Composition and Mass Closure of PM 2 . 5 in Urban Environment ( Athens , Greece )

Daily PM2.5 concentrations and composition were monitored at an urban site at 14 m above ground level, at the National Technical University of Athens campus from February to December 2010. Total sulfur, crustal origin elements and elements of a major crustal component (Al, Si, Fe, Ca, K, Mg, Ti) trace elements (Mn, V, Pb, Cu, Zn, Ni) and water soluble ions (Cl, NO3, SO4, Na, K, NH4, Ca, Mg) were determined. Carbonaceous compounds were determined for a period of one month. Sulfur (sulfates) and carbonaceous material were the most abundant constituents (35% and 30% respectively). Ionic balance calculations revealed a cation deficit in PM2.5 attributed to H associated with sulfates. The concentrations of the elements of crustal origin presented abrupt increases during Saharan dust transport events. Mass closure was attempted considering Secondary Inorganic Aerosol (SIA), Particulate Organic Matter (POM), Elemental Carbon (EC), Dust, Mineral Anthropogenic Component (MIN) and Sea Salt (SS). The sum of these components accounted for about 75% of the measured PM2.5 concentrations. SIA and carbonaceous material (OM + EC) contributed almost equally for about 30% in the PM2.5 mass, and while the dust contribution was significant during dust transport events, it was only about 5% in absence of such events. The contribution of sea salt was calculated to be about 4%.


INTRODUCTION
The quantitative estimation of aerosols physicochemical characteristics are of particular interest due to their impact on air quality, human health and climate.The urban environments of the Mediterranean countries are characterized by enhanced concentrations of locally emitted and long range transported particulate matter, absence of precipitation from spring to late autumn and favorable conditions for the formation of photochemical smog.Monitoring of concentration levels and quantitative estimations on particles composition together with their source identification, in order to acquire data for the estimation of their impact, is an effort that has significantly strengthen during the last decade for the Greek urban environments.Results from systematic monitoring of PM 10 concentrations and composition determinations are the most frequent in literature for Greek urban environments and have shown that the EU limit of 50 μg/m 3 is very often exceeded (Chaloulakou et al., 2003;Gerasopoulos et al., 2006;Grivas et al., 2008;Terzi et al., 2010;Theodosi et al., 2011).Results on PM 2.5 concentration levels and composition are few and recent and have also shown frequent exceedances of the annual target value of 25 μg/m 3 (Manoli et al., 2002;Vassilakos et al., 2005;Sillanpaa et al., 2006;Karageorgos and Rapsomanikis, 2007;Lazaridis et al., 2008;Theodosi et al., 2011).The relative contribution of local and long range transported particulate matter as well as the relative contributions of natural and anthropogenic sources on fine atmospheric particulate matter are critical issues for the greater Athens area in which is accumulated more than a half of the population of the country.Every new data set on quantitative estimations of aerosol components is necessary and will significantly contribute in order to complete and clarify the picture.
As far as the long range transported atmospheric particulate matter is concerned, more than half of the air masses arriving in Athens, originate from the North sector, which covers Central and Eastern Europe as well as part of the western Turkey.This northern contribution reaches almost 85% during summer.On the other hand, Saharan dust transport events are very frequent in the area during spring and autumn contributing up to 25% of the prevailing air masses (Borge et al., 2007;Mihalopoulos et al., 2007;Kassomenos et al., 2010;Markou and Kassomenos, 2010;Gerasopoulos et al., 2011).
The present paper reports results on the composition of PM 2.5 from 24-hour PM 2.5 samples at an urban site in Athens, 14 m above ground level, at the National Technical University of Athens campus, collected from February to December 2010.Total sulfur, crustal origin elements and elements of a major crustal component (Al, Si, Fe, Ca, K, Mg, Ti) trace elements (Mn, V, Pb, Cu, Zn, Ni) and water soluble ions (Cl -, NO 3 -, SO 4 2-, Na + , K + , NH 4 + , Ca 2+ , Mg 2+ ) were determined in PM 2.5 samples.Carbonaceous compounds (OC, EC) were also determined for a period of one month.The composition of PM 2.5 is discussed in terms of elemental, ionic and EC/OC concentrations levels, their relative contribution in PM 2.5 and their correlations.For the ionic component, the ionic balance is presented and the neutralization capacity of the atmospheric acidity is discussed.Finally, mass closure is attempted considering the following components: Secondary Inorganic Aerosol (SIA), Organic Matter (OM), Elemental Carbon (EC), Dust -Mineral anthropogenic component (MIN) and Sea Salt (SS).

PM 2.5 Sampling and Concentrations Determinations
Twenty four hours aerosol samples were collected at the top of the building of the School of Mining and Metallurgical Engineering at the National Technical University of Athens campus at 14 m above ground level.A hundred sixteen samples corresponding to about 10-15 samples per month have been collected from February to December 2010.The sampling point is fully exposed to wind and free all around of other obstacles (Remoundaki et al., 2011).PM 2.5 sampling was carried out using a TCR TECORA (Sentinel PM) operating at 38.33 L/min, constructed and calibrated in order to comply with European Standard EN14907 for standard sampling of PM 2.5 .The sampling device operates with autonomy of 16 samples loaded in a sequential cassette holder by programming the sampling span and duration.Aerosol samples were collected on PTFE membranes (PM 2.5 air monitoring membranes Whatman).For organic carbon (OC) and elemental carbon (EC) determinations high purity quartz filters were used prefired at 450° for 12 h.
Sampling material and filter keeping petri-dishes were pretreated by soaking in dilute nitric acid solution and thorough rinsing by ultra-pure water (18 MΩ/cm) and dried under the laminar flow hood of the laboratory.In order to determine PM 2.5 concentrations, the membranes were weighted before and after sampling according to the procedure described in EN12341 (Annex C) using a Mettler Toledo MS105 with a resolution of 10 μg in the air conditioned weighing room of the laboratory.
The pre weighted membranes were loaded to the filter supports and sampler cassette under the laminar flow hood.Filter blanks and blank field samples were also prepared and analyzed together with samples.

Analytical Techniques Elemental Concentrations Determinations by WDXRF
The elemental composition determinations for Al, Si, Fe, Ca, K, Mg, Ti and S have been carried out by WDXRF (Thermo ARL Advant XP, sequential XRF) with Rh X-ray tube at 30 kV, 30 mA.Single element standards purchased from Micromatter have been used for calibration for each element.Two NIST standard SRM 2783 have been also used as calibration points and for calibration verification.The elements Si, Al, Fe, K, Ca, Mg, Ti and S have been determined.The detection limits were calculated to be: Al: 100 ng, Si: 300 ng, Fe: 300 ng, Ca: 40 ng, S: 20 ng, Ti: 40 ng, K: 20 ng, Mg: 100 ng.The estimated precision of the method was estimated to be < 5%.
Following WDXRF analysis, one half of the filter was extracted with 10 mL of ultrapure water for ion chromatography analysis and acid digestion was performed to the second part for trace metals analysis.The extracted samples were kept refrigerated and analyzed by ion chromatography within one week after extraction.Acid microwave digestion (Prolabo Microdigest,401) in PTFE bombs was performed to the second half of the filter using Nitric acid 50:50 (suprapur Merck) for trace metals analysis by ICP-MS.

ICP-MS
Trace metals: Zn, Mn, V, Pb, Cu, Ni have been determined by ICP-MS Perkin Elmer (Elan 6100), Dual Detector Mode: Pulse.The calculated detection limits were: 20 μg/L for Zn and 5 μg/L for the other trace elements.

Organic Carbon (OC) and Elemental Carbon (EC) by Thermal Optical Transmission
Fifteen samples were collected from March 16 th to April 19 th 2010 on quartz filters for organic carbon (OC) and elemental carbon (EC) analysis.Organic carbon (OC) and elemental carbon (EC) analysis concentrations were determined on quartz filters by Thermal Optical Transmission method in a Sunset Laboratory OC/EC Analyser using EUSAR2 protocol (Koulouri et al., 2008;Terzi et al., 2010).

Air Mass Trajectories and Local Meteorological Data
Four ending air mass backward trajectories were calculated for the Athens site using the Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT) to gather information about the origin of the observed aerosols and the synoptic patterns corresponding to the period under study.The calculations were made for the arrival heights of 750, 1500 and 3000 m a.s.l.over Athens, Greece (NTUA site).
Wind roses have also been calculated using the local meteorological data from the ground meteorological station of NTUA.

Elemental, Ionic and EC/OC Concentrations Levels
Table 1 presents mean, minimum and maximum values of the elemental, ionic and elemental and organic carbon concentrations determined in the present study.The mean value of daily PM 2.5 concentrations of 20 μg/m 3 for the whole period is also reported on Table 1.Similar or higher values were reported for Athens: 21 μg/m 3 , for a suburban sampling point (Vassilakos et al., 2005), 23.5 and 29.4 μg/m 3 for downtown sampling points (Theodosi et al., 2011), 40.5 μg/m 3 32 μg/m 3 for high traffic downtown sampling points at ground level and 25 m above ground level respectively (Karageorgos and Rapsomanikis, 2007).Because of the choice of the sampling site at 14 m above ground level, it is expected that our results may represent lower values than those reported for ground level and downtown sampling sites.Most of the recorded values approached 20 μg/m 3 .PM 2.5 concentrations equal and/or higher than 25 μg/m 3 corresponded to thirteen periods of 1-5 days representing 22% of the values determined.Unusually high PM 2.5 concentrations occurred between February 17 th and 20 th where PM 2.5 concentrations reached 100 μg/m 3 due to an intense Saharan dust transport event.
From Table 1, it is apparent that the dominant element in PM 2.5 is sulfur.Sulfates and carbonaceous material represented by elemental and organic carbon (EC and OC) are major fractions of PM 2.5 .If the mean values reported on Table 1 are summed, sulfates represent 35% of the mean total mass of the determined species and OC + EC represent 30%.These results are in agreement with other studies on PM 2.5 composition in Athens and other urban areas in the world (Borge et al., 2007;Karageorgos and Rapsomanikis, 2007;Theodosi et al., 2011).For elemental and organic carbon, recent results from PM 10 monitoring at a down town urban site in Athens reported average concentrations of 2200 and 6800 ng/m 3 respectively (Grivas et al., 2012), and similar values are reported by Terzi et al. (2010) for Thessaloniki.The ratio of OC/EC calculated from our samples gives a mean value of 2.92 (± 0.6) showing a clear prevalence of organic carbonaceous species over EC which indicates significant secondary aerosol formation.Other major elements include Si, Ca, Al, Fe which are typical dust constituents.They strongly correlate among them (Table 2(a)), their concentrations presented the highest variability among all the elements determined with maxima corresponding to Saharan dust transport events.Titanium, potassium and magnesium strongly correlated with these elements (Table 2) and their variability has been also found strongly influenced by Saharan dust transport events.(Karageorgos and Rapsomanikis, 2007;Karanasiou et al., 2007;Theodosi et al., 2011).Lead concentrations have further decreased through the last decade since non-catalyst equipped vehicles were gradually removed from circulation.
Some interesting results shown on Table 2(a), merit to be commented.Sulfur correlated significantly only with lead and did not exhibit a positive correlation with typical dust constituents.Trace metals exhibited positive correlations among them.Manganese correlated positively both with the typical constituents of dust but also with sulfur and zinc since this element has both natural and anthropogenic origin.Significant correlation was also obtained between lead, potassium and sulfur due to their co-existence in fuels.Metals like zinc, nickel and vanadium have shown positive correlations with elements which are typical dust constituents because their concentrations may be enhanced during dust transport episodes.
From the determinations of the ionic concentrations in our PM 2.5 daily samples (Table 1), sulfates are by far the most abundant, followed by ammonium.Mean values of all ionic species are also in very good agreement with those reported for PM 2.5 for Athens (Karageorgos and Rapsomanikis, 2007;Theodosi et al., 2011).A detailed discussion on ionic composition results will be presented in following section.The ratios of water soluble (ionic) concentration over the total elemental concentration [S as SO 4 2-]/[S], [Ca 2+ /Ca], [Mg 2+ /Mg] and [K + /K] have been found to be on average 0.90, 0.80, 0.90 and 0.65 respectively.These ratios indicate the dominance of the water soluble fraction of these species and are in very good agreement with those reported by Karageorgos and Rapsomanikis (Karageorgos and Rapsomanikis, 2007).

Temporal Variability of the Atmospheric Concentrations of the PM 2.5 Constituents
No significant seasonal variability was deduced from the measurement of PM 2.5 concentrations which is also in agreement with previous studies (Vassilakos et al., 2005;Theodosi et al., 2011).
Fig. 1 compares the temporal variability of the daily concentrations for two elements of different origin: (a) sulfur which has a net anthropogenic origin and significant local sources emitting constantly during the year and (b) aluminum which is a tracer of dust and its concentrations are significantly influenced by sporadic episodes of Saharan dust transport.We observe that the temporal variability of the concentrations of these elements is completely different.Sulfur concentrations present higher values during spring and summer: 1699 (± 632) ng/m 3 and lower values during autumn and winter: 1076 (± 529) ng/m 3 , a behavior which is confirmed for urban sites.The higher values for the warm season are associated to enhanced photochemistry, lack of precipitation, low air masses renovation at regional scale or the increment of the summer mixing layer depth favoring the regional mixing of polluted air masses.Moreover, during summer, long range transported sulfur from northern regions is added on top of the locally emitted (Mihalopoulos et al., 2007;Terzi et al., 2010;Theodosi et al., 2011).
As it can be seen from Fig. 1, aluminum concentrations do not follow such a seasonal pattern and present abrupt increases mainly during spring and autumn.These abrupt increases are associated to Saharan dust transport events occurring during spring and autumn (Fotiadi et al., 2006;Gerasopoulos et al., 2006;Kalivitis et al., 2007).From the air mass trajectories, it was obvious that all [Al] maxima were associated to Saharan dust transport events.The episode of February 17 th was of exceptional intensity, as it can be seen from Fig. 1, and corresponded to daily PM 2.5 concentrations as high as 100 μg/m 3 .
We may therefore distinguish two main groups of major elements-species: (a) The first group includes species of a net anthropogenic origin mainly emitted from local sources in an urban environment such as sulfur, sulfates, carbonaceous material, and ammonium.For these major PM 2.5 constituents, a seasonal pattern of higher values during the warm period was observed (e.g., for sulfur and sulfates).Recently reported values of EC and OC in PM 10 for Athens also demonstrated this seasonal pattern especially for OC which is also strongly influenced by photochemistry during the warm season, while a seasonal pattern for EC is less obvious due to the stable vehicular traffic which is the dominant EC source (Grivas et al., 2012).As far as nitrates are concerned, the seasonal trend is opposite to that for the other elements-species of anthropogenic origin with higher values during winter (a mean value of 600 ng/m 3 ) and lower values during summer (300 ng/m 3 ).The higher values of nitrate during winter are due to the instable ammonium nitrate formation, while ammonium nitrate formation does not occur during summer because of the higher temperature of the air (Theodosi et al., 2011).(b) The second group includes the elements which are typical constituents of dust: Si, Al, Ca, Fe, K, Mg, Ti, strongly correlated among them (Table 2(a)) and their temporal variability is strongly influenced by Saharan dust transport episodes implying concentration maxima during spring and autumn.

Ionic Balance and Neutralization of Atmospheric Acidity
Fig. 2(a) presents the mean relative contribution of the main inorganic ionic species for the period of sampling.Sulfates, ammonium and nitrates represent 79% of the total mass of inorganic ions determined.Sulfates percentage is at the same level to that reported for the fine fraction and the Eastern Mediterranean region, and taking also into consideration the sulfates levels reported for urban and background sites in the region, it is apparent that sulfates levels above Greece are also largely controlled by longrange transport and processes evolving at a large spatial scale (Bardouki et al., 2003;Koulouri et al., 2008;Theodosi et al., 2011).The mean percentage of ammonium determined in our samples is relatively low, indicating an ammoniumpoor aerosol system (Karageorgos and Rapsomanikis, 2007).
The ionic balance may be used to determine potentially missing ionic species.For this purpose, ionic balance was calculated and the obtained results are shown on Fig. 2(b) where the regression line corresponds to samples not affected by dust events.On the same figure the points corresponding to dust events are also shown.
The slope of the regression line is equal to 1.3, meaning that there is a cation deficit indicating presence of acidity (H + ) not determined by ion chromatography.These results are in very good agreement with results for the fine fraction previously reported (Karageorgos and Rapsomanikis, 2007;Theodosi et al., 2011).The points corresponding to dust events and mainly the intense Saharan dust transport event which occurred between February 17 th and 20 th were characterized by high calcium concentrations and a net excess of sum of cations.In this case, the anion deficit can be explained by the presence of carbonates (not determined in our study) due to an enhanced concentration of calcium carbonate (Coz et al., 2009;Remoundaki et al., 2011).
Table 2(b) reports correlation coefficients between the major inorganic anions and cations.The correlation between NH 4 + and SO 4 2-indicates that NH 4 + is an important neutralizing agent of sulfates.However, the mean percentage of ammonium was found to be relatively low, indicating an ammonium-poor aerosol system.Only partial neutralization of sulfates by ammonium may be expected since the molar ratio NH 4 + /SO 4 2-was lower than 2 and presented a mean value of 1.2.The respective correlation between NH 4 + and NO 3 -was low confirming the prediction that formation of NH 4 NO 3 is not expected to occur to a significant extend.NO 3 -correlated well with Mg 2+ and Ca 2+ which is indicative of formation of salts like Mg(NO 3 ) 2 and Ca(NO 3 ) 2 .These results are in perfect agreement with those reported by Karageorgos and Rapsomanikis (Karageorgos and Rapsomanikis, 2007) for the fine fraction.Finally, strong correlations are observed between Na + and Cl -and Na + , K + and Mg 2+ as expected, indicating the presence of NaCl, KCl and MgCl 2 from both sea salt and dust contributions.

Mass Closure and Origin of Air Masses
For the purpose of chemical mass closure the following components have been considered: Secondary Inorganic Aerosol (SIA), Organic Matter (OM), Elemental Carbon (EC), Dust, Mineral anthropogenic (MIN) and Sea Salt (SS).
Secondary Inorganic Aerosol (SIA) is represented by the sum of nss-SO 4 2-, NH 4 + and NO 3 -.The amount of dust was calculated from the determined concentrations of Si by considering that Si/dust = 15.8% (Andreae et al., 2002;Guieu et al., 2002).
Because the elements Na, Ca, Fe, and K have a mixed origin, their elemental ratios to Al for Saharan dust aerosols were used: for Na/Al = 0.11, Ca/Al = 1.25, Fe/Al = 0.63 and K/Al = 0.19, Mg/Al = 0.3 (Andreae et al., 2002;Guieu et al., 2002) in order to calculate their dust component.
The anthropogenic contribution of Fe, Ca and K was included as a mineral anthropogenic fraction (MIN) by subtracting from the total elemental concentration the contribution of dust.The anthropogenic component of these three elements was then multiplied by the appropriate factor to pass from the element to the corresponding oxide and summed according to Eq. (1).MIN = 1.95Ca anthr + 1.43Fe anthr + 1.21K anthr (1) Ca was multiplied by a factor of 1.95 to account equally for CaO and CaCO 3 which are considered as its most abundant forms.
Particulate organic matter was estimated by multiplying OC by a conversion factor which corresponds to the organic molecular carbon weight per carbon weight, and form previous work in the area, was found to range from 1.6 to 2.1 (Sciare et al., 2005).In this work a conversion factor of 1.6 was chosen to be in accordance with other studies for Greek urban environments and Athens (Terzi et al., 2010;Grivas et al., 2012).Since OC and EC were determined in our samples for a period of one month from March 16 th to Aril 19 th , mass closure refers to that period.
Sea salt was calculated by using Na + concentrations measured in our samples multiplied by 3.248 (Brewer, 1975).Since during the sampling period, PM 2.5 concentrations were significantly influenced by Saharan dust transport events, sodium concentrations were corrected before this conversion from the dust contribution in them.
The results on the mass closure, reported on Table 3, show that the sum of the considered fractions justified about 73% (± 12) of the measured PM 2.5 concentrations, which is in good agreement with values already reported in literature for the fine fraction in urban environments and Athens (Sillanpaa et al., 2006).The remaining as unidentified may be attributed to aerosol bound water estimated by Tsyro (Tsyro, 2005) to vary between 20 and 35% of PM 2.5 .Finally, another source of systematic error may be the OC multiplication factor.
According to the results shown on Table 3, SIA and carbonaceous material (OM + EC) contributed almost equally (30%) in the measured PM 2.5 concentrations explaining more than 60% of the measured PM 2.5 concentrations, a result in good agreement with values already reported in literature for the fine fraction in urban environments and Athens (Sillanpaa et al., 2006;Karageorgos and Rapsomanikis, 2007;Grivas et al., 2008).
The period between March 16 th -April 19 th for which the mass closure was attempted, was characterized by absence of rain events, frequent south winds and dust transport events for three periods: On March 23 rd -25 th , March 30 th -April 3 rd , and April 17 th -19 th as shown in Fig. 3 from the wind roses and the air mass trajectories presented.
These events explain the increase in dust contribution in PM 2.5 on the corresponding dates (Table 3).In general, if the periods of Saharan dust transport events are excluded from our data set, dust contribution in PM 2.5 was calculated to be about 5%.
Finally, the mineral anthropogenic contribution also accounts for about 3%.
Fig. 4 presents the contributions of SIA, dust, mineral fraction of anthropogenic origin and sea salt for the sampling period.From this figure, it is apparent that SIA contribution is close to 30-40% for most samples.Dust contribution becomes significant in cases of Saharan dust transport events while the anthropogenic mineral fraction contribution was calculated to be about 2%.Sea salt contribution may enhance simultaneously with dust since the air masses transporting desert dust mix with maritime air masses.A mean value of 3.6 (± 3)% for sea salt contribution in PM 2.5 was calculated from the whole data set in the present study.

CONCLUSIONS
Our results on the determination of the composition of PM 2.5 from 24-hour samples in an urban site at the University campus of NTUA have shown that the most abundant constituents of PM 2.5 are sulfur (sulfates) and carbonaceous material.The elemental, ionic and EC/OC levels are in good agreement with results reported in literature for Athens and other urban environments.The concentrations of the elements having crustal origin and the elements with significant crustal component presented abrupt increases during Saharan dust transport events.Trace metals concentrations were found in low levels.The ratios of water soluble (ionic) concentration over the total elemental concentration [S as SO 4 2-]/[S], [Ca 2+ /Ca], [Mg 2+ /Mg] and [K + /K] have been found to be on average 0.85, 0.80, 0.90 and 0.65 respectively indicating the dominance of the water soluble fraction of these species.As far as the temporal variation of the atmospheric concentrations is concerned, PM 2.5 concentrations did not present significant seasonal variation while sulfate concentrations have shown a seasonal pattern with higher values during the warm season and lower values during the cold season.During the warm season, long range transported sulfur from northern regions added on top of the locally emitted, enhanced photochemistry, lack of precipitation, low air masses renovation at regional scale and the increment of the summer mixing layer depth which favors the regional mixing of polluted air masses, are main reasons for the observed higher concentrations of sulfur/sulfates (Mihalopoulos et al., 2007;Terzi et al., 2010;Theodosi et al., 2011).A similar pattern to that of sulfates is also reported for organic carbon.
From the calculation of the ionic balance, a cation deficit indicated presence of acidity except for the samples corresponding to the Saharan dust transport episodes where a net excess of cations was recorded due to the presence of significant amounts of calcium carbonate.The correlation between NH 4 + and SO 4 2-indicated that NH 4 + is an important neutralizing agent of sulfates.However, the mean percentage of ammonium was relatively low, indicating an ammonium-poor aerosol system.The respective correlation between NH 4 + and NO 3 -was low confirming the prediction that formation of NH 4 NO 3 is not expected to occur to a significant extend.NO 3 -correlated well with Mg 2+ and Ca 2+ which is indicative of formation of salts like Mg(NO 3 ) 2 and Ca(NO 3 ) 2 .
Mass closure was attempted for a period of one month.For the mass closure, Secondary Inorganic Aerosol (SIA = nss-SO 4 2-+ NH 4 + + NO 3 -), Organic Matter (OM), Elemental Carbon (EC), Dust, a mineral anthropogenic fraction (MIN) and Sea Salt (SS) have been considered.The sum of the considered fractions justified about 73% (± 12) of the measured PM 2.5 concentrations.The remaining as unidentified may be mainly attributed to aerosol bound water estimated to vary between 20 and 35% of PM 2.5 .SIA and carbonaceous material (OM + EC) contributed almost equally (30%) in the measured PM 2.5 concentrations explaining more than 60% of the measured PM 2.5 concentrations, a result in good agreement with values already reported in literature for the fine fraction in urban environments and Athens.Dust contribution was higher during dust transport events while, dust contribution in PM 2.5 was calculated to be about 5% in absence of dust events.Sea salt contribution was about 3.6 (± 3)% for the whole period of sampling, a result expected since sea salt particles are mainly associated to the coarse fraction.Finally, anthropogenic mineral fraction contribution was calculated to be about 2%.

Fig. 1 .
Fig. 1.Temporal variability of S and Al concentrations in PM 2.5 in ng/m 3 .

Fig. 2 .
Fig. 2. (a) Mean relative contribution of the main inorganic ionic species (b) Ionic balance (equivalents), the points corresponding to dust events are also shown on the figure.

Fig. 3 .
Fig. 3. Wind roses and air mass trajectories ending over Athens on March 22 nd , 31 st and April 19 th , 2010.

Table 1 .
Arithmetic mean and range of the PM 2.5 concentrations, elemental, ionic and OC/EC concentrations.

Table 2 .
The correlation coefficients (r 2 ) a) between the elements b) between the major inorganic anions and cations.