-volatility product is the result of more extensive oxidation of organic aerosol in the area, within a few days, and has, thus, a more regional character.
Since the identification of BBOA is mainly based on the two fragments of m/z 60 and 73, 460 considered as the "fingerprint" fragments of levoglucosan and biomass burning tracers, BBOA exhibits indeed excellent correlation with these two fragments (R 2 =0.94 and 0.9, respectively).
Nss-K + is also proposed as a very good tracer for biomass burning and as is reported by Fourtziou et al. (2017), it shows a significant correlation with BC coming from wood burning (BCwb), during wintertime in Athens.Consequently, the time series of nss-K + provided by PILS-IC and m/z 60 465 are studied together.It appears that during both winters (2013-14 and 2016-17) for which nss-K + data is available, m/z 60 is in very good agreement with nss-K + (R 2 =0.85) (Figure 7a).Furthermore BBOA correlates well with BCwb (R 2 =0.78), nss-K + (R 2 =0.62) and with CO (R 2 =0.52).SV-OOA correlates significantly with both wood burning "fingerprint" fragments of m/z 60 and 73 (R 2 =0.99 for both), BCwb (R 2 =0.82),CO (R 2 =0.61) (Figure 7b) and nss-K + (R 2 =0.61), demonstrating the 470 direct link between SV-OOA and primary combustion sources (mainly biomass burning).It can be seen in Figure 8, that increased concentrations of both BBOA and SV-OOA are linked to air masses originating from Northern and Eastern Europe.During wintertime, these flow categories are associated with the prevalence of synoptic-scale northern winds and a decline in temperature in the area, leading to the appearance of PM episodes due to local combustion for residential 475 heating (Paschalidou et al., 2015).
Therefore, during the cold period, the organic aerosol in the area linked to secondary organic aerosol formation contributes around 65% to the total organic fraction.In contrast to summer, the semi-volatile product has a very clear origin, linked to the fast oxidation of primary combustion sources (BBOA and HOA), which is also reflected on its diurnal variability as will be discussed shortly (Fig. 6).Its affinity to biomass burning tracers points out that the largest part of it originates from the fast oxidation of BBOA.The low-volatility product is in this case of more local than long-range transport nature, as also highlighted by the almost two-fold higher values during nighttime.

490
The diurnal cycles of the five factors are shown in Figure 6 (bottom left panel).HOA, originating from fossil fuel combustion, exhibits maximum values during night, associated with result of enhanced photochemistry, limited precipitation and higher regional transport.
Based on the source apportionment of the organic aerosol, four factors were identified during summer, namely hydrocarbon-like OA (HOA), cooking-like OA (COA), semi-volatile oxygenated OA (SV-OOA) and low-volatility OA (LV-OOA), and five factors during winter, the same as in summer with the addition of primary biomass burning emissions (BBOA).During 545 summer, HOA makes up 5.7% of the total organic fraction, COA around 12%, and the rest is linked to secondary organics (SV-OOA and LV-OOA).HOA has peaking values during the morning traffic rush hour, and COA mainly during night-time.SV-OOA exhibits two-fold higher concentrations during night-time while LV-OOA exhibits a peak during mid-day, consistent with photochemical processes.The semi-volatile product is clearly of mixed origin, linked to quick 550 atmospheric processing within a few hours, of VOCs emitted from primary sources like vegetation, traffic and to some limited extent to processed regional biomass burning.The low-volatility product, on the other hand, is the result of more excessive oxidation, in the order of several days, having thus a more regional character.

200
During the intensive winter 2015-2016 campaign, the concentrations of the ACSM components are compared to those determined from the ion chromatography, based on concurrent filter samples collected at the same site, twice per day, (06:00 -18:00 pm and t 18:00 -06:00 local time).Results indicate an excellent agreement for sulfate (R 2 =0.88, slope of 0.92), ammonium (R 2 =0.81, slope of 1.04), and nitrate (R 2 =0.87, slope of 1.08) (Figure SI.1.2).During the intensive winter 2013-205 2014 campaign, the ammonium concentrations from the ACSM showed significant correlation with the respective ones from the PILS (R 2 =0.81, slope of 0.97).Finally, the sum of the ACSM component concentrations plus BC, measured with the 7wavelength aethalometer was compared with the mass concentrations determined by a Scanning Mobility Particle Sizer (SMPS; TSI 3034) measuring particle number concentrations in the range 210 of 10.4-469.8nm (since February 2017 at Thissio).The results obtained using a constant collection efficiency of 0.5, are portrayed in Figure 1 and indicate strong correlation (squared Pearson correlation coefficient R 2 =0.88) a slope of 0.90 and an insignificant intercept of 0.01.The slight underestimation of the SMPS-derived submicron mass can be attributed to the instrument's upper discrepancy could be attributed to the uncertainty in the estimation of aerosol density used for the conversion of volume to mass concentration of the SMPS (Bougiatioti et al., 2014). 250 275 cycle, with maximum values in summer and minimum during winter and spring.The higher 9 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-356Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 16 May 2018 c Author(s) 2018.CC BY 4.0 License.
, using the aforementioned customized software.Solutions were chosen based on specific indicators, such as the Q/Qexp ratio, residual analysis, reproducibility of factors for different model seeds, the affinity of the deconvolved Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-356Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 16 May 2018 c Author(s) 2018.CC BY 4.0 License.
All constituents exhibit significantly higher concentrations during night-time, with HOA being affected by combustion from central heating and presenting a secondary peak during the morning traffic rush hour.COA has a similar diurnal profile to the one observed during summer.BBOA is also characterized by a pronounced diurnal cycle with peaking values during night from 560 combustion for heating purposes.SV-OOA has almost 6-fold higher concentrations during night, consistent with its link to the oxidation of primary combustion sources, while even LV-OOA exhibits almost 2-fold higher concentrations during night.In contrast to summer, the semi-volatile product during winter has a very clear origin, linked to the fast oxidation of primary combustion sources (HOA and BBOA) with BBOA being the major source, due to the affinity of SV-OOA 565 with biomass burning tracers.Part of the LV-OOA, as well, could originate from the extensive oxidation of the local primary combustion sources, showing that LV-OOA during winter is of more local than regional character.Concluding, it is clear that organic aerosol constitutes a large fraction of submicron aerosol throughout the year, in the urban environment of Athens.During wintertime, a large part of this 570 OA, as high as 50%, originates from combustion sources for heating purposes, such as central heating and biomass burning, causing significant air quality deterioration.Night-time contribution of BBOA is 7-fold higher than the one during day, while the respective contribution of SV-OOA is increased by a factor of 2.6.Given that during wintertime, fine PM concentrations reach up to 240 μg m -3 , the significance of these sources contribution becomes even more striking, 575 demonstrating the necessity for strategic, long-term mitigation actions.("IKY Fellowships of Excellence for Postgraduate Studies in Greece -Siemens Programme, 2016-2017"), in the framework of the Hellenic Republic-Siemens Settlement Agreement.The authors would also like to acknowledge support from Francesco Canonaco and Andre Prévôt from PSI, who developed SoFi and provided valuable input related to Positive Matrix Factorization.This study contributes to ChArMEx work package 1 on emissions and sources.585 19 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-356Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 16 May 2018 c Author(s) 2018.CC BY 4.0 License.from meat charbroiling, Atmos.Chem.Phys., 17, 7143-7155, doi: 10.5194/acp-17-7143-2017, 2017.Kassomenos, P., Kotroni, V., and Kallos, G.: Analysis of climatological and air quality Park, S.S., Ondov, J.M., Harrison, D., and Nair, N.P.: Seasonal and short-term variations in particulate atmospheric nitrate in Baltimore, Atmos.Environ., 39, 2011-2020, doi:10.1016/j.atmosenv.2004.12.032, 2005.Paschalidou, A.K., Kassomenos, P., Karanikola, P.: Disaggregating the contribution of local 770 dispersion and long-range transport to the high PM10 values measured in a Mediterranean urban environment, Sci.Total Environ., 527-528, 119-125, doi: 10.1016/j.scitotenv.2015.04.094, 2015.Pateraki, St, Asimakopoulos, D.N., Bougiatioti, A., Maggos, Th, Vasilakos, Ch, and Mihalopoulos, N.: Assessment of PM2.5 and PM1 chemical profile in a multiple-impacted Mediterranean urban area: origin, sources and meteorological dependence, Sci.Total Environ.

Figure 2 :
Figure 2: Time series of the main submicron aerosol components.On the top panel the one-year period starting on 26 July 2016 and ending on 31 July 2017, on the middle panel the 2013-2014 winter campaign (18 December-21 February), and on the bottom panel, the 2015-2016 winter campaign (23 December-17 February).

Figure 3 :
Figure 3: Monthly average concentrations of the main aerosol constituents.Organics are shown on the top panel for the one year 2016-2017 period as well as the 2013-2014 and 2015-2016 winter periods, while sulfate and ammonium on the middle panel, and BC, nitrate and chloride on the bottom panel

Figure 4 :
Figure 4: Average daily cycle of the main submicron aerosol constituents for the cold period 2016-17 on the top panel and the warm period of 2017 on the bottom panel.The size of the markers indicates the 855 normalized values relative to each species' daily mean value.

Figure 5 :
Figure 5: Time series of the contribution of the different factors identified by PMF between 1 May -31 July 2017 (top) along with their average diurnal cycles (bottom left) and the respective hourly average contributions (bottom right).

Figure 6 :
Figure 6: Time series of the contribution of the different factors identified by PMF between 21 Nov. 2016 -1 March 2017 (top) along with their average diurnal cycle (bottom left) and respective hourly contribution (bottom right).

Table 1 :
Seasonal average concentrations ± standard deviation (range) and total mass of the main submicron aerosol components for the one-year study period and the two winter campaigns.Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-356Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 16 May 2018 c Author(s) 2018.CC BY 4.0 License.

Table 2 :
Contribution of the five organic aerosol components to the total organic fraction during the three individual winter campaigns.