Investigation of Desert Dust Contribution to Source Apportionment of PM 10 and PM 2 . 5 from a Southern Mediterranean Coast

In order to identify the source apportionment of particulate matter PM10 and PM2.5 in the southern Mediterranean coast of Tunis (Tunisia), three different sites characterized respectively by traffic, industries and being an urban background area are studied. The chemical characterization included a gravimetric determination of atmospheric particles mass concentration, measurements of the major anions (SO4, NO3, Cl) and cations (Ca, Mg, K, NH4) concentrations in the aerosol samples by ion chromatography and analysis of 18 elements by energy dispersion X-ray fluorescence. Aerosol ion balance of various PM10 constituents are used to identify possible sources of the particulate matters. Thanks to these analysis, the particulate masses were reconstructed from the main possible constituents: crustal matter, primary and secondary pollutants, marine aerosols and organic matter. Wherever PM10 and PM2.5 were studied, PM10 crustal elements and sea salt aerosols were mainly associated with the coarse fraction whereas primary and secondary anthropogenic pollutants as well as organic matter rather compose PM2.5 fraction. In all the sites, PM10 mass was mainly composed of crustal matter (41–50%) and poorly of sea salt (3–4%). And so aerosols chemical composition is heavily affected by dust winds from Sahara desert, with some contribution of local traffic and industries and only a slight direct impact of the nearby Mediterranean sea.


INTRODUCTION
Aerosols are important components of the earth system.The composition of atmospheric aerosols impacts on many parameters, among them the decreased visibility (Watson, 2002), deposits of pollutants affecting the ecosystems (Bytnerowicz and Fenn, 1996), tropospheric toxicity and human health (Carter et al., 1997;Peters et al., 2001;Sattler et al., 2001, Pope andDockery, 2006;Oesterling et al., 2008;Hsieh et al., 2009), the hygroscopic and optical properties of the aerosol (Tsai and Kuo, 2005;Tsai et al., 2007), the earth's radiation budget, the global climate (Facchini et al., 1999;Charlson et al., 2001;Acker et al., 2002) and the direct and indirect effect of aerosols on the planet energy balance (Seinfeld and Pandis, 1998;Cabada et al., 2004;De Carlo et al., 2008).Yet the mass size distributions and chemical characteristics of the aerosols are still insufficiently understood.
The need for analysis of Particulate Matter (PM) sources contributing to PM 10 and PM 2.5 atmospheric concentrations is particularly strong in Tunis, situated in North Africa, in the southern Mediterranean coast.Specific meteorological circulations and natural sources like Mediterranean sea and the proximity of Sahara create specific patterns of aerosol concentrations that could influence not only the particulate concentrations through Europe but also global climate changes due to the migration of Sahara desert dust.
The aim of this paper is to investigate the origin and the variability of PM 10 and PM 2.5 in the urban area of Tunis.It involves the measurement of the mass concentrations and the different chemical compositions of atmospheric suspended particulate matters.It compares the particulate matter characteristics in three different areas and estimates the source apportionment of the particulate matters in this area.

STUDY AREA
The study area is located in Tunis city, the capital of Tunisia in Northern Africa.Three urban sites were chosen for their specificities (Fig. 1).The near white grey level in Fig. 1 covers the areas outside Tunis.The other shades of grey increases with the population density inside Greater Tunis.
Site 1 (Bab Saadoun) is located at a big crossroads in the center of Tunis (latitude 36, 80769°N, longitude 10, 15941°E).It is a point of high density traffic roads and congestions all day long.Site 2 (Ben Arous) is located in Tunis suburb, about 10km south of Tunis center, in an industrial area (latitude 36, 74120°N, longitude 10, 19161°E).The industrial activities include a fuel oil power station, a cement factory and several chemical, food and welding factories.Site 3 (El Mourouj) is also located in the southern suburb, about 12 km from Tunis, inside a city park surrounded by a crowded dormitory area (latitude 36, 74120°N, longitude 10, 27442°E).

Sample Collection and Analysis
The sampling of PM 10 and PM 2.5 was performed in 2008 from June 6 to 27, 24 hours a day, 7 days a week.The filters were changed every day.There were 21 samples from each site for PM 10 and additionnal 21 samples for PM 2.5 in Bab Saadoun.Of the 85 samples, only three were missing (one in Ben Arous and two in El Mourouj).
Daily sampling of PM 10 and PM 2.5 was carried out by a dual-channel sequential sampler operating at a flow rate of 2.3 m 3 /h.Sampling heads met the specifications of the European Standard (EN12341, 1998).The sampler (HYDRA Dual Sampler -FAI Instruments) has two independent simultaneous sampling lines for PM 10 and PM 2.5 .Each sampler was equipped with two filters: a PTFE membrane and a quartz fiber membrane.
The sampler was placed inside constant temperature housing, set at 20°C in order to reduce the loss of volatile chemical species from the collected particles (Perrino et al., 2008b, c).The gravimetric determinations of the PM 10 and PM 2.5 mass concentrations were carried out on the PTFE filters in the sampler by β attenuation method.After these samplings and measurements, the filters were stored at 5°C for further analysis.

Chemical Composition of PM 10 and PM 2.5
The PM on the PTFE filters were analyzed by energydispersion X-ray fluorescence (ED-XRF, X-Lab 2000, SPECTRO) for Na, Mg, Al, Si, S, Cl, K, Ca, Fe, As, Cr, Cu, Mn, Ni, Pb, Ti, V, Zn, yielding results directly in mass concentrations (µg/m 3 ).Some metals (As, Cr, Cu, Mn, Ni, Pb, Ti, V, Zn) were not used in the estimation of PM since only the major metals were used to calculate the reconstructed masses.Afterwards, PTFE filters were ultrasonicated two times in 5 mL ultra-pure water.The resulting solution was analyzed by ion chromatography (DX120, Dionex) for anions (chloride, nitrate, sulphate) and cations (sodium, ammonium, potassium, magnesium, calcium).The ion chromatography conditions are described in Table 1.The insoluble fraction (Cins) of Na, K, Mg and Ca was calculated as the difference between the concentration resulting from XRF (total fraction) and the concentration resulting from ion chromatography (soluble fraction only).
Quartz filters were analyzed for their elemental carbon and organic carbon content (EC/OC) using a thermo-optical analyzer integrating a flame ionization detector (OCEC Carbon Aerosol Analyzer, Sunset Laboratory).The filter is first heated in helium atmosphere up to 870°C for organic carbon compounds analysis then up to 900°C in helium and oxygen atmosphere for the determination of elemental carbon compounds.
Previous studies have shown that in most environments this overall procedure allows the identification of more than 90% of the particulate mass (Perrino et al., 2009b(Perrino et al., , 2010)).
The cation and anion in all the PM samples were calculated as follows: where the ions are in µg/m 3 Crutal matter (CM, sea salt aerosol (SSA), primary anthropogenic pollutants (PA), organic matter (OM) and inorganic secondary species (IS) cannot be measurd, they are estimated by the use of the previously analysed elements (by XRF) and ions (by ionic chromatography).
Crustal matter (CM) was estimated (Eldred et al., 1987;Chan et al., 1997) by the following equation, which takes into account the main elements considered as components of the earth's crust: [CM] = oxides + carbonates Seven elements are assumed to form the oxides.Their concentration was multiplied by the ratio between the molecular weight of the oxide and the molecular weight of the element.Calcium and magnesium carbonates were calculated as the soluble (sol) fractions of Ca and Mg, added to the calculated CO 3 2-concentrations: and so Sea-salt aerosol (SSA) was estimated by the following equation using sodium and chloride and estimating the minor constituents S, Mg, Ca and K through a multiplicative factor (Perrino et al., 2008a, b): Primary anthropogenic pollutants (PA) were estimated as the elemental carbon concentration (EC) plus the primary organic carbon content estimated by the elemental carbon multiplied by 1.1 (Viidanoja et al., 2002): Secondary organic matter (OM) was estimated as the remaining amount of organic carbon content multiplied by a factor α that takes into account the non-carbon component of organic molecules.The factor α is the average molecular weight per carbon weight ratio studied by Turpin and Lim (2001).They state that the currently used ratio of 1.4 is the lowest reasonable ratio for an urban aerosol and that a ratio of 1.4 is too low for nonurban aerosols.Based on the above evaluation, they recommend using a ratio of 1.6 ± 0.2 for urban aerosols and 2.1 ± 0.2 for aged (nonurban) aerosols.Considering the detailed study of these authors, in our calculations, this factor α was set to 1.6 for the urban sites 1 and 2 and to 1.8 for the urban background site 3: Inorganic secondary species (IS) were estimated (Wang and Shooter, 2001) as the sum of ammonium, nitrate and non sea-salt (nss) sulphate: The reconstructed mass considers all five origins:

Mass and Water Soluble Ions Concentrations in PM 10 and PM 2.5
Three sites around Tunis city (Tunisia, Africa) are analyzed for their PM 10 .They are chosen for their characteristics: Site 1 as a traffic site, Site 2 as an industrial site and Site 3 as an urban park in a dormitory area.Site 1 is also studied for its PM 2.5 .For all three sites (Table 2), if the variability (SD) of the results is taken into account, the mass concentrations of PM 10 are very close, around 56 µg/m 3 .Although the characteristics of the sites are notably different, all the concentrations of the studied water soluble ions (chloride Cl -, nitrate NO 3 -, sulfate SO 4 2-, sodium Na + , ammonium NH 4 + , potassium K + , magnesium Mg 2+ and calcium Ca 2+ ) are similarly close.
The higher Cl -concentrations in PM 10 were found at Site 2 the nearest site to the sea.It is known that the spatial variability of Na + and Cl -in atmospheric aerosols are mainly derived from the marine source, more commonly referred to as sea-salts (Quinn et al., 2004;Kumar et al., 2008).This indicates that all the sites experienced a certain marine influence.In the PM 10 samples, Ca 2+ , SO 4 2-and NO 3 were the most abundant ions, and Ca 2+ alone accounted for 35.0% of PM 10 mass, ranking at the top of the ions for the three sites.At Site 1 (Bab Saadoun) ion masses were in the order Ca 2+ = SO 4 2-> NO 3 -> Na + > Cl -= NH 4 + > K + > Mg 2+ while for Site 2 (Ben Arous) and at Site 3 (El Mourouj) ion masses were in the order Ca 2+ > NO 3 -> SO 4 2-> Na + > NH 4 + > Cl -> K + > Mg 2+ .Prior studies have shown that high Ca 2+ occurs during dust storms (Choi et al., 2001;Arimoto et al., 2004;Cao et al., 2005b;Shen et al., 2007;Matassoni et al., 2011;Galindo et al., 2013).This calls, in first hypothesis, for the possibility of these particulate matter to mainly originate from Sahara dust.The concentrations of secondary ions SO 4 2-and NH 4 + , during the sampling period, were the lowest in PM 10 .They accounted for 31% of the total PM 10 ion concentrations at Site 1, 28% at Site 2 and 23% at Site 3.
In site 1, the mean mass concentrations analyzed for PM 10 and PM 2.5 were 54 µg/m 3 and 25 µg/m 3 (Table 2) respectively and the ion masses order for PM 2.5 was different from the one for PM 10 .In PM 2.5 , the ion masses order was SO 4 2-> NH 4 + > Ca 2+ > Na + > NO 3 -> K + = Cl -> Mg 2+ .PM 2.5 contributes to 46% of the PM 10 mass.Chlorides, nitrates, sodium, magnesium and calcium ion mass concentrations are lower in PM 2.5 than in PM 10 suggesting that these ions are concentrated in the coarse fraction, between PM 10 and PM 2.5 .It is not the case for SO 4 2-and NH 4 + of PM 10 that originated entirely from PM 2.5 .These two secondary ions accounted for 66% of the total PM 2.5 ion concentrationsand only 31% of the PM 10 .Other authors (Wang et al., 2006;Pey et al., 2013) has also seen secondary aerosol species being the most abundant of the ions studied in PM 2.5 .The relatively high SO 4 2-concentration may be due to gas to particle conversion of SO 2 over sea regions (Reddy et al., 2008, Zhao et al., 2011) into H 2 SO 4 (Stockwell and Calvert, 1983) which may further react with atmospheric NH 3 resulting in (NH 4 ) 2 SO 4 aerosols (Seinfeld and Pandis, 1998).High SO 4 2-concentrations associated with land breeze would also affect significantly the long range transport of SO 2 .In site 1, the PM 2.5 chemical composition domination by SO 4 2and NH 4 + indicates thus the influence of land breeze and various anthropogenic activities.
When compared to other cities in the world, either on the Mediterrannean coast like Bari in Italy (Amodio et al., 2008), on other coasts like Erdemli in Turkey (Kocak et al., 2004(Kocak et al., , 2007) ) and Mangalore in India (Hegde et al., 2007) or in industrial or high traffic areas like Madrid in Spain (Salvadore et al., 2004), Yampa Valley in USA (Waston et al., 2001) and Pamplona in Spain (Aldabe et al., 2011), no prevailing pattern comes out.The only similarity that can possibly be noted is the high PM 10 mass concentrations on our sites (58, 56 and 54 µg/m 3 ) that are quiet close to Calexico's (61.90 µg/m 3 ) in USM (Chow et al., 2001), possibly due to both proximity to a desert land.

Particulate Matter Ion Balance
Ion balance is a useful tool to determine any possibly missing ionic species.The linear regression of total cation equivalents against total anion equivalents for each size class show a linear relationship with a significant positive correlation coefficients between the sum of cations concentrations and that of anions, parallel to the theoretical line 1:1.The deviation from the theoretical line indicates a deficiency of anions, since bicarbonate, organic ions (formate and acetate), F -, NO 2 -, PO 4 3-and Br -were not determined in the present study.
According to the ion balance, PM 10 and PM 2.5 samples are slightly dominated by cations and anions, respectively.A probable relationship between Ca 2+ and anion deficiency implies that CO 3 -is most probably the missing anion in the PM 10 samples whereas significant correlation between SO 4 2-and cation deficiency in the fine fraction is might be a consequence of H + associated SO 4 2-.These results are consistent with those of PM studies conducted by Khoder and Hassan (2008).
The regression coefficient for PM 10 was 0.89 at S1, 0.88 at S2 and 0.84 at S3, and 0.9 for PM 2.5 at site 1, lower than 1, probably due to the missing compounds.Similar values have been reported elsewhere (Tsitouridou and Samara, 1993;Karakas and Tuncel, 1997;Cheng et al., 2000;Tsitouridou et al., 2003;Kocak et al., 2007).Additionnaly, if the sites are ranked by their regression coefficient between anions and cations: site 1 PM 2.5 (Traffic) > site 1 PM 10 (Traffic) > site 2 PM 10 (industrial) > site 3 PM 10 (Urban), it is probably due to the decrease of the missing organic acids and bicarbonate ions in the particulate matters of these sites.

Percent Contribution of Different Water Soluble Ions
In water soluble extract of PM 10 , Ca 2+ was found to be the major ion with its respective contribution of 33%, 35% and 29% for sites 1, 2 and 3 (Figs.2(b), 2(c) and 2(d)).Generally soil is considered to be the main source of Ca 2+ .High calcium concentration may be due to dust transport from the Sahara at the south of the country.NO 3 -is the second major ion for PM 10 in site 2 and 3 whereas SO 4 2-is the second major ion for PM 10 in site 1, in percent by mass concentration.
Nitrate is a secondary aerosol formed by the combination of NH 3 and nitric acid issued from the conversion of NO x .NO x is known to be the most significant precursor of nitrate.Dynamic distributions of nitrate in gas-particle phase had been reported by many authors (Willison et al., 1985;Suzuki et al., 2008).They suggested that the main factor to impact distributions of nitrate in gas-particle phase is temperature (Zhao et al., 2011).Indeed in Tunis, high temperatures, above 40°C, can be reached.
For PM 2.5 in site 1 (Fig. 2(a)), the major ion, in percent by mass concentration, is SO 4 2-(42%).The contribution to PM 10 in the same site for this same ion is only 27%.The SO 4 2-high concentrations in site 1 are probably due to the heavy traffic in this area.The influence of sea spray does not seem to be relevent: Na + , Cl -and Mg 2+ concentrations are quiet low, even though the distance between the sites and the Mediterranean sea do not exceed 20km.

Source Apportionment of Aerosol Particles during the Sampling Period in Three Sites
In order to study the different sources of the PM 10 and PM 2.5 of the three sites and their apportionment, five origins were considered and estimated as mass concentrations: crustal matter (CM), sea-salt aerosol (SSA), primary anthropogenic pollutants (PA), organic matter (OM) and secondary inorganic species (SI).All together, these five concentrations constitue the reconstructed mass concentrations.Each of the five concentrations was calculated using the ions analyzed by ion chromatography and the insoluble elements analyzed by energy-dispersion X-ray fluorescence (Table 2).This reconstructed mass (RM) is compared to the gravimetric mass (GM) issued from the gravimetric experimental determination (Table 3).The calculated and measured values are very close: Pearson coefficient values are 0.95, 0.90 and 0.88 respectively for PM 10 of site 1, site 2 and site 3; and 0.91 for PM 2.5 of site 1.The data show that a very satisfactory reconstruction of the PM mass concentrations can be obtained and used to understand the particulate matter origins for all the sites.
For PM 10 , the mass concentration is mainly composed of crustal matter which constitutes 41%, 44% and 50% at the sites 1, 2 and 3, respectively (Fig. 3).The higher value of the crustal matter (CM) fraction at site 3 (calcium and silicium respective mean concentrations are 8.8 and 2.1 µg/m 3 ), compared to site 1, is probably due to calcium oxide of local origin: site 1 is central urban whereas site 3 is in urban outskirts.These values of CM are high if compared to CM registered in northern Mediterranean cities the like Lazio region (25-30%) in Italy (Perrino et al., 2008a).It is probably due to the higher proximity of the Sahara desert to Tunis, situated in the southern side of the Mediterranean sea.
Primary anthropogenic pollutants, mainly emitted by traffic, is higher at the traffic site 1 (14%) than at the other two sites (9% at site 2 and 6% at site 3).Organic matter fraction (OM) constitute about 30% of the total PM 10 mass at the three sites.This result is similar to the one of Perrino et al. (2009b) at the traffic station in Rome (Italy) and where this contribution decreases to 10% at the urban background station at the urban stations in Latina and Viterbo and at the near-city station of Montelibretti.
The primary anthropogenic compounds (PA) account for 6% of the PM 10 at the urban park in site 3, 11% at the traffic site 1 and 15% at the industrial site 2.They probably originate from local emission, increasing with the level of pollution.
The mass concentration of inorganic secondary pollutants (IS) have very close values, around 10%, at all the stations.This is the typical behaviour of secondary aerosols, which exhibit a quite homogeneous space distribution on a regional scale.
With 4% of total PM 10 mass concentrations, sea-salt (SSA) was the source with the least weight (3-4%) of the three sites of this study even though they are located in a coastal area.
In site 1, primary anthropogenic compounds (Fig. 4) are almost completely (91% of the total PA) in the fine fraction (PM 2.5 ), more than organic matter (68%) and inorganic secondary species (71%).Crustal material is mostly in the coarse fraction (80% of the total CM) as well as sea salt aerosols (81%).During the study, an important variation was recorded for crustal elements, which showed a relevant increase during the last two days of the sampling exceeding 50% of the total PM 10 mass.This increase is recorded for PM 10 at all three stations and at a less extent also for PM 2.5 .This extrem event characterizes and coincide with a desert dust wind event from Sahara.
When the three sites are compared in terms of average mass composition (Fig. 5), it is perceptible that the primary anthropogenic fraction decreases from the traffic to industrial to urban background stations.The difference in the crustal fraction between the values recorded at site 1 and site 3 is probably due to the role played by dust re-suspension, which is traffic-related.In the site 3 the mean mass concentrations are 10 µg/m 3 for Ca and only 2.9 µg/m 3 for Si.The higher value of the crustal fraction at site 3, compared to site 1, is due, in particular, to calcium oxide, probably of local origin.
The concentration of secondary pollutants at the industrial site 3 and background urban stations site 2 is very similar, about 4.9 µg/m 3 .The higher values at the traffic site 1 are probably related to local emissions.The contributions of these elements to total mass of particulate matter were similar but higher than those of desert dust from North Africa found and reported in Southern European countries (Aldabe et al., 2011).Even if there is some contribution of sea salt from the near Mediterranean sea, the majority of the PM 10 is composed of crustal matter originating probably from the near Sahara desert and some primary and secondary pollutants coming from the local traffic and industries as well as a significant fraction of organic matter of various origins.These characterizations would help the stakeholders and the decision makers to take actions against the tropospheric pollution due to particles by knowing its origins.It can also be useful for the implementation of epidemiologic studies and for the understanding of the effects on health and environment of aerosols.It also allows the evaluation of the dispersion of Sahara desert dust, from the south of the country to the north, near the Mediterranean coast and so their movements to further destinations like Europe.It would also be possible to use these results within the framework of the study of global climate change that is known to be influenced by dust and its desertic component.In the same context, modelization of futur climates that needs local atmospheric data that are presently lacking in this area could usefully exploit the above results.

Fig. 2 .
Fig. 2. Ionic composition of PM 10 and PM 2.5 , in percent by mass concentration, at the three sampling sites.

Fig. 3 .
Fig. 3. Mass concentration composition of PM 10 at the three sampling sites.

Fig. 4 .
Fig. 4. Contribution of each source to the three size fractions at the traffic site 1.

Fig. 5 .
Fig. 5. Relative contributions of aerosol species to PM 10 fraction for the three sites.

Table 1 .
Ion chromatography conditions for the analysis of cations and anions in the samples.

Table 2 .
Mean, Standard Deviation (SD)and extreme event (ee) of PM 2.5 and PM 10 mass, ionic, elemental concentrations and estimated sources concentrations at the three sampling sites (µg/m 3 ) *CM: crustal matter, SSA: Sea Salt Aerosols, OM: Organic Matter, PA: Primary Anthropogenic pollutant, IS: Inorganic Secondary species.

Table 3 .
Mean values, standard deviations SD and ranges (in µg/m 3 ) of PM 10 and PM 2.5 at all the sampling sites.GM and RM are the gravimetric and reconstructed mass concentrations, respectively.