PM Dimensional Characterization in an Urban Mediterranean Area: Case Studies on the Separation between Fine and Coarse Atmospheric Aerosol

: Fine particulate matter (PM) is object of particular attention due to its health effects. It is currently regulated by adopting PM 2.5 as an indicator to control anthropogenic combustion emissions. Therefore, it is crucial to collect aerosol samples representative of such sources, without including PM from natural sources. Thus, a clean separation between coarse and ﬁne mode aerosol should be set. With this purpose, aerosol size mass distribution was taken in the aerodynamic diameter range from 0.5 to 10 µ m. In comparison with a base scenario, characterized by local pollution sources, three case studies were considered, involving desert dust advection, sea salt advection and forest ﬁre aerosol from a remote area. In the base scenario, PM 2.5 represented a suitable ﬁne-mode indicator, whereas it was considerably affected by coarse PM in case of desert dust and sea salt aerosol advection. Such interference was considerably reduced by setting the ﬁne/coarse separation at 1.0 µ m. Such separation underrepresented ﬁne PM from forest ﬁre long-range transport, nonetheless in the case studies considered, PM 1 represented the best indicator of ﬁne aerosol since less affected by coarse natural sources. The data presented clearly support the results from other studies associating the health effects of PM 2.5 to PM 1 , rather than to PM 1–2.5 . Overall, there is a need to reconsider PM 2.5 as an indicator of ﬁne atmospheric aerosol.


Introduction
Particulate matter (PM) atmospheric pollution started becoming a serious health problem at the beginning of the 19th century, when steadily increasing industrialization caused the release into the atmosphere of growing amounts of pollutants. Nonetheless, it was in 1987 that PM 10 limits were set for the first time by the United States Environmental Protection Agency (USEPA) (52 FR 24634) and in 1999 by the European Union (EU) [1]. Therefore, epidemiological associations with the health effects observed have been established with the available PM 10 data. Human exposure to high PM 10 levels has been associated to different adverse health effects involving the cardiovascular and respiratory systems, especially in subjects with pre-existing diseases; besides, PM 10 exposure can be determine lung cancer, and various kind of allergies [2][3][4][5]. Finally, in 2013, more than 1000 studies and pieces of scientific evidence led the International Agency for Carcinogenic Research (IARC) to define outdoor air pollution, with particular attention to particulate matter, as carcinogenic to humans [6].

Materials and Methods
Aerosol measures carried out at the INAIL's building in downtown Rome (41 • 53 46 N, 12 • 29 46 E) were used for the analysis of the basic, desert dust and sea salt aerosol advection scenarios, whereas those acquired at the Italian National Institute of Health (ISS)'s building (41 • 54 16 N, 12 • 31 02 E) were used for the forest fire burning scenario. Aerosol size number distributions were measured in the aerodynamic diameter (d a ) range of 0.5 µm-10 µm, by means of an Aerodynamic Particle Sizer (mod. 3321 APS, TSI, Shoreview, MN, USA). The instrument relies on the correlation of the lag of a particle behind a carrier gas in an accelerating flow field with its aerodynamic diameter [25,26]. It measures the time of flight of each sampled particle, once it is accelerated through a nozzle, by means of two overlapping laser beams. Particles are counted and sized in fifty size channels, in the range 0.5-20 µm. Measurements were taken with 5 min and 1 h time resolution at the INAIL and ISS buildings, respectively. Aerosol size mass distribution have been calculated supposing spherical particles with 1.5 g cm −3 density [27][28][29]. The dynamic properties of the Planetary Boundary Layer (PBL) were estimated by means of a PBL Mixing Monitor (FAI Instruments, Fonte Nuova, Italy) at 1 h time resolution [30]. The instrument samples atmospheric PM on 47-mm membrane filters, where a Geiger detector measures the β-radioactivity of short-lived decay products of Radon ( 222 Rn). For the sea salt advection scenario, the ionic fraction of PM 2.5 has been measured by means of an URG-9000 ambient ion monitor (Chapel Hill, NC, USA) equipped with two ion chromatograph, which allows direct measurements with hourly time resolutions of cations and anions [31]. For the desert dust advection scenario, the PM 10 chemical analysis were obtained from the data of Institute of Atmospheric Pollution Research (IIA), National Research Council (CNR) in Montelibretti (about 30 km north-east from Rome), made available in the framework of the European Monitoring and Evaluation Programme (EMEP) (data available at http://ebas.nilu.no/ (last access on 18 December 2021).

Basic Scenario-Local Pollution Sources
Atmospheric aerosols, based on their origin and on their particle size, are currently described by two main fractions: a coarse one, identified as PM 10 -PM 2.5 and a fine one, identified as PM 2.5 , which includes ultrafine (da < 0.1 µm) and accumulation mode particles (0.1-1 µm). Fine mode aerosol derives mainly from primary anthropogenic combustion sources and from secondary processes, whereas coarse mode PM derives from soil and road dust resuspension [32][33][34], and from natural sources, such as desert dust [35,36] and sea salt aerosol [37,38]. The two modes, given that their emission sources are completely different, are independent one from the other. Due to the well-known effects of exposure to air pollution and PM on human health, it is crucial to work on both reducing emissions and regulating and controlling ambient concentrations of PM 2.5 as the indicator for fine.
In downtown Rome, in winter period, on average, PM 10 is made of primary anthropogenic aerosol (31%), secondary aerosol (36%), crustal material (25%) and sea salt aerosol (8%) [39]. Figure 1 describes the aerosol mass size distributions peculiarly measured in downtown Rome, during the rush hours in wintertime, in pollution conditions mainly determined by local sources. In such conditions, the cut-off point between fine and coarse mode can be set in between 1.0 µm and 1.5 µm. Nonetheless, a 2.5 µm cut-off point represents a suitable separation between the two fractions, since it includes only a small portion of the lower tail of the coarse size mass distribution.

Desert Dust Advection-Transition from a Mainly Fine to a Mainly Coarse Mod
Dust from desert areas can be uplifted at high altitude into the atmosph transported over long distances on a synoptic scale [40,41]. Saharan dust advec sents a source of coarse aerosol in the Mediterranean basin [42,43]. Saharan transported over Rome on 30% of the days of 2001, causing an average cont PM10 concentrations of about 15.6 µ g m −3 [36]. Worldwide, Kok et al. [44] estim 22-29 Tg global loading of dust derives 50% from North African regions, 40% f regions and 10% from North American and Southern Hemisphere regions.
Surface aerosol concentrations at the receptor sites become affected whe mixing height increases to the extent that it intercepts the dust laden air mass. due to the increased atmospheric dilution, this occurrence causes the reduction centrations of ground emitted pollutants, whereas the surface concentration of transported at high altitude increases [41,45]. In Figure 2a the temporal trends mass concentrations measured by APS are reported for some size channels in 0.5-3.5 µ m. Figure 2b,c show for the same time interval the temporal trends of radioactivity associated to short lived decay products of radon and of CO an measurement of natural radioactivity represents a useful tool to infer informa dynamic evolution of the PBL mixing height. Furthermore, 222 Rn is a natura radioactive gas from soil ground emitted gas, and for a given geographical regio sion rate is constant in the regional scale of some kilometers [46,47]; it does n

Desert Dust Advection-Transition from a Mainly Fine to a Mainly Coarse Mode Aerosol
Dust from desert areas can be uplifted at high altitude into the atmosphere and be transported over long distances on a synoptic scale [40,41]. Saharan dust advection represents a source of coarse aerosol in the Mediterranean basin [42,43]. Saharan dust was transported over Rome on 30% of the days of 2001, causing an average contribution to PM 10 concentrations of about 15.6 µg m −3 [36]. Worldwide, Kok et al. [44] estimated that a 22-29 Tg global loading of dust derives 50% from North African regions, 40% from Asian regions and 10% from North American and Southern Hemisphere regions.
Surface aerosol concentrations at the receptor sites become affected when the PBL mixing height increases to the extent that it intercepts the dust laden air mass. Peculiarly, due to the increased atmospheric dilution, this occurrence causes the reduction of the concentrations of ground emitted pollutants, whereas the surface concentration of pollutants transported at high altitude increases [41,45]. In Figure 2a the temporal trends of aerosol mass concentrations measured by APS are reported for some size channels in the range 0.5-3.5 µm. Figure 2b,c show for the same time interval the temporal trends of the natural radioactivity associated to short lived decay products of radon and of CO and NO. The measurement of natural radioactivity represents a useful tool to infer information on the dynamic evolution of the PBL mixing height. Furthermore, 222 Rn is a natural occurring radioactive gas from soil ground emitted gas, and for a given geographical region its emission rate is constant in the regional scale of some kilometers [46,47]; it does not undergo any chemical transformation. Therefore, 222 Rn and its short-lived decay products concentrations merely depend on the degree of atmospheric dilution, namely on PBL mixing height and on the presence of advective conditions [48,49]. On 20 January 2013 natural radioactivity sharply decreased ( Figure 2b); concurrently, the concentrations of CO and NO, primary pollutants associated to local combustion sources, also decreased ( Figure 2c). On 20 January 2013 the daily concentration trends of mineral dust markers such as aluminum, silicon and titanium [53,54]  As a result of the increased dilution, the 0.5 µm particle concentration also abruptly decreased (Figure 2a). Conversely, the concentration of particles with sizes above 1 µm steeply increased. Such behavior can be explained by the entrainment into the PBL mixing height of an aerosol source present in a higher atmospheric layer. Indeed, the dust forecasts carried out by the Dust REgional Atmospheric Model (DREAM) developed by the Earth Sciences Division of the Barcelona Supercomputing Center (BSC) ( Figure S1) [49,50], and the NOAA HYSPLIT backward trajectories ( Figure S2) [51,52], show the advection over Rome on 20 January 2013 of dust from north-western Saharan desert reaching the ground. PM 10 chemical analysis data carried out by IIA-C.N.R. at the EMEP monitoring site (data available at http://ebas.nilu.no/, last access on 18 December 2021) conclusively confirms such an interpretation (Figure 3).
Atmosphere 2022, 13, x FOR PEER REVIEW 6 of 14 was 2.17, concurrent with the range between 1 and 7 reported for northern African and eastern Asian dust samples [55,56]. Moreover, the elemental (Ca + Mg)/Fe ratio of 2.89 was also in agreement with the dust provenience from northern Africa. Such a ratio has been reported to be greater than 1 for dust coming from Atlas region, central Algeria, Libya, and Egypt and lower than 1 for the sub-Saharan region [54,56].  Figure 4 shows the temporal evolution of the aerosol size mass distributions for the 1h-time intervals highlighted with arrows in Figure 2a. Before the marked increase of the PBL mixing height (Figure 2b), aerosol pollution was dominated by the fine fraction, with a mode below 1 µ m (Figure 4a). When the PBL mixing height steeply increased, the fine aerosol concentration markedly dropped, because it was associated to local sources of pollution and due to the increased atmospheric dilution (Figure 4b). Few hours later, because of the increased PBL mixing height, the Saharan dust transported at high quote reached the ground and the aerosol size mass distribution exhibited an intense broad coarse mode centered at about 2.5 µ m (Figure 4c). In such conditions, the mass size distributions changed from having a maximum at below 1 µ m (Figure 4a) to maximum values at about 2.5 µ m (Figure 4c). Before this transition occurred the aerosol levels dramatically dropped (Figure 4b) due to strong dilution of local sources. When desert dust reached the ground the local source contribution was very low and PM2.5 was almost completely made of desert dust. It is worth observing that, in agreement with the findings of Gobbi et al. [40] and of Manigrasso et al. [41,45], the tail of the coarse fraction extended below 1 µ m, as also shown by the 0.77 µ m and 0.89 µ m size fractions that, even if at markedly lower On 20 January 2013 the daily concentration trends of mineral dust markers such as aluminum, silicon and titanium [53,54] exhibited peak values (Figure 3a). Specifically, in agreement with HYSPLIT back-trajectories ( Figure S2), on 20 January 2013 the Al/Si ratio was 2.17, concurrent with the range between 1 and 7 reported for northern African and eastern Asian dust samples [55,56]. Moreover, the elemental (Ca + Mg)/Fe ratio of 2.89 was also in agreement with the dust provenience from northern Africa. Such a ratio has been reported to be greater than 1 for dust coming from Atlas region, central Algeria, Libya, and Egypt and lower than 1 for the sub-Saharan region [54,56]. Figure 4 shows the temporal evolution of the aerosol size mass distributions for the 1h-time intervals highlighted with arrows in Figure 2a. Before the marked increase of the PBL mixing height (Figure 2b), aerosol pollution was dominated by the fine fraction, with a mode below 1 µm (Figure 4a). When the PBL mixing height steeply increased, the fine aerosol concentration markedly dropped, because it was associated to local sources of pollution and due to the increased atmospheric dilution (Figure 4b). Few hours later, because of the increased PBL mixing height, the Saharan dust transported at high quote reached the ground and the aerosol size mass distribution exhibited an intense broad coarse mode centered at about 2.5 µm (Figure 4c). In such conditions, the mass size distributions changed from having a maximum at below 1 µm (Figure 4a) to maximum values at about 2.5 µm (Figure 4c). Before this transition occurred the aerosol levels dramatically dropped ( Figure 4b) due to strong dilution of local sources. When desert dust reached the ground the local source contribution was very low and PM 2.5 was almost completely made of desert dust. It is worth observing that, in agreement with the findings of Gobbi et al. [40] and of Manigrasso et al. [41,45], the tail of the coarse fraction extended below 1 µm, as also shown by the 0.77 µm and 0.89 µm size fractions that, even if at markedly lower concentrations, shared the same temporal trend of the size fractions above >1 µm (Figure 2a).

A Coarse Aerosol Period-Desert Dust Followed by Sea Salt Aerosol Advection
Desert dust advection occurred over Rome on 27 April 2009, as shown by the DREAM dust loading ( Figure S3) and by the NOAA Hysplit backward trajectories (Figure

A Coarse Aerosol Period-Desert Dust Followed by Sea Salt Aerosol Advection
Desert dust advection occurred over Rome on 27 April 2009, as shown by the DREAM dust loading ( Figure S3) and by the NOAA Hysplit backward trajectories ( Figure S4). As previously discussed, the size fractions from 0.78 µm to 3.5 µm (Figure 5a) reached peak concentrations when the level of natural radioactivity dropped, due to the increased PBL mixing height (Figure 5b). Coherently, in the same period, such size fractions followed the same temporal trend of Ca 2+ ions measured on PM 2.5 fraction (Figure 5c) because of the strong desert dust contribution. At the end of 27, on 28 and on 29-30 April 2009 the size fractions from about 0.8 µm to 3.5 µm ceased to follow the Ca 2+ temporal trend and followed the trends of Na + and Cl − ions, suggesting the advection of sea salt aerosol. Coherently, the wind direction in the same periods changed from an offshore to an onshore direction ( Figure S5).
Atmosphere 2022, 13, x FOR PEER REVIEW 8 of 14 S4). As previously discussed, the size fractions from 0.78 µ m to 3.5 µ m ( Figure 5a) reached peak concentrations when the level of natural radioactivity dropped, due to the increased PBL mixing height ( Figure 5b). Coherently, in the same period, such size fractions followed the same temporal trend of Ca 2+ ions measured on PM2.5 fraction (Figure 5c) because of the strong desert dust contribution. At the end of 27, on 28 and on 29-30 April 2009 the size fractions from about 0.8 µ m to 3.5 µ m ceased to follow the Ca 2+ temporal trend and followed the trends of Na + and Cl − ions, suggesting the advection of sea salt aerosol. Coherently, the wind direction in the same periods changed from an offshore to an onshore direction ( Figure S5).  The size mass distribution characterizing the two periods of desert dust (a) and of sea salt aerosol advection (b) are reported in Figure 6. In both cases, a broad coarse mode was observed, centered at about 4 µm, with a strong contribution at 2.5 µm and a lower tail extending below 1.0 µm.
The size mass distribution characterizing the two periods of desert dust (a) and of sea salt aerosol advection (b) are reported in Figure 6. In both cases, a broad coarse mode was observed, centered at about 4 µ m, with a strong contribution at 2.5 µ m and a lower tail extending below 1.0 µ m.

Forest Fires
Drought and high temperature weather conditions, more and more frequent due to the global climate change, have been causing the worldwide alarming increase of the frequency and of the extension of forest fires [57,58]. As a result of that, huge amounts of aerosol are released into the atmosphere and are transported over long distances.
In 2019 Ager et al. [59] expressed a growing concern over the fires ignited in the Chernobyl contaminated areas, with the risk of significant 137 Cs resuspension. In particular, they estimated that the risk of large wildfire was highest in the Ukrainian Chernobyl Exclusion Zone. At the beginning of April 2020 fires broke out in the Chernobyl Exclusion Zone [60-62] and the smoke plume completely invested the Italian peninsula (https://www.youtube.com/watch?v=drBEy4V0j3I, last access on 18 January 2022; IRSN, 2020) [63], as also shown by the NAAPS smoke surface concentration map ( Figure S6). Figure 7 shows the daily trends of the 1h-average mass concentrations of the APS (ISS Building) aerosol fractions with da in the size range from 0.5 µ m to 3.5 µ m. These data show an intense peak of the fractions in between 0.7 µ m and 1 µ m on 9 April 2020 and a less intense one on 13 April 2020.
The NOAA HYSPLIT back trajectories at 500 m height over Rome on 9 April ( Figure  S7) are coherent with the transport over Rome of an air mass coming from Ukraine.
The hourly average aerosol mass size distribution before (8 April) and after (10 April), the first intense fine episode (Figure 8a,c), show that the separation between the coarse and the fine fraction can be set in between 1.0 µ m and 1.5 µ m. However, the mass size distributions measured on the peak days (Figure 8b,d) are somewhat different, the fine mode is broader and its upper tail spans slightly over 1.5 µ m, overlapping to some extent with the lower tail of the coarse mode. Indeed, particles of 1-2 μm in diameter were detected in forest fire plumes at high altitude (above 10 km) by Dahlkötter et al. [64]. Moreover, the mode of the mass size distributions observed at 0.8-0.9 µ m agrees with the findings of Sapkota et al. [65] who assessed the impact of Quebec Forest fires on the air quality in Baltimore and reported particle mass size distributions below 2.5 µ m and peak PM concentration in the 0.8-0.9 µ m size range. Therefore, on 9 ( Figure 8b) and 13 April ( Figure 8d) the coarse-fine separation would be properly set at an aerodynamic diameter of about 1.5 µ m. In any case, a 2.5 µ m cut-off point would include a portion of the coarse mode aerosol.

Forest Fires
Drought and high temperature weather conditions, more and more frequent due to the global climate change, have been causing the worldwide alarming increase of the frequency and of the extension of forest fires [57,58]. As a result of that, huge amounts of aerosol are released into the atmosphere and are transported over long distances.
In 2019 Ager et al. [59] expressed a growing concern over the fires ignited in the Chernobyl contaminated areas, with the risk of significant 137 Cs resuspension. In particular, they estimated that the risk of large wildfire was highest in the Ukrainian Chernobyl Exclusion Zone. At the beginning of April 2020 fires broke out in the Chernobyl Exclusion Zone [60-62] and the smoke plume completely invested the Italian peninsula (https:// www.youtube.com/watch?v=drBEy4V0j3I, last access on 18 January 2022; IRSN, 2020) [63], as also shown by the NAAPS smoke surface concentration map ( Figure S6). Figure 7 shows the daily trends of the 1h-average mass concentrations of the APS (ISS Building) aerosol fractions with da in the size range from 0.5 µm to 3.5 µm. These data show an intense peak of the fractions in between 0.7 µm and 1 µm on 9 April 2020 and a less intense one on 13 April 2020.
The NOAA HYSPLIT back trajectories at 500 m height over Rome on 9 April ( Figure S7) are coherent with the transport over Rome of an air mass coming from Ukraine.
The hourly average aerosol mass size distribution before (8 April) and after (10 April), the first intense fine episode (Figure 8a,c), show that the separation between the coarse and the fine fraction can be set in between 1.0 µm and 1.5 µm. However, the mass size distributions measured on the peak days (Figure 8b,d) are somewhat different, the fine mode is broader and its upper tail spans slightly over 1.5 µm, overlapping to some extent with the lower tail of the coarse mode. Indeed, particles of 1-2 µm in diameter were detected in forest fire plumes at high altitude (above 10 km) by Dahlkötter et al. [64]. Moreover, the mode of the mass size distributions observed at 0.8-0.9 µm agrees with the findings of Sapkota et al. [65] who assessed the impact of Quebec Forest fires on the air quality in Baltimore and reported particle mass size distributions below 2.5 µm and peak PM concentration in the 0.8-0.9 µm size range. Therefore, on 9 ( Figure 8b

Conclusions
A clean separation of fine and coarse atmospheric aerosols is relevant for understanding the anthropogenic sources emissions and for taking important policy for the public health. The case studies reported show that, when local combustion pollution sources along with road dust and crustal material are prevalent, a clear separation between fine and coarse fractions is observed and can be set at 2.5 µm. Under these circumstances, PM 2.5 can be considered a suitable indicator of the fine PM, although slightly affected by the lower tail of the coarse distribution. For such reason, the separation between the two fractions, specifically at 1.0 µm, would be more effective. In fact, under special conditions, when desert dust and sea salt aerosol advection affect the aerosol mass distribution, PM 2.5 includes a strong contribution from these other natural sources. Thus, under these circumstances, PM 2.5 would be an ineffective indicator of anthropogenic fine aerosol. The point is to avoid to erroneously include, as a 2.5 µm separation would, a relevant fraction of desert dust and/or of sea salt aerosol in PM samples that are meant to represent combustion aerosol. It is worth observing that, in these cases, the lower tail of the size mass distributions extends below 1.0 µm, then PM 1 as well is affected, even if to considerably minor extent than PM 2.5 , by the contribution of these natural sources. Forest fires are sources of pollution that are intensifying due to the effects of global climate change. The relevant aerosols share with anthropogenic combustion aerosols components that have a great impact on human health, since they elicit carcinogenic potential and are causing agents of cardiovascular and respiratory pathology exacerbation. Therefore, they should be efficiently sampled upon collecting fine PM samples. The case study presented, concerning long-range transport from a remote area, shows that the upper tail of the fine mode, where this kind of aerosol is distributed, and the lower tail of the coarse mode are to some extent overlapped, so that a proper cut-off point between the two fractions should be placed at 1.5 µm. Setting the separation between fine and coarse PM at 1.5 µm would allow a more efficient collection of aerosols from forest fires but would be subjected to include a still important contribution from desert dust advection.
Overall, the data discussed show that PM 1 clearly represents anthropogenic combustion sources, whereas PM 2.5 may be greatly affected by natural sources. Following the WHO air quality guidelines, addressing the importance of measuring PM below 2.5 µm, in view of the human health protection, PM 1 measurements should be included in air monitoring plans.  Figure S6: NAAPS smoke surface concentration on 8 April 2020 at 18.00 UTC; Figure S7: NOOA HYSPLIT 72 h backward air trajectories passing over Rome, 1000 above ground level, on 9 April 2020.