Interactive comment on “ PM 1 geochemical and mineralogical characterization using SEM-EDX to identify particle origin – Agri Valley pilot area ( Basilicata , Southern Italy ) ” by S . Margiotta

Abstract. A PM1 geochemical and mineralogical study using Scanning Electron Microscopy (SEM) was performed on a pilot site in the Agri Valley which is close to the oil pre-treatment plant (C.O.V.A) of Europe's largest on-shore hydrocarbon reservoir. The study identified PM1 geochemical and mineralogical characters in the period before, during and immediately after a burning torch flare event. The finer fraction (DFe μm ) consisted mainly of secondary particles and soot. In the coarser fraction (DFe ≥ 0.7 μm ), natural particles originating from crustal erosion and soot were abundant. Fine quartz particles and lower Al / Si ratios are markers for desert dust origin, proving that a Saharan dust episode which occurred during the observation period played a significant role in supplying geogenic aerosol components to the PM1. Largest amounts of ≥ 0.7 μm fraction particles observed on the day of flare event may be due to a greater supply of Saharan geogenic particles. Soot had been significantly increasing long before the flare event, suggesting that this increase is also related to other causes, although we cannot exclude a contribution from flaring. S-rich aerosol consisted mainly of mixed particles originating from deposition and heterogeneous nucleation of secondary sulfates on mineral dust. Only-S particles were identified in the ≥ 0.7 μm fraction following the flare event. These particles may be indicators of larger amounts of sulphur in the atmosphere.


Introduction
In recent years, there has been an increasing scientific interest in atmospheric pollution and its effects on human health. Such interest has focused on atmospheric aerosols and their formation, transformation, transportation and deposition processes. Atmospheric aerosols are composed of solid and liquid particles, originating from anthropogenic sources (such as industrial activities, urban traffic and domestic heating) and natural sources (such as wind erosion of soils, pollen, volcanic eruption and desert dust), which can impact negatively on the environment, air quality and human health (e.g. Migon et al., 1997;Sokolik and Toon, 1999;Nickovic, 2002;Pope et al., 2002;Moshammer and Neuberger, 2003;Jawad Al Obaidy and Joshi, 2006;Pope and Dockery, 2006;Middleton et al., 2008;Kleanthous et al., 2009;Pope et al., 2009;Klein et al., 2010;Nickovic et al., 2012;Paternoster et al., 2014).
In order to obtain a complete typological characterization of atmospheric aerosols and to assess their role in the environment, it is important to apply measurement and characterization methods integrating conventional techniques. Scanning Electron Microscopy with Energy-Dispersed Analysis (SEM-EDX) play a very important role, providing morphological, chemical and mineralogical data fundamental in understanding the formation mechanisms of aerosols and distinguishing between natural and anthropogenic origin.
A PM 1 geochemical and mineralogical characterization was carried out using SEM, in a pilot site in the Agri Valley, to distinguish between the natural and anthropogenic origin of the finer atmospheric aerosols in an area of great envi- ronmental concern due to the presence of the largest European on-shore reservoir and an oil pre-treatment plant. In order to identify anomalies in the geochemical and mineralogical characters of the particles, observations were carried out from 22 September to 01 October 2012, before, during and immediately after a burning torch flare event on 28 September 2012.

Pilot site
The pilot site (40 • 20 8 N, 15 • 54 7 E, 844 m a.s.l.) is located in the village of Viggiano (Potenza, southern Italy), in the vicinity of C.O.V.A., a source of anthropogenic emissions located in the bottom of the valley, more than 200 m downstream (Fig. 1). There are also some industrial settlements nearby, the surrounding zones are mainly rural, agricultural activities are prevalent and there are extended woodlands and pastures in middle mountains, partially protected by the Appennino Lucano -Val d'Agri -Lagonegrese National Park. The area has local road networks, modest traffic volumes and there is a busy motorway (S.S. 598) linking the Agri Valley with the cities of Potenza and Taranto.
From a geological point of view, the area is characterized by a wide range of the lithotypes represented, with variable compositional characters. The mountain ridges of the Agri Valley are mainly composed of white and grey limestone and subordinate dolostone of the Apenninic Carbonate Platform, tectonically overlapped upon radiolarites, siliceous argillites, calcilutites and marls referable to the Scisti Silicei Formation and Galestri Formation of the Lagonegro Units (Scandone, 1971;Carbone et al., 1988Carbone et al., , 1991Pescatore et al., 1999). The area north of the Pietra del Pertusillo Lake, at orographic left of the Agri River, is also characterized by the Albidona Formation (marls, clayey marls and silty clays) and Gorgoglione Flysch (sandstones and clays) terrains (Selli, 1962;Lentini et al., 1987;Carbone et al., 1991). The Agri valley is filled with continental clastic Quaternary units represented by coarsegrained slope deposits and clastic deposits from alluvial and lacustrine environment (Di Niro and Giano, 1995;Giano et al., 2000;Zembo, 2010;Giano, 2011;Gueguen et al., 2015).
This sampling site was chosen because Viggiano is the most populated town in the area, making it a more representative site with respect to potential risk to human health. Some key meteorological parameters were also available, such as atmospheric pressure, temperature, relative humidity and precipitations, which were provided by the Viggiano Civil Protection weather station.
The site has a mountain climate influenced by Mediterranean atmospheric circulation, resulting in dry summers and cold winters with precipitation concentrated in autumn and winter. This high rainfall is due to a proximity to the southwest Lucanian Apennine mountains, one of the wettest zones in Basilicata due to its exposure to Atlantic humid currents (Basilicata Region, 2006).
From 22 September to 1 October 2012, the weather station recorded average temperatures between 18.6 and 25.6 • C, with a peak of 30.9 • C on 29 September. There was no rainfall and average relative humidity ranged between 26 and 62 % (Fig. 2).
During this period the Mediterranean Basin also experienced a dust episode, which affected the studied area (Fig. 3).

Sampling, SEM-EDX analysis procedures and settings
As reported in Caggiano et al. (2010), PM 1 samples were collected using a low-volume gravimetric sampler equipped with a PM 1 cut-off inlet and polycarbonate filters. Sampling time was 24 h (starting from 12:00 p.m.). Each filter was humidity-conditioned in a filter-conditioning cabinet (T = 20±2 • C and RH = 50 ± 5 %) for 48 h, before and after sampling.
Microscopic analyses were carried out using a Field Emission Scanning Electron Microscope (FESEM, Zeiss Supra 40) equipped with an Energy Dispersive X-ray Spectrometer (EDX, Oxford Instruments). Portions of filter (about 0.5 cm 2 ) were attached to aluminum stubs (diameter 12 mm) using carbon sticky tabs and subsequently carbon coated. SEM images were obtained using both secondary (SE) and back- scattered (BSE) electrons. X-ray analyses were carried out using an energy-dispersive Si(Li) detector able to detect elements with Z ≥ 5, nevertheless carbon and oxygen were not taken into account because they are components of polycarbonate substrate and carbon is used for coating the samples. Elemental composition characterization of particles was performed using the Inca Energy 350 Suite software.
Morphological and chemical analyses of particles were performed both manually and automatically. Automatic analyses were possible when an intense and clear signal was collected by BSE. By considering the Feret diameter (D Fe ), used for determining particle sizes (Merkus, 2009), finer particles (D Fe < 0.7 µm) were excluded due to the very weak image provided by the detector, and were only manually analyzed. On coarser particles (D Fe ≥ 0.7 µm) automatic analyses were carried out by using Inca Feature software, presetting a BSE intensity threshold and using the following instrument parameters: working distance of 8.5 mm, acceleration voltage of 20 kV, aperture size of 60 µm and magnification of 15 000 X. X-ray acquisition time was fixed at 20 s. The settings and area layout chosen allowed the detection and analysis, for each sample, of about 500 particles in random fields of view. Mamane et al. (2001) reported that physical and chemical properties of a sample can be well represented by analyzing several hundred particles. Accordingly, 500 particles detected were considered representative of the entire filters, showing a good homogeneous distribution of particulate. Elemental composition and morphological fea- tures of each particle were determined. The error associated with automated analysis, such as overlapping particles, contrast artifacts and sizing error, were corrected or eliminated by both manual off-line data review and manual data reacquisition from selected particles, in order to improve data quality.
Clustering of coarser particles (D Fe ≥ 0.7 µm) was carried out according to the rules indicated by Coz et al. (2009), with some modifications. In particular, we could distinguish between quartz (Si ≥ 90 %) and kaolinite (Si + Al ≥ 90 %), thanks to the use of polycarbonate filters instead of aluminium foils. The rules of Coz et al. (2009) were also considered in identifying sub-categories within the group of aluminosilicates (kaolinite, smectite and illite/mica). The average Al / Si ratio of the ≥ 0.7 µm fraction was calculated for each S. Margiotta et al.: PM 1 geochemical and mineralogical characterization using SEM-EDX filter using the results of the particle elemental characterization. Particles without peaks in their X-ray spectra were reallocated and manually characterized in order to classify them from their morphological features.
Particle concentration on filter was also calculated as a ratio between the number of particles counted and the whole scanned surface.
Manual characterization of finer particles (D Fe < 0.7 µm) was performed using SE images, reducing acceleration voltage to 15 kV and aperture size to 30 µm, and increasing magnification to 25 000 X. About 300 particles were analyzed in each sample. This number is comparable to other studies carried out using the manual procedure (e.g. Paoletti et al., 1999;Ebert and Weinbruch, 2001;Lettino et al., 2012).

Results
The analyses performed allowed us to identify eight main types of particulate: silica particles, aluminosilicates, carbonates, biogenic particles, non biogenic C-rich particles, metal particles, S-rich particles and secondary particles with low-Z elements (Fig. 4). These particles were distributed differently in the ≥ 0.7 and < 0.7 µm fractions. Relative amounts of several particle typologies detected in the coarser fraction (D Fe ≥ 0.7 µm) are shown in Fig. 5. Mineral component consisted of aluminosilicates, silica and carbonates, in this decreasing order.
Silica particles, composed of silica and oxygen alone, were represented by residual quartz (Fig. 4a). Silicates and aluminosilicates (Fig. 4b), the most abundant particles, were mainly composed of silica and aluminum, with variable amounts of calcium, magnesium, sodium, potassium, iron and titanium. For the most part, aluminosilicates consisted of clay minerals, with smectite kaolinite > illite/mica (Fig. 6).
Carbonates included both calcite and dolomite. Calcite (Fig. 4c) was the more abundant phase, however dolomite was also well represented.
Mineral particles were often rounded because of wind erosion and long transport processes; clay particles preserved a plate morphology, but also showed crushed or chamfered edges linked to mechanical impacts due to mobilization and transport by wind.
Non biogenic C-rich particles consisted of soot, clusterlike structures made up of a number of individual spheroids, with a diameter of a few tenths of a nanometre (30-50 nm), combined to form linear or branched-chain amorphous structures (Fig. 4e). Soot results from incomplete hydrocarbons combustion processes and forms at high temperatures by gasto-particle conversion (e.g. Wolff, 1981;Han et al., 2010).  (kaolinite); (c) carbonate particle (calcite); (d) biogenic particle (brocosomes); (e) non-biogenic C-rich particle (soot); (f) metal particle (iron oxide); (g) S-rich particle (gypsum); (h) secondary low-Z element particles.  Metal particles were not frequent and were mainly composed of Fe and Ti. A smaller amount of Cr-and Pb-rich particles were also identified. A few metal particles detected (about 14 %) showed a spherical shape and were almost all composed of Fe or Pb (Fig. 4f).
The S-rich aerosol consisted mainly of polimineralic aggregates and mixed particles composed of secondary regular crystals of calcium sulfates grown directly onto Ca-rich particles (mainly carbonates) and sulphur compounds coating preexisting particles. Calcium sulfates were also present as residual gypsum (Fig. 4g). Smaller amounts of sodium sulfates, potassium sulfates and barium sulfates were recognized. A few droplet shaped particles counting only S (following OSP) were also identified.
In the finer fraction (D Fe < 0.7 µm), particles with low-Z elements (i.e. C, N, O) were prevalent, such as dropletshaped ammonium nitrates (Fig. 4h) and soot. Secondary phases represented by amorphous or droplet-shaped particles, in which the EDX microanalysis detected only sulphur and/or sodium (ammonium sulfates, sodium nitrates, sodium sulfates or their mixtures) were also observed.
Mineral component of the aerosol is less abundant than that observed in the ≥ 0.7 µm fraction, with no change in composition: aluminosilicates, silica particles and carbonates were present in decreasing order. EDX spectra of these particles sometimes show a small peak in sulphur. Calcium sulfates were also present as residual gypsum.

Discussion
The ≥ 0.7 and < 0.7 µm particle size fractions show very different compositional characters with respect to the distribution of natural and anthropogenic components.
In the coarser fraction (D Fe ≥ 0.7 µm) anthropogenic component consists mainly of soot. A few metal particles de- tected show a spherical morphology unequivocally indicative of an anthropogenic origin associated with combustion processes. The remaining metal particles (mainly iron or titanium oxides), show an irregular shape which did not allow the identification of origin. Natural component is dominated by geogenic particles, originating from crustal erosion. They are composed of aluminosilicates (mainly clay minerals), quartz and carbonates (calcite and dolomite), in decreasing amount order. These mineral phases are consistent with lithological and pedological characters in the pilot area, characterized by soils on calcareous, marly or arenaceous substrates, moderately evolved by brunification and removal of carbonates (Basilicata Region, 2006). Biogenic particles (e. g. pollens, vegetable fragments, brocosomes) also contribute to the natural component.
As shown in Fig. 7, S-rich aerosols consist mainly of mixed particles with a composed origin, as it can originate from deposition and heterogeneous nucleation  of secondary sulfates (anthropogenic component) on mineral dust (natural component). Formation of CaSO 4 regular crystals accreted upon the surface of the pre-existing carbonate phases can result from interaction between H 2 SO 4 and rich-Ca mineral particles (mainly carbonates), according to the following reaction (Harrison and Kitto, 1990;Clarke and Karani, 1992;Zhuang et al., 1999;Alastuey et al., 2004): H 2 SO 4(aq, g) + CaCO 3(s) → CaSO 4(s) + H 2 O + CO 2(g) , (R1) A geogenic S-rich component, represented by residual gypsum, is also present in minor amounts. This phase is a constituent of desert soil (Schwikowski et al., 1995;Claquin et al., 1999) and it is supposedly linked to a Saharan dust episode rather than an autochthonous contribution, hardly justifiable in the light of the lithologies outcropping in the Figure 8. Surface concentration trends of total particles (a), geogenic particles (b) and soot (c) in the ≥ 0.7 µm fraction. The flare event date is highlighted in red. The largest amounts of particles were observed on 28 September 2012, due to increased geogenic particle quantities. The highest soot surface densities on the filters were identified on the day after the flare at C.O.V.A., but soot amounts had been significantly increasing since 25 September, long before the event.
area. Conversely, the OSP consist of secondary sulphur compounds of anthropogenic origin.
Total particle surface concentrations are reported in Fig. 8a. The largest amounts of particles in the coarser fraction were observed on 28 September 2012, due to increased geogenic particle quantities (Fig. 8b). Indeed, the dry depo- sition of crustal particles seems to make a significant contribution to the PM 1 accumulation on the filters, limited to the ≥ 0.7 µm fraction.
As shown in Fig. 7, highest values of S-rich particle surface concentration were detected on the day of the flare, due to a significant increase in mixed particles. This increase is probably favoured by larger amounts of crustal particles that serve as substrate to nucleation or deposition of secondary sulphates.
Highest values of gypsum particles were detected on 24 and 28 September, allowing the hypothesis that the Saharan dust episode influenced the PM 1 sampled on these dates, which also show a relative and absolute peak of geogenic particles respectively (Fig. 8b). Furthermore, these particles often have rounded morphologies or chamfered edges, and consist mainly of clay minerals (smectite, kaolinite and illite/mica in variable proportions) and quartz, which are the main components of atmospheric Saharan dust (e.g. Krueger et al., 2004;Brooks et al., 2005;Coz et al., 2009;. Particularly, on these dates the highest amounts of quartz were detected (Figs. 5 and 6), in accordance with the high mechanical stability of this mineral phase.
Average Al / Si ratios for each sample are consistently close to the value of 0.3, with lowest values on 24 and 28 September (respectively 0.23 and 0.24, Fig. 9).
Al / Si ratios > 0.3 are generally considered indicators of desert dust (Molinaroli, 1996;Guerzoni et al., 1997;Blanco et al., 2003;Kandler et al., 2007;Matassoni et al., 2011), however Coude-Gaussen et al. (1987 have already found that the Al / Si ratio can vary with particle size. They noted that finest fractions have an Al / Si ratio significantly lower than that of mean aerosols, with values below 0.3, due to larger amounts of quartz. These fine quartz particles seem to be good indicators of aeolian and desert dust, because their formation can be attributed to impacts between sand grains (Krinsley and MacCoy, 1978), crystalline rock wind corrosion (Wilding et al., 1977) and amorphization of grains (Le Ribault, 1971Ribault, , 1977, related to saltation and rolling aeolian mobilization. These processes are very efficient in desert areas due to the absence of vegetation. In the light of these considerations, the Al / Si ratios detected are in accordance with the assumption that the Saharan episode, particularly on 24 and 28 September, had a significant influence on the PM 1 . A wind mark on the quartz particles was made on their morphological characters: they are often very worked, with rounded corners and microfeatures probably originating from impacts during long periods of transport. Furthermore, particles potentially associated with local mechanical abrasion processes, such as road dust resuspension and fugitive dust due to agricultural operations in rural environments are generally characterized by coarse fraction mode (i.e. particles with an aerodynamic diameter between 2.5-10 µm), as indicated by Watson et al. (2000), Thorpe and Harrison (2008), Colbeck et al. (2011). Thus, the contribution of these emission sources to PM 1 can be considered limited.
As shown in Fig. 8c, the highest soot surface densities on the filters was identified on the day after the flare at C.O.V.A. However, soot amounts had been significantly increasing since 25 September, long before the event, suggesting that this increase is related to different causes, such as long range transport, according to previous studies (Alastuey et al., 2005;Kallos et al., 2007;Formenti et al., 2011;Rodríguez et al., 2011). However, we cannot exclude a soot contribution from flaring.
Greatest amounts of OSP were detected on the filter the day following the flare event. These particles may be indicators of larger amounts of sulphur in the atmosphere.
The finer particles (D Fe < 0.7 µm) consist mainly of anthropogenic or composite origin aerosols (soot and secondary particles such as ammonium sulfates, sodium nitrates, sodium sulfates or their mixtures), whereas natural component is much less significant than that in the coarser fraction.
Origin of secondary compounds can be attributed to different formation mechanisms.
HNO 3(g) + NaCl (s,aq) → NaNO 3(s,aq) + HCl (g) . (R2) The supply of sea salt could be related to the pathways of air mass coming from North Africa across the Mediterranean Sea. Only a few unprocessed sea salt particles were detected, probably due to nitrates quickly replacing chloride (e.g. Laskin et al., 2002). Furthermore, the behavior of the hygroscopic NaCl particles can encourage aggregation with silicate particles, abundant in Saharan dust outbreaks (Levin et al., 2005;Matassoni et al., 2011).

Conclusions
The pilot site's PM 1 contains a significant allochthonous component. Fine quartz particles and lower Al / Si ratios represent good markers for a desert dust origin, proving that Saharan dust episodes can play a significant role in supplying geogenic aerosol components to the PM 1 . Therefore, studies of mineral dust and its influence on atmospheric processes, terrestrial environment and human health should not be neglected in this fraction.
Soot is the main anthropogenic component in the ≥ 0.7 µm fraction. However, it's not possible to identify a certain cause-effect relationship between soot increment and flaring, because this increase started long before the event and could also be due to the Saharan dust episode which supplies allochthonous soot particles.
Deposition and heterogeneous nucleation of secondary sulfates on mineral dust and formation of composed origin (natural and anthropogenic) particles are very important mechanisms during the period studied, due to the presence in the atmosphere both of sulphur compounds and of geogenic substrates. Furthermore, the reactions of sulphur and nitrogen compounds with dust particles can be considered an important removal mechanism of SO 2 gaseous pollutants and their reaction products present in the atmosphere. This removal mechanism due to dust particles was already indicated by Kerminen et al. (1997), Zhuang et al. (1999), .
In the light of these considerations, it can be concluded that Saharan episodes mark the PM 1 composition with regard to both geogenic component and to soot amounts and secondary aerosols, also supplying allochthonous pollutants as already indicated by Formenti et al. (2011) and Rodríguez et al. (2011). In the pilot site, the Saharan episode seems to have, concurrently, a negative influence on air quality, in relation to the contribution of allochthonous contaminants (primarily soot), and a positive scavenging effect in removing gaseous contaminants (primarily sulphur), as a result of its deposition or nucleation on the geogenic particles that serve as substrate.
The Saharan episode hides the identification of soot contribution to a flaring event, suggesting that soot autochthonous contribution in the pilot site, and in a flaring condition, can be easily superceded by allochthonous soot carried by a Saharan episode, which frequently occurs in the Mediterranean Basin. Conversely, the presence of OSP in the ≥ 0.7 fraction on the day after the flaring suggest these particles are a possible indicator of increased sulphur in the atmosphere.
Therefore, monitoring activities providing geochemical and mineralogical characterization of the atmospheric particulate through continuous and systematic observations using Scanning Electron Microscopy and Energy Dispersive Xray Spectrometer measurements could be useful to identify chemical, geochemical and mineralogical anomalies of the PM composition potentially linked to flaring events and oil pre-treatment processes ongoing in the area studied.