Chemical characteristics of size-resolved atmospheric aerosols in Iasi , north-eastern Romania : nitrogen-containing inorganic compounds control aerosol chemistry in the area

This study assesses the effects of particle size and season on the content of the major inorganic and organic aerosol ionic components in the Iasi urban area, north-eastern Romania. Continuous measurements were carried out over 2016 using a cascade Dekati low-pressure impactor (DLPI) performing aerosol size classification in 13 specific fractions over the 0.0276–9.94 μm size range. Fine-particulate Cl, NO−3 , NH + 4 , and K + exhibited clear minima during the warm season and clear maxima over the cold season, mainly due to trends in emission sources, changes in the mixing layer depth and specific meteorological conditions. Fine-particulate SO2− 4 did not show much variation with respect to seasons. Particulate NH+4 and NO − 3 ions were identified as critical parameters controlling aerosol chemistry in the area, and their measured concentrations in fine-mode (PM2.5) aerosols were found to be in reasonable good agreement with modelled values for winter but not for summer. The likely reason is that NH4NO3 aerosols are lost due to volatility over the warm season. We found that NH+4 in PM2.5 is primarily associated with SO2− 4 and NO − 3 but not with Cl . Actually, indirect ISORROPIA-II estimations showed that the atmosphere in the Iasi area might be ammonia rich during both the cold and warm seasons, enabling enough NH3 to be present to neutralize H2SO4, HNO3, and HCl acidic components and to generate fine-particulate ammonium salts, in the form of (NH4)2SO4, NH4NO3, and NH4Cl. ISORROPIA-II runs allowed us to estimate that over the warm season∼ 35 % of the total analysed samples had very strongly acidic pH (0– 3), a fraction that rose to∼ 43 % over the cold season. Moreover, while in the cold season the acidity is mainly accounted for by inorganic acids, in the warm ones there is an important contribution by other compounds, possibly organic. Indeed, changes in aerosol acidity would most likely impact the gas– particle partitioning of semi-volatile organic acids. Overall, we estimate that within the aerosol mass concentration the ionic mass brings a contribution as high as 40.6 %, with the rest still being unaccounted for.


Introduction
Despite dramatic progress made to improve air quality, global air pollution continues to harm people's health and the environment.The problem of aerosols or particulate matter (PM) is still a great concern (Olariu et al., 2015) in Europe and many other areas in the world (e.g.China, India, USA).Atmospheric aerosols, described as complex mixtures of liquid and/or solid particles suspended in a gas (Olariu et al., 2015), mainly originate from anthropogenic and natural sources (Querol et al., 2004).Fine PM 2.5 particles (airborne particles with an equivalent aerodynamic diameter < 2.5 µm) are air pollutants with significant effects on human health (Pope et al., 2004;Dominici et al., 2006;WHO, 2006a;Directive 2008/50/EC, 2008;Sicard et al., 2011;Ostro et al., 2014) as well as air quality (Directive 2008/50/EC, 2008;Freney et al., 2014), visibility (Directive 2008/50/EC, 2008;Tsai and Cheng, 1999), ecosystems, weather, and climate (Ramanathan et al., 2001;Directive 2008/50/EC, 2008; Published by Copernicus Publications on behalf of the European Geosciences Union.IPCC, 2013).Aerosols are also known to play a significant role within the chemistry of the atmosphere (Prinn, 2003), acting as surfaces for heterogeneous chemical reactions (Ravishankara, 1997).
Safety threshold values for both PM 2.5 (10 or 17 µg m −3 air, as annual mean) and PM 10 (airborne particles with an equivalent aerodynamic diameter < 10 µm; 20 or 28 µg m −3 air, as annual mean) are addressed by the WHO (2006b) or by the 2008/50/EC Directive (2008) on ambient air quality and cleaner air for Europe.With regard to the PM 2.5 fraction, the EEA Report (2015) indicates that in 2013 the European Union daily limit values for PM 10 and PM 2.5 were exceeded in, respectively, 22 and 7 out of the 28 European Union member states.However, a decreasing trend was observed when compared with the WHO report data (2006b).Moreover, on a global scale the PM 2.5 exposure leads to about 3.3 million premature deaths per year (predominantly in Asia), a figure that could double by 2050 (Lelieveld et al., 2015).Indeed, PM air pollution imparts a tremendous burden to the global public health because it ranks as the 13th leading cause of mortality (Brook, 2008).
Up to now the chemical composition of atmospheric aerosols has been reported for various urban European sites (Bardouki et al., 2003;Hitzenberger et al., 2006;Tursic et al., 2006;Gerasopoulos et al., 2007;Schwarz et al., 2012;Laongsri and Harrison, 2013;Wonaschutz et al., 2015;Sandrini et al., 2016), but such information is very scarce in eastern Europe (Arsene et al., 2011).Sulfate (SO 2− 4 ), nitrate (NO − 3 ), and ammonium (NH + 4 ) ions, which are major inorganic particle constituents (Wang et al., 2005;Bressi et al., 2013;Hasheminassab et al., 2014;Voutsa et al., 2014), are mainly secondary species formed in the atmosphere by chemical reactions of their precursor gases and by physical processes (nucleation, condensation, etc.) (Aksoyoglu et al., 2017).Ammonium aerosols, with an atmospheric lifetime of 1-15 days, have a clear tendency to deposit at large distances from their emission sources (Aneja et al., 2000) and seem to play a very important role in atmospheric chemistry.In urban air, the abundance of NO − 3 on fine particles seems to mainly depend on the reaction between HNO 3 and NH 3 (Stockwell et al., 2000).On a global scale, HNO 3 heterogeneous reactions on mineral dust and sea salt particles might be the predominant source of particulate NO − 3 (Athanasopoulou et al., 2008;Karydis et al., 2011).The human health effects of atmospheric ammonia, primarily exerted through particulate NH 4 NO 3 , are gradually acquiring importance compared to NO x emissions (Sutton et al., 2011).
Unfortunately, the aerosols' role in the global atmospheric system is not yet sufficiently understood.The main related challenges are the occurrence of multiple sources (e.g.soil erosion, sea spray, biogenic emissions, volcano eruptions, soot from combustion, condensation of precursor gases) and the complexity of interactions with other atmospheric constituents (Zhang et al., 2015).Today, sources, distribution, and behaviour of natural and anthropogenic aerosols are still a matter of debate, exacerbated by the scarcity of aerosolrelated work for eastern European Union countries (EEA Report, 2015) but also by the existing discrepancies between models and field measurements.The main uncertainties are related to secondary inorganic aerosols that control the availability of atmospheric sulfuric acid (H 2 SO 4 ), nitric acid (HNO 3 ), and ammonia (NH 3 ) (Ianniello et al., 2011).The existing knowledge gaps bring high uncertainty in the estimated radiative forcing, although they do not impair the conclusion that warming of the climate system is unequivocal (IPCC, 2013).
Despite a growing international recognition of the importance of air pollution and air quality problems, there is a definite need to assess air pollution patterns in Romania.The existing data for north-eastern Romania concern the chemical characteristics of ambient air pollutants such as the watersoluble ionic constituents of aerosols (Arsene et al., 2011) and rainwater (Arsene et al., 2007).Recent work performed by Arsene's group in the field of atmospheric chemistry has clearly shown that, in the Iasi urban environment (northeastern Romania), ∼ 59 % of the total aerosol mass concentration is still unaccounted for.The present work reports, for the first time, detailed information on the chemical composition and seasonal variation in size-segregated, water-soluble ions in aerosol samples collected throughout 2016 in the Iasi urban area, also taking into account the potential ongoing chemistry and the contributions of critical driving forces such as meteorological factors (RH, temperature), mixing layer depth, and emission source intensity.As a first attempt to assess particle acidity in the area, the present work highlights the existence of significant aerosol fractions characterized by pH values in the very strong acidity range (0-3 pH units), with potentially important implications for acid rain.Moreover, the potential importance of gaseous precursors (e.g.NH 3 , HNO 3 , HCl) in secondary inorganic PM is also discussed.

Measurement site
Measurements were performed in Iasi, north-eastern Romania, at the Air Quality Monitoring Station (AMOS, 47 • 9 N latitude and 27 • 35 E longitude) of the Integrated Centre of Environmental Science Studies in the North-Eastern Development Region, "Alexandru Ioan Cuza" University of Iasi, CERNESIM-UAIC, Romania.AMOS is located north-east from the city centre, on the rooftop of the highest university building (∼ 35 m above the ground level), in a totally open area that characterizes the site as an urban receptor point most probably influenced by well-mixed air masses.A comprehensive demo-geographical characterization of Iasi is described in detail by Arsene et al. (2007Arsene et al. ( , 2011)).However, according to a more recent estimate of the Romanian National Institute of Statistics, in 2016 the population in Iasi reached about 362 142 inhabitants (Ichim et al., 2016).

Field measurements
Size-resolved atmospheric aerosols were collected on ungreased aluminium filters (25 mm diameter) using a cascade Dekati low-pressure impactor (DLPI) operating at a flow rate of 29.85 L min −1 .Similar devices have been successfully used in other studies (Kocak et al., 2007;Wonaschutz et al., 2015).The DLPI unit performs aerosol size classification in 13 specific fractions (with size cuts at 0.0276, 0.0556, 0. 0945, 0.155, 0.260, 0.381, 0.612, 0.946, 1.60, 2.39, 3.99, 6.58, and 9.94 µm at 50 % calibrated aerodynamic cut-point diameters and at 21.7 • C, inlet pressure 1013.3 mbar, outlet pressure 100 mbar, and 29.85 L min −1 flow rate).Sampling performance of the DLPI unit was verified through comparison with simultaneous measurements performed with a stacked filter unit (SFU) system previously used by Arsene et al. (2011).Before each reuse the sampler's components were cleaned with ultra-pure water and methanol.Dispensable polyethylene gloves were always used to avoid hand contact with sampler's components, which were assembled and disassembled in a Labgard Class II Biological Safety Cabinet, NuAire.The DLPI sampler was transported to and from the field in tightened polyethylene bags.
Sampling was performed in 2016 twice a week, on the weekend and working days, for a total of 84 sampling events (41 during the cold season from October to March, and 43 during the warm season from April to September), generating 1092 size-resolved aerosol samples.Sampling took place on a 36 h basis, with each sampling event starting at 18:00 local time.An average volume of 64.33 ± 0.85 m 3 per sampling was collected.At least two field blanks (consisting of a loaded sampler taken to and from the field but never removed from its tightened bag) were generated and simultaneously analysed together with the laboratory blanks, in or-der to assess possible contamination during sampler loading, transport, or analysis.
Meteorological parameters including atmospheric temperature, RH, wind speed, wind direction, and global radiation were provided by a Hawk GSM-240 weather station, running at the AMOS site.Information about the mixing layer depth (atmospheric boundary layer) and its ability to dilute atmospheric pollutants at the investigated site was obtained from the NOAA Air Resources Laboratory (ARL) website (Stein et al., 2015;Rolph et al., 2017).

Sample analyses
Aerosol masses for both the PM 2.5 and PM 10 fractions were gravimetrically determined with a Sartorius microbalance (MSU2.7S-000-DF,±0.2 µg sensitivity) by weighing aluminium filters before and after sampling.Before weighing, filters stored in petri dishes were kept for at least 3 days in a conditioned room at 40 ± 2 % RH and at 20 ± 2 • C. In between different procedures, they were appropriately stored in zipped plastic bags.
After sampling and all other required preparative steps, one half of each collected filter was ultrasonically extracted for 45 min in 5 mL deionized water (resistivity of 18.2 M .cm)produced by a Milli-Q Advantage A10 system (Millipore).Filtered extracts (0.2 µm pore size cellulose acetate filters, Advantec) were analysed on an IonPac CS12A (4 × 250 mm) analytical column for cations and an IonPac AS22 (4 × 250 mm) column for anions, running simultaneously on the IC system.The chromatographic instrumental set-up was completed by CSRS 300 × 4 mm and AERS 500 × 4 mm electrochemical suppressors and conductivity detectors.Ion analyses were performed under isocratic elution mode, using CO 2− 3 / HCO − 3 (4.5/1.4 mM, 1.2 mL min −1 ) as the mobile phase for anions and methanesulfonic acid (20 mM, 1.0 mL min −1 ) for cations.
The sum of the detected ions, or of the gravimetrically determined mass concentration, over all DLPI stages is hereafter termed "PM 10 fraction", while the sum over impactor stages from 1 to 10 is termed "PM 2.5 fraction".Modal diameters of the size-segregated aerosol particles or of the analysed ionic components (individual or as a sum) were determined by fitting lognormal distributions.

Estimation of the aerosol acidity
The thermodynamic model proposed by Fountoukis and Nenes (2007), i.e.ISORROPIA-II (http://isorropia.eas.gatech.edu/),was used to get an estimate of the in situ potential acidity of our PM 2.5 fractions.The ISORROPIA-II thermodynamic equilibrium model calculates the gas-liquidsolid equilibrium partitioning of K + , Ca 2+ , Mg 2+ , NH + 4 , Na + , SO 2− 4 , NO − 3 , Cl − , and aerosol water content, and it can predict particle pH.Up to now the model has been used in various field campaigns' data analysis (Nowak et al., 2006;Fountoukis et al., 2009).
To obtain the best predictions of aerosol pH, ISORROPIA-II was run in the "forward mode" for the metastable aerosol state, as preliminary runs of the experimental data in the "reverse mode" did not supply suitable information.In the metastable mode the aerosol is assumed to be present only in the aqueous phase, either supersaturated or not (Fountoukis and Nenes, 2007).As model input data we used just aerosolphase ion concentrations measured with IC, along with RH and temperature data from the Hawk GSM-240 weather station.Actually, in the absence of accompanying gas-phase data required to constrain the thermodynamic models, the accuracy of aerosol pH predictions can be enhanced by using the aerosol concentrations in forward mode calculations (Guo et al., 2015;Hennigan et al., 2015), which seem to be less sensitive to measurement errors than the reverse mode.Where required, NH 3 data predicted by ISORROPIA-II were used for the interpretation of the results.

Air mass back trajectories and air mass origin
Air mass back trajectories were calculated using the HYS-PLIT 4 model of the NOAA Air Resources Laboratory (Stein et al., 2015;Rolph et al., 2017).Arriving at the investigated site at 18:00 local time (15:00 UTC), 48 h back trajectories were computed at 500, 1000, and 2000 m altitude above the ground level.Four major sectors of air mass origin were distinguished, and their contributions are shown in Fig. 1.The most and least frequent sectors were the north-eastern (N-E, 36.6 %) and the south-south-eastern (S-SE, 11.4 %) ones, respectively.The W-SW sector mainly prevails during winter, while the NE sector is most common during summer.The NW sector had a slightly enhanced frequency in winter and summer that, according to James (2007), could reflect a possible European monsoon circulation.Events from the S-SE sector, prevailing mainly in spring, carried out marine chemical features highly influenced by the Black Sea.Air masses undertaking faster vertical transport most probably due to the locally or continentally driven buoyancy (Holton, 1979;Seinfeld and Pandis, 1998), travelling above large (long-range transport) or short (local) continental areas, were also identified.For instance, in April 2016 five sampled events out of a total of eight were highly influenced by fast vertical air mass transport.In these events, air masses from both 500 and 1000 m altitude went down to below 500 m (brushing the ground surface), with a strong impact on the chemical composition of the collected particles (vide infra).

Results and discussion
3.1 Variability in PM 10 and PM 2.5 mass concentrations Table 1 shows summary statistics (median, geometric mean, arithmetic mean, standard deviation, minimum, and maximum) for PM 10 and PM 2.5 mass concentrations at the AMOS site for both working days and weekends.Statistical tests were applied to determine whether there are significant differences among working days and weekends.The Shapiro-Wilk normality test applied to both PM 2.5 and PM 10 indicated that the entire database was normally distributed (detailed statistics are given in Table S1 in the Supplement).Moreover, the difference in the mean values of the two groups was not high enough to exclude random sampling variability, thereby suggesting that the differences are not statistically significant.Therefore, local anthropogenic activities seem to bring similar contribution to the aerosol burden in the area on both working days and weekends.However, the PM 10 annual mean mass concentration (18.95 µg m −3 ) did not exceed the WHO 20 µg m −3 air annual mean value, while the PM 2.5 annual mean mass concentration (16.92 µg m −3 ) exceeded the 10 µg m −3 air annual mean value set by WHO (2006b).
Table 2 presents the annual and/or seasonal arithmetic means of PM 10 and PM 2.5 mass concentrations in Iasi, compared to other European sites (mean ± SD).The annual averages obtained in the present work show differences in comparison with those reported by Arsene et al. (2011) for the same site.Arsene et al. (2011) have used a SFU system consisting of a 8.0 µm pore size, 47 mm diameter Isopore polycarbonate filter mounted in front of a 0.4 µm pore size, 47 mm diameter Isopore filter.However, the values determined in the present work for the fractions PM 0.027-1.6(15.6 ± 8.7 µg m −3 ) and PM 0.381-1.6(9.1 ± 5.6 µg m −3 ) are much closer to those reported by Arsene et al. (2011) for PM 1.5 (10.5 ± 11.2 µg m −3 ).The potential influence on the PM levels of particle size cut-off, differences in sampling site altitude, occurrence of precipitation events, long-range trans-port phenomena, and air mass buoyancy is presented in detail in Sect.S1 in the Supplement.
Figure 2 shows monthly arithmetic mean mass concentrations (±standard deviations) of PM 10 and PM 2.5 in Iasi.The distribution of the PM 2.5 / PM 10 ratio is presented in the same figure.The relative contribution of PM 2.5 toward PM 10  0 ± 7.1 (warm)  16.8 ± 8.3 (warm)  21.3 ± 13.0 (cold)  22.6 ± 13.1 (cold)  15.6 ± 8.7 2  -9.1 ± 5.5  showed low variability amongst the months of the year, with ratios ranging from ∼ 0.75 to ∼ 1.0.For urban background and/or traffic sites, WHO (2006a) suggests PM 2.5 / PM 10 ratios in the 0.42-0.82range.Seasonal patterns, with maxima during the cold season and minima during the warm season, are observed for all of the PM 2.5 , PM 10 , and PM 2.5 / PM 10 profiles and might be the combined result of seasonal emissions variations, local-and long-range air transport and dispersion, chemical processes, and deposition (Wang et al., 2016).The maxima of PM 2.5 / PM 10 during the cold season are most probably caused by combustion processes as the burning of coal or petroleum for heating purposes enhances secondary aerosol generation (Li et al., 2012).The lower PM 2.5 / PM 10 values during the warm season might be due to dust events and to more intense anthropogenic activities near the sampling site (e.g.excavation, construction, and building renewal) that would cause higher loading of coarse particles in the atmosphere.As suggested by Zhang et al. (2001), various land use categories (e.g.grass, crops, mixed farming, shrubs) corroborated with other particle-related characteristics (i.e.particle density, relevant meteorological variables) may enhance the dry deposition of sub-micron particles during the warm season and hence their fine / coarse ratio.Another clear seasonal pattern was observed for the mass concentration size distribution (Fig. 3).Over the cold season, it had a clear monomodal feature with a maximum at 381 nm.In contrast, the warm season was characterized by the same dominant fine mode, but also by the occurrence of a supermicron mode between 1.60 and 2.39 µm.Again, changes in source contributions and meteorological conditions could account for the observed differences (details in Sect.S2).3.2 Ionic balance, seasonality of water-soluble ions, and stoichiometry of (NH 4 ) 2 SO 4 and NH 4 NO 3

Ionic balance and potential aerosol acidity
The completeness of the ionic balance was checked for the identified and quantified species ( 2+ , and Ca 2+ ) in both PM 10 and PM 2.5 .The slopes in the raw IC data related to cations vs. anions were < 1 in both PM 2.5 and PM 10 (detailed statistics in Table S2), pointing to an important cation deficit that was higher in the cold compared to the warm season.However, for each sampling event, either cation or anion deficit was observed in various impactor stages.It should also be noted that, at the investigated site, a RH of ∼ 82 % was even observed during the cold season.
Predicting pH is suggested as the best method to analyse particle acidity (Guo et al., 2015).The ion balance method is usually based upon the principle of electroneutrality, and any deficit in measured cationic compared to anionic charge is assigned to the presence of unmeasured protons (H + ).The reverse occurs for unmeasured hydroxyl (OH − ) (Hennigan et al., 2015) or bicarbonate / carbonate (HCO − 3 / CO 2− 3 ) (Fountoukis and Nenes, 2007).In the present work, HCO − 3 / CO 2− 3 was assigned as the missing anion (details in Sect.S3 for HCO − 3 / CO 2− 3 estimation) while NH + 4 was assigned as the main missing cation (details in Sect.S4 for the missing NH + 4 assumptions).Detailed statistics of the cations vs. anions dependences, with HCO − 3 / CO 2− 3 and missing NH + 4 included in the ionic balance, are presented in Table S3.
In an attempt to investigate whether or not the H + species would bring an important contribution within the ionic balance, the ISORROPIA-II thermodynamic equilibrium model proposed by Fountoukis and Nenes (2007) has been used in the present work (more details about ISORROPIA-II can be found in Sect.S5).We investigated the relationship between ISORROPIA-II-predicted aerosol pH and the ionic balance for the present database, and the results are presented in Fig. 4a.The data remarkably follow a traditional titration curve, and they also clearly show that many of the analysed particles were approximately neutral (dashed lines at 0).A very important fraction of the investigated samples were in the acidic range (pH < 3 if samples were in cation deficit mode), while the remaining fraction were alkaline (pH slightly above 7 with anion deficit mode).However, as suggested by Hennigan et al. (2015), small uncertainties in the ionic balance (mainly due to measurement uncertainties and more likely in approximately neutral conditions) may lead to shifts that span over about 10 pH units.Moreover, the sensitivity to changes in the aerosol NH + 4 concentration has been checked in predicted aerosol pH under forward-mode calculation.As shown in Fig. 4b, it seems that the predicted pH might decrease by 2 % when the ionic balance takes into account NH + 4 (total) concentration (defined as the sum between that derived from raw IC data and the part estimated by using the rationale of Arsene et al., 2011).Details on the pH sensitivity tests for NH + 4 concentrations are given in Sect.S6.In Iasi, north-eastern Romania, an important fraction of the total analysed samples were alkaline while the remaining samples were acidic.A more detailed view of the samples' pH distribution can be obtained from the data presented in Fig. 4c, d.It seems that over the warm season about 55-56 % of the analysed samples were alkaline (pH > 7) and about 44-45 % were acidic (pH < 7), with the last fraction mainly distributed in the very strong acidity fraction (∼ 35 % of the samples with pH in the 0-3 range, and about 2 % with an aerosol pH less than 0).Over the cold season only 47 % of the total analysed samples were alkaline (pH > 7) and 53 % were acidic (pH < 7).The acidity was also mainly distributed in the very strong acidity fraction (∼ 43 % of the acidic samples with pH in the 0-3 range).Sulfuric, nitric, hydrochloric, and formic acids are the most likely contributors to aerosol pH in the 0-3 range.Note that strongly acidic aerosols affect air quality, health of aquatic and terrestrial ecosystems (especially through acid deposition), and atmospheric visibility and climate (Dockery et al., 1996;Gwynn et al., 2000;Hennigan et al., 2015).Possible impacts of strongly acidic aerosols are presented in more detail in Sect.S7.Moreover, aerosol acidity can impact the gas-particle partitioning of semi-volatile organic acids.While under strongly acidic conditions (pH 1-3) the pH contribution of organic acids is expected to be negligible as these conditions prevent their dissociation, the scenario may change completely at pH values of 3-7 (vide infra).Under these circumstances, formic acid with pK a = 3.75 (Bacarella et al., 1955)   while Fig. 5c, d present ISORROPIA-II estimates for pH and H + mass concentration distributions, for both the cold (Fig. 5a, c) and the warm (Fig. 5b, d) seasons.Clear monomodal distribution seems to be specific for the cold season, while for the warm season the second mode (1.60-2.39µm) mass concentration distribution seems to be predominated by NO − 3 .For the 155-612 nm size range, from details presented in Fig. 5c, d, it is quite clear that pH ≤ 2. In the present work, the aerosol H + levels inferred indirectly from the ion balance as proposed by Hennigan et al. (2015) showed statistically significant correlation with the H + loadings predicted by ISORROPIA-II in the forward mode (Pearson coefficient of 0.72, p < 0.001).However, despite the good correlation, there were important discrepancies between the two estimated H + levels (intended as absolute values), with those from ISORROPIA-II being considerably lower than those from the ionic balance.The main issue with the model is that it may account only for partial dissociation, while the ionic balance may be affected by the uncertainty due to the propagation of measurement error.The latter may be particularly important in the presence of a slight anion deficit balance, which is interpreted as an H + -loaded system.Despite the uncertainties in the actual pH values, it is very likely that the 155-612 nm aerosol particles are strongly acidic and that H + is mainly contributed by completely dissociated strong acids such as H 2 SO 4 and HNO 3 .Contributions from free acidity (dissociated H + ) or total acidity (free H + and undissociated H + bound to weak acids) are expected to be more important in all other remaining fractions, and especially in the 27.6-94.5nm particle size range (vide infra).Of course, a higher confidence in the estimate of particle pH would allow better prediction of the chemical behaviour of organics that, if dissociated at relatively low acidities, would significantly contribute to the ion balance.
In the literature it is suggested that the molar ratio approach may be a proxy for aerosol pH estimation (Hennigan et al., 2015).However, such a procedure is highly susceptible to bias in the results, either due to exclusion of minor ionic species or because it does not consider the effects of aerosol water or species activities on particle acidity.In this work, even when the aerosol was inferred to be highly acidic (samples with a molar ratio of NH + 4 / (Cl − + NO − 3 + 2 × SO 2− 4 ) < 0.75), there was no statistically significant correlation between the cation / anion molar ratio and [H + ] from either ion balance or model predictions.Therefore, the molar ratio does not appear to be a suitable tool to infer the acidity of atmospheric particles at the study site, but it could be a good parameter to distinguish between alkaline and acidic particles.
Data on gaseous NH 3 were not available, but the potential relationship was also investigated between NH 3 / NH + 4 phase partitioning (with NH 3 values predicted by ISORROPIA-II) and particle pH.The hypothesis of phase partitioning equilibrium is justified by the fact that the sampling time interval (36 h) was much longer than the equilibration time for submicron particles (seconds to minutes; Meng et al., 1995).As previously mentioned, in the present work the number of samples with a pH < 0 was significantly lower compared to those with a pH > 0.Moreover, ISORROPIA-II predicted that in the 94.5-612 nm size range there would be a significant NH 3 fraction in the gas phase.
The detailed NH 3 / NH + 4 partitioning as a function of RH is presented below, and considerations on the potential effects on the partitioning brought about by changes in the SO 2− 4 and NO − 3 concentrations, and by temperature affecting both SO 2− 4 and NO − 3 production, is given in Sect.S8.The potential role played by temperature on NH 3 / NH + 4 partition seems to be minimal, but one should also consider that highly acidic aerosols will affect a variety of processes and definitely the partitioning of HNO 3 to the gas phase, producing low nitrate aerosol levels.ISORROPIA-II runs predicted gas-phase NH 3 concentrations in Iasi as high as 0.52 ± 0.28 (0.46) (mean ± SD (median)) µg m −3 at RH < 40 %, 0.61 ± 0.26 (0.49) µg m −3 at RH = 40-60 %, and 0.96 ± 0.54 (0.92) µg m −3 at RH > 60 %.These warm season values are smaller than those reported in a modelling study by Backes et al. (2016), who predicted NH 3 abundances as high as 1.6 to 2.4 µg m −3 (data extracted from NH 3 concentration for the reference case, i.e.Fig. 3 in Backes et al., 2016, for north-eastern Romania).In contrast, the 0.96 ± 0.54 (0.92) µg m −3 value at RH > 60 %, which would be mainly found during the cold season, seems to be in reasonable agreement with the ≤ 0.8 µg m −3 value modelled by Backes et al. (2016) over winter.Moreover, from ISORROPIA-II runs performed at RH < 40 %, it was estimated that (77.6 ± 28.4) % or (79.3 ± 26.2) % (mean ± SD) of the NH 3 predicted by ISORROPIA-II could be present in the gaseous phase (with reference to both NH + 4 derived from raw IC data and to the NH + 4 (total) fraction).Similar but slightly decreasing values were predicted for gas-phase NH 3 as RH increased, namely (76.5 ± 30.9) % or (78.6 ± 28.2) % (raw IC and NH + 4 (total), respectively) at RH = 40-60 % and (68.3 ± 36.7) % or (74.5 ± 29.7) % at RH > 60 %.
In other studies, thermodynamic equilibrium calculations predicted that all of the NH 3 was mainly susceptible to partitioning to the particle phase at the equilibrium, and also that > 44 or 51 % of the investigated samples presented an aerosol pH < 0 (Hennigan et al., 2015).However, it seems that at the investigated Romanian site the atmosphere could be rich enough in NH 3 so as to allow its occurrence in the gas phase while also promoting particle-phase partitioning.The seasonal trends in the NH 3 concentrations derived from ISORROPIA runs for Iasi are reported in Fig. S1 (Sect.S9).The same section reports considerations on possible interrelated emission factors governing the distribution in the NH 3 concentration levels in Iasi.

Seasonality of the major water-soluble ions
Table 3 shows monthly statistics for meteorological variables and mass concentrations of PM 10 , PM 2.5 , and major water-soluble ions in PM 2.5 .Compared to the PM 2.5 fraction, in the PM 10 fraction we observed increases in the mass concentrations of the following ions (notation for the % increase: min-max (mean)): 13-80 % (35 %) for Cl − ; 1-107 % (32 %) for NO − 3 ; 1-170 % (17 %) for SO 2− 4 ; 38-185 % (63 %) for HCO − 3 ; 14-171 % (41 %) for acetate; 4-136 % (22 %) for formate; 0-294 % (27 %) for oxalate; 16-48 % (32 %) for Na + ; 6-58 % (20 %) for K + ; 1-105 % (12 %) for NH + 4 (total); 28-83 % (46 %) for Mg 2+ , and 33-123 % (61 %) for Ca 2+ .However, the PM 10 and PM 2.5 mass concentration fractions show statistically significant correlation with a ratio of 1.1 (Pearson coefficient of 0.99, p < 0.001).Higher mass concentrations of specific water-soluble ions (Cl − , NO − 3 , K + , NH + 4 , and, to some extent, SO 2− 4 ) were observed during the cold compared to the warm season, probably because of the combination of increased strength of pollution sources and meteorological effects (inducing lower mixing heights or even temperature inversion), or due to different chemical or photochemical processing.Although lowering mixing heights over the cold season might increase pollutant concentration in the atmosphere, for some species additional phenomena should be taken into account in order to explain their distribution.For particulate SO 2− 4 , high concentrations can be observed during winter and autumn but also in summer, and in the latter case they can be due to higher temperature and solar radiation that enhance photochemical reactions and the atmospheric oxidation potential because of the elevated occurrence of oxidant species such as ozone, hydroxyl, and nitrate radicals.These conditions favour the oxidation of SO 2 to particulate SO 2− 4 .Also particulate C 2 O 2− 4 was maximum in summer, possibly due to enhanced photochemical processing.Moreover, the maxima observed for SO 2− 4 during the cold season might be due to the intensifica-tion of coal burning for heating purposes.Higher abundances of particulate NO − 3 , SO 2− 4 , NH + 4 , and K + in winter compared to summer are reported for other European (Schwarz et al., 2012;Voutsa et al., 2014) and non-European sites as well (Sharma et al., 2007).Sharma et al. (2007) also suggest a potential role of CaCO 3 in controlling particulate NO − 3 abundance in Kanpur, India.
Seasonal variations for selected water-soluble ionic components in PM 2.5 are shown in Fig. 6a-h, while Fig. 6i shows the variation in the mixing layer depth at the investigated site.Fine-particulate Cl − , NO − 3 , K + , NH + 4 (total), and to some extent even SO 2− 4 seem to exhibit distinct seasonal variations with maxima during the cold season and minima over the warm season, which might be related to changes in the mixing layer depth.The summer minima observed for both NO − 3 and NH + 4 are not surprising because NH 4 NO 3 is volatile and tends to dissociate to gas-phase NH 3 and HNO 3 at high temperatures.Coarse particulate C 2 O 2− 4 , Ca 2+ , and Na + did not show much variation with respect to seasons.However, SO 2− 4 and C 2 O 2− 4 showed similar patterns (implying most probably common sources), and the Ca 2+ trend suggests prevalent contribution from soil dust.Higher ion concentrations in winter than in summer are reported by Sharma et al. (2007) for Kanpur (India), while Ianniello et al. (2011) report opposite trends for Beijing (China).
Particulate Cl − mass concentrations show a clear seasonal pattern, with higher values during the cold season than during the warm season (Fig. 6a).The chloride mass concentration in both PM 2.5 and PM 10 had a statistically significant correlation with RH, temperature (only for PM 10 fraction), particle loading, and mixing layer depth (detailed statistics in Table S4).The chloride maxima during the cold season might be the result of increased coal burning for heating purposes or of the use of NaCl in winter on icy and snowy roads.These observations are in agreement with other studies at eastern European sites (Arsene et al., 2011;Alastuey et al., 2016).However, the Cl − mass concentration follows a pattern similar to that of K + (tracer of biomass burning), thereby suggesting that over the cold season wood burning might become an important heating source (Christian et al., 2010;Akagi et al., 2011).
Nitrate also shows cold season maxima and warm season minima (see Table 3 and Fig. 6b).The inset distribution presented within NO − 3 seasonal variation (Fig. 6b) suggests that, in summer, the coarse PM fraction can bring significant contributions to the aerosol atmospheric burden of nitrate.In our work, fine-particulate NO − 3 mass concentrations varied from 0.31 to 3.62 µg m −3 (Table 3) and these data are in very good agreement with those predicted for Europe in a modelling study performed by Backes et al. (2016).The data obtained in the present work over the cold season (3.62 ± 1.10 µg m −3 in January, February, and December as the coldest months of the year, and 2.65 ± 0.38 µg m −3 over January, February, March, October, November, and Decem-    ber) seem to be in reasonably good agreement with those predicted for Europe by Backes et al. (2016).In contrast, the 0.59 ± 0.30 µg m −3 NO − 3 concentration in PM 2.5 measured over the warm season (including the April month characterized by predominant air mass buoyancy, and decreasing down to 0.44 ± 0.12 µg m −3 if the April month is excluded) is lower than that predicted by Backes et al. (2016) (abundances as high as 0.8 µg m −3 over summer).Similar to the NH + 4 case, this might reflect the susceptibility of NO − 3 to be transferred to the gas phase over the warm season.
The NO − 3 mass concentrations in both PM 2.5 and PM 10 had a statistically significant correlation with RH, temperature, mixing layer depth, and particle loading (Table 3, detailed statistics in Table S4).However, it has to be observed that highly acidic aerosols expected over all seasons have the potential to affect the partitioning of HNO 3 to the gas phase, producing low nitrate aerosol levels.Guo et al. (2015) also report low nitrate aerosol levels during summer.Moreover, NO − 3 heterogeneous formation (i.e.condensation or absorption of NO 2 in moist aerosols or N 2 O 5 oxidation and HNO 3 condensation) generally relates to RH and the partic-ulate loading (Wang et al., 2006;Ianniello et al., 2011).At the investigated site, this process might be of similar importance as the gas-particle conversion, which implies oxidation of precursor gases, such as NO x , to nitrate via HNO 3 formation and involving photochemical processes.The high concentration of NO − 3 during the cold season might also be caused by higher NH 3 atmospheric levels from yet unaccounted sources, which could neutralize gas-phase H 2 SO 4 and HNO 3 to produce ammonium salts (vide infra).Reactive nitrogen species are emitted to the atmosphere mainly in the forms of NO x (from transport or power generation) and NH 3 (agriculture).In Iasi, the animal husbandry sector (open and closed barns, manure storage or spreading) is most likely an important NH 3 source.Moreover, the high RH observed in the cold season could offer suitable conditions for significant fractions of HNO 3 and NH 3 to be dissolved in humid particles, therefore enhancing particulate-phase NO − 3 and NH + 4 (Pathak et al., 2009(Pathak et al., , 2011;;Ianniello et al., 2010;Sun et al., 2010).From measurements performed in October 2004 in Beijing, China, Kai et al. (2007) also concluded that high RH, stable atmosphere, and high NH 3 levels can enhance transformation of NO x into NO − 3 .Particulate SO 2− 4 maxima are observable during both cold and warm seasons.However, the particulate SO 2− 4 mass concentrations showed statistically significant correlation with the measured meteorological parameters only at a 68 % confidence level (detailed statistics in Table S4).The data presented in Fig. 6c show that the seasonal trend of the monthly SO 2− 4 mean mass concentrations is not as clear as that of NO − 3 and NH + 4 , which might suggest the occurrence of regional SO 2− 4 sources as well (Wang et al., 2016).Moreover, high RH (in Iasi, especially during cold months) may aid the conversion of SO 2 to SO 2− 4 (Kadowaki, 1986), with a significant enhancement of SO 2− 4 production rate in the aqueous phase (Sharma et al., 2007).However, the oxidation of SO 2 to sulfate may be induced not only by H 2 O 2 in the aqueous phase, but also by gas-phase radical hydroxyl (Vione et al., 2003).This issue can explain the PM sulfate maxima during the warm season because of higher sunlight irradiance and temperature (Stelson and Seinfeld, 1982;Stockwell and Calvert, 1983;Kadowaki, 1986;Wang et al., 2005).
The data reported in Table 3 and Fig. 6f show that particulate NH + 4 (total) has a clear seasonal pattern with maxima during the cold and minima over the warm season, in agreement with reports at other European (Schwarz et al., 2012;Bressi et al., 2013;Tositti et al., 2014;Voutsa et al., 2014) or non-European sites (Sharma et al., 2007;Wang et al., 2016).This behaviour is opposite to that reported by Ianniello et al. (2011).The cold season maxima and warm season minima we observed can be due to the effect of variations in the mixing layer depth, combined with gas-phase transfer of NH 4 NO 3 to NH 3 and HNO 3 as temperature increases.In our dataset, NH + 4 (total) varied from 0.8 to 2.09 µg m −3 , a much lower range than that reported by Meng et al. (2011) for a more polluted site (Beijing, China, with concentrations varying between 4.73 and 9.04 µg m −3 among various seasons).However, the data we measured in the cold season (2.03 ± 0.30 µg m −3 over January, February, and December; 1.65 ± 0.23 µg m −3 over January, February, March, October, November, and December) seem to be in reasonable agreement with those predicted for Europe by Backes et al. (2016).In contrast, the 0.90 ± 0.09 µg m −3 NH + 4 (total) measured concentration over the warm season is much lower than that predicted by Backes et al. (2016), and the discrepancy might actually reflect the limitation of the experimental measurement techniques concerning NH 4 NO 3 volatility.Moreover, the NH + 4 (total) mass concentration correlated significantly with RH and particle loading in both the PM 2.5 and PM 10 fractions, and it anticorrelated significantly with temperature and mixing layer depth (detailed statistics in Table S4).The PM 2.5 fraction also showed statistically significant correlation with the mixing depth (Pearson coefficient higher than 0.67, p = 0.016).The seasonal variation in particulate NH + 4 (total) especially follows that of particulate NO − 3 and Cl − , which would indicate that most probably NH + 4 (total) largely originates from neutralization among NH 3 , HNO 3 , and HCl (Wang et al., 2006), or that the cited particulate species derive from similar gas-to-particle processes (Huang et al., 2010).Although in the present work gaseous NH 3 was not measured, ISORROPIA-II runs predicted that the atmosphere was often in a gaseous ammonia-rich state, independent of the RH values ([NH 3 ] / ([HNO 3 ] + [HCl]) 1).
At the investigated site, C 2 O 2− 4 and SO 2− 4 show similar behaviour and the C 2 O 2− 4 maxima during summer may suggest photochemical and/or biogenic contributions to its abundance (Laongsri and Harrison, 2013).The Na + ion, tracer of sea salt or NaCl aerosols, shows higher concentrations during spring when one has a predominant long-range transport of air masses from the S-SE sector, with contributions from natural sources and especially sea spray aerosols from the Black Sea.Particulate K + mass concentrations also show a clear seasonal pattern, with higher values during the cold season compared to the warm season (Fig. 6e).This phenomenon could be due to increased wood burning combined with a limited mixing layer depth.However, particulate K + mass concentrations also had some maxima when one expects intense agricultural biomass burning for field clearing (i.e.April, July, and September).The K + mass concentrations follow a pattern similar to Cl − (Pearson coefficient of 0.79, p = 0.002) but a different one compared to SO 2− 4 .Therefore, we suggest that intense wood burning may be a common source for K + and Cl − species in the study area (Christian et al., 2010;Akagi et al., 2011).
High mass concentrations of Ca 2+ and Mg 2+ (with Mg 2+ shown in Table 3 but not in Fig. 6), as soil or dust tracers, were observed especially during spring and summer.Over these seasons, lower precipitation frequency and high wind speed contribute to the observed behaviour.During the cold season, low wind speeds might prevent mineral dust resuspension and produce low values for these ions.However, Mg 2+ and Ca 2+ as mineral ions did not correlate with either PM 2.5 or PM 10 , which suggests that inorganic particles would be mainly produced by NH 3 -triggered secondary processes.
3.2.3Stoichiometry of (NH 4 ) 2 SO 4 , NH 4 NO 3 , and NH 4 Cl Table 4 presents the correlation matrix (Pearson coefficients) for the major ionic species (Cl − , NO − 3 , SO 2− 4 , CH 3 COO − , HCOO − , C 2 O 2− 4 , HCO − 3 , Na + , NH + 4 (total), K + , Mg 2+ , Ca 2+ ) in PM 2.5 , for both the cold (Table 4a) and the warm seasons (Table 4b).Despite similar correlations in PM 10 as well, PM 2.5 was selected for the correlation matrix analysis because of higher representativeness.For the cold season there are significant correlations (at the 99.9 % confidence level) among many chemical pairs, suggesting an important occurrence of (NH 4 ) 2 SO 4 , NH 4 NO 3 , and (NH 4 ) 2 C 2 O 4 .In 1.00 the warm season, (NH 4 ) 2 SO 4 , Mg(NO 3 ) 2 , and NaCl seem to be the most important.However, Ca(HCO 3 ) 2 might play a role during both cold and warm seasons.Over the cold season, K + shows statistically significant correlation with many inorganic (NO − 3 , Cl − , and SO 2− 4 ) and organic (HCOO − and C 2 O 2− 4 ) anions, thereby suggesting that at least some of these species might have common biomass burning sources (Ianniello et al., 2011, and references therein).However, in the case of SO 2− 4 and NO − 3 the most likely explanation would rather be the use of both coal and wood for burning, as well as the effect of the mixing layer depth.
In the ambient atmosphere, inorganic ammonium salts such as NH 4 HSO 4 , (NH 4 ) 2 SO 4 , NH 4 NO 3 , and NH 4 Cl are known to be produced by gas-to-particle conversion processes.In the present work, from the ionic balance analysis the NH + 4 ion was assigned as the most critical parameter in the chemical composition analysis of aerosol particles (with 274 analysed samples, i.e. about one-quarter of the total, being highly deficient in cations).Figure 7a, b present the relationship between the molar concentrations of fine-particulate NH + 4 (i.e. from raw IC data and also in the total form estimated under the assumptions from Arsene et al., 2011) and that of particulate SO 2− 4 for both the cold (Fig. 7a) and warm (Fig. 7b) seasons.Correlations between particulate NH + 4 and SO 2− 4 are statistically significant in both cases (detailed statistics in Table S5).
During cold and warm seasons, the [NH + 4 ] / (2× [SO 2− 4 ]) molar ratio was either ∼ 1 (raw IC data) or (NH + 4 (total) values) equal to 1.76 (cold season) and 1.02 (warm season).These data suggest the existence of enough NH 3 for the complete neutralization of H 2 SO 4 , and also a predominance of particulate (NH 4 ) 2 SO 4 in agreement with the observations of Ianniello et al. (2011).Moreover, as shown in Table 4, for particulate SO 2− 4 and NH + 4 (total) the correlation was statistically significant (with Pearson coefficients of 0.96, p < 0.001 for the cold season and 0.97, p < 0.001 for the warm season), thereby suggesting that (NH 4 ) 2 SO 4 could be formed from H 2 SO 4 (g) and NH 3 (g) in either case.However, the 1.76 value for the [NH + 4 ](total) / (2 × [SO 2− 4 ]) molar ratio during the cold season will indicate that there should still be Unfortunately, at present, no measured NH 3 values are available for the site of interest but it is still believed that, in the atmosphere of Iasi, sufficient gas-phase NH 3 occurs to promote both the homogeneous and heterogeneous formation of nitrate salts in the collected aerosol particles.The NH 4 NO 3 formation routes might involve either the homogeneous reaction between gaseous HNO 3 and NH 3 (Ianniello et al., 2011) or the heterogeneous reaction between NH 3 and the products formed upon hydrolysis of N 2 O 5 that could be present on the surface of the pre-existing moist aerosols under relatively high humidity (Pathak et al., 2011;Shon et al., 2013).Actually, gaseous NH 3 can influence both the inorganic ions and the aqueous-phase H + distribution in aerosols.The concentration of H + in aqueous aerosols is mainly determined by the balance of the acidic ionic components with the basic ones.During both cold and warm seasons, about half or slightly more than half of the collected samples were found to be alkaline with pH values fluctuating between 7 and 8.The remaining samples were acidic and with pH values ranging mainly from 1 to 3. Zhao et al. (2016) report that, on average, a +25 % perturbation in the NH 3 level could lead to a 0.14 unit pH increase, and a -25 % perturbation could cause a 0.23 unit pH decrease.They concluded that sufficient NH 3 was frequently present in the wintertime atmosphere, and also that the fine collected particulates were almost fully neutralized by NH 3 .
Figure 7c, d, e, f show the relationships between the fineparticulate molar concentrations of (i) NH + 4 and the sum of SO 2− 4 and NO − 3 (Fig. 7c, d), and (ii) NH + 4 and the sum of is formed through Reaction (R3) involving Mg 2+ (but not Ca 2+ ): During the warm season, however, higher concentrations of (NH 4 ) 2 SO 4 compared to NH 4 NO 3 are expected because (NH 4 ) 2 SO 4 is less volatile than NH 4 NO 3 (Utsunomiya and Wakamatsu, 1996).Moreover, NH 4 NO 3 will be formed only when excess NH 3 is available to react with HNO 3 .A modelling study by Backes et al. (2016) suggests that a reduction of NH 3 emissions by 50 % may lead to a 24 % reduction of the total PM 2.5 concentrations in north-western Europe, mainly due to reduced formation of NH 4 NO 3 .However, the NH 3 concentration in the atmosphere over Europe seems to be high enough to saturate the reaction forming (NH 4 ) 2 SO 4 particles, even in a scenario of reduced NH 3 levels, while on the contrary it is not high enough to saturate the reaction with HNO 3 to form NH 4 NO 3 particles.A reduced formation of NH 4 NO 3 particles may lead to an increase in gasphase HNO 3 during winter.In our study, results from the ISORROPIA-II thermodynamic model foresee an increase in gas-phase HNO 3 at higher RH values during the cold season.Higher levels of gas-phase HNO 3 may increase its condensation onto existing particles such as sodium chloride (NaCl), and the replacement of Cl − with NO − 3 may enhance the concentration of HCl in the atmosphere (similar processes are described in Arsene et al., 2011).
In the atmosphere, additional non-volatile species containing nitrate and chloride might also be present; thus we investigated the potential of fine-particulate NO − 3 and Cl − to be chemically bound to Ca 2+ , Mg 2+ , K + , or Na + .The free NO − 3 and Cl − concentrations, defined as the fractions of excess nitrate and chloride that are not bound to alkali or alkaline earth metals, were estimated for both in the 0.946-1.6µm range).Such behaviour most likely reflects contributions from biomass burning over the entire year (Schmidl et al., 2008;Pachon et al., 2013).For both Ca 2+ and Mg 2+ ions, clear monomodal mass distributions with maxima in the 1.6 to 2.39 µm size range were observed over the investigated period.Over the warm season, Ca 2+ accounts for (7.0 ± 2.9) % of the PM 10 fraction and for (5.5 ± 2.9) % of the PM 2.5 fraction, while over the cold season it accounts for only (3.0 ± 0.6) % of the PM 10 and (2.2 ± 0.5) % of the PM 2.5 fraction.These observations indicate that the impact from soil dust resuspension could be more important during the warm (dry) season.Mineral dust may also explain the higher coarse fraction of Mg 2+ (mineral source being MgCO 3 ).
Clear evidence was obtained in this work that air mass origin highly influences the aerosol chemical composition at the investigated site.Annual averaged sector contributions, in terms of PM long-range transport, are shown in Fig. S2 while the seasonal contributions of PM and particulate inorganic and organic ions associated with different air mass origins are reported as radar charts in Fig. S3 (within Sect.S10). Figure S4 (within Sect.S10) highlights the percentage distributions for the identified and quantified ions and the gaseous concentrations of NH 3 , HNO 3 , and HCl for selected investigated events, predicted by ISORROPIA-II.Wonaschutz et al. (2015) suggested that in Vienna, Austria, air mass origin is the most important factor for bulk PM concentrations, chemical composition of the coarse fraction (> 1.5 µm), and mass size distribution, while it is less important for the chemical composition of the fine fraction (< 1.5 µm).Although Iasi is located far from the Mediterranean or Black Sea, over the warm season the sea salt chloride contribution to the aerosol budget in the area is not entirely excluded (Arsene et al., 2011).Moreover, dust particles originating from the Sahara are acknowledged as travelling across the tropical Atlantic Ocean (10-90 µg m −3 ) and across the Mediterranean, affecting air quality in southern Europe (10-60 µg m −3 ) ( Karydis et al., 2016).In the present work, particulate Na + and Cl − ions as tracers of sea salt aerosols (Tositti et al., 2014) were mainly observed in conditions predominated by contributions from air masses arriving in Iasi from the S-SE direction.However, one of the most interesting collected events was that of 9-11 April 2016.For this event, the PM 10 fraction mass concentration was as high as 43.9 µg m −3 , a value which is about 2 times higher than the average of the total events.These conditions were highly influenced by air masses originating from both the Saharan desert and the Mediterranean Sea.As shown in Fig. 10a, the size distributions of particulate Na + , Ca 2+ , Mg 2+ , Cl − ions, and their mass concentrations present a highly dominating mode with maxima at 2.39 µm.For this event, the (Ca 2+ , Mg 2+ ) and (Na + , Cl − ) pairs showed statistically significant correlations (respectively, r = 0.94, p < 0.001 and r = 0.85, p < 0.001), suggesting common contributions from mineral Saharan dust (Ca 2+ , Mg 2+ ) and from sea salt marine aerosols (Na + , Cl − ).Moreover, Fig. 10a clearly shows that Na + , Ca 2+ , Mg 2+ , and Cl − make a very significant contribution to the total aerosol mass in the super-micron mode, with the maxima at 2.39 µm.
In April 2016, an interesting behaviour was also observed for the averaged mass size distributions of particulate NH + 4 , NO − 3 , SO 2− 4 , and Mg 2+ (Fig. 10b).This month is highly affected by the atmospheric air mass buoyancy phenomenon, as shown by trajectory analysis for selected events and, while particulate NH + 4 and SO 2− 4 mainly resided in the fine mode with clear maxima at 381 nm, NO − 3 and Mg 2+ also presented a predominant mode in the 1.6-2.39µm fraction.Such distributions, corroborated with meteorological conditions, would actually suggest a possible heterogeneous formation route for SO 2− 4 (Wang et al., 2012).In contrast, the adsorption of HNO 3 on mineral dust and sea salt particles (Karydis et al., 2016) would become more important for NO − 3 .

Conclusions
The atmospheric concentrations of particulate species including acetate, formate, fluoride, chloride, nitrite, nitrate, phosphate, sulfate, oxalate, sodium, potassium, ammonium, magnesium, and calcium were measured over 2016 at an urban site in Iasi, north-eastern Romania.The measurements were carried out using a cascade Dekati low-pressure impactor (DLPI), performing aerosol size classification in 13 specific fractions evenly distributed over the 0.0276-9.94µm size range.The entire data set was analysed to investigate the seasonal variations in fine-particulate species and the meteorological effects and to examine the contributions of local and regional sources.ISORROPIA-II thermodynamic model runs were used to estimate the pH values of the collected atmospheric particles, as on the present database it was the best method to analyse particle acidity.
Within the aerosol mass concentration, the identified ions' mass brings contributions as high as 40.6 % with the rest being unaccounted for yet.Fine-particulate Cl − , NO − 3 , NH + 4 , and K + exhibited clear seasonal variations with minima during the warm season, mainly due to cold season enhancement in the emission sources, changes in the mixing layer depth, and specific meteorological conditions (e.g. higher RH values prevailing in Iasi during the cold season).Fineparticulate SO 2− 4 did not show much variation with respect to seasons.The measured concentrations of NH + 4 and NO − 3 in fine-mode (PM 2.5 ) aerosols were in reasonably good agreement with modelled values for the cold season but not for the warm season.This observation reflects the susceptibility of NH 4 NO 3 aerosols to be lost due to volatility.
Clear evidences were obtained that NH + 4 in PM 2.5 was primarily associated with SO 2− 4 and NO − 3 .However, indirect ISORROPIA-II estimations showed that the atmosphere of Iasi might be ammonia rich during both cold and warm seasons, so that enough NH 3 would be present to neutralize the H 2 SO 4 , HNO 3 , and HCl acidic components and to generate fine-particulate ammonium salts in the form of (NH 4 ) 2 SO 4 , NH 4 NO 3 , and NH 4 Cl.Significant amounts of fine-particulate NO − 3 were in fact detected during the cold season.The presence of possibly large amounts of NH 3 , the domination of (NH 4 ) 2 SO 4 over NH 4 NO 3 and NH 4 Cl, and the high RH conditions in the cold season (likely leading to dissolution of a significant fraction of atmospheric HNO 3 and NH 3 ) are among the most important driving forces enhancing the fine-particulate NO − 3 and NH + 4 distribution at the investigated site.
Most probably, gaseous NH 3 is not only a precursor of NH + 4 formation, but it also affects the occurrence of NO − 3 and eventually Cl − in PM 2.5 via neutralization processes.The chemical composition database concerning PM 2.5 (and PM 10 ), combined with predictions from ISORROPIA-II run in the forward mode, allow us to suggest that NH 3 was most probably present at a sufficiently high concentration to promote fine-particle acidity neutralization during both the cold and the warm seasons.Although it is known that running ISORROPIA-II in the forward mode with only aerosol concentrations as input may result in a bias in the predicted pH due to the repartitioning of ammonia in the model, this approach was the only one that allowed for a reasonable interpretation of the obtained results.Over the warm season, ∼ 35 % of the total analysed samples presented pH values in the very strong acidity fraction (0-3 pH units range), while over the cold season the fraction of samples in this pH range was ∼ 43 %.However, while during the warm season ∼ 24-25 % of the acidic samples were in the 1-2 pH range (reflecting mainly contributions from very strong inorganic acids), over the cold season an ∼ 40 % fraction in the 1-3 pH range would reflect possible contributions from other acidic type species (i.e.organics).The observed changes in aerosol acidity could potentially impact the gasparticle partitioning of semi-volatile organic acids.

Figure 1 .
Figure 1.Sector contributions identified from classification of 2-day back trajectories of air masses ending at Iasi and representative backward trajectories of long-and short-range transport, Black Sea influence, and African dust (shown trajectories correspond to sampling events).

Figure 2 .
Figure 2. Patterns of the monthly arithmetic mean concentrations and standard deviations in the PM 10 , PM 2.5 , and PM 2.5 / PM 10 variables at Iasi, north-eastern Romania.

Figure 3 .
Figure 3. Size distribution histograms of aerosol particle mass concentration gravimetrically determined over both the cold and warm seasons.

Figure 4 .
Figure 4. Distribution of the aerosol pH predicted by ISORROPIA-II (forward mode) vs. the ion balance (a), sensitivity of aerosol pH predicted with the model to small changes in the input aerosol NH + 4 concentration (b), and bar chart distribution in aerosol pH over the warm and cold seasons for both NH + 4 derived from raw IC data (c) and NH + 4 (total) (d).

Figure 5 .
Figure 5. Size distribution of averaged aerosol mass and NO − 3 SO 2− 4 , and NH + 4 concentrations over the cold (a) and warm (b) seasons accompanied by the size distribution of ISORROPIA-II-estimated pH and H + mass concentration over both the cold (c) and the warm (d) seasons.

Figure 6 .
Figure6.Seasonal variations for selected water-soluble ionic components in the PM 2.5 fraction (a-h) and variation in the mixing layer depth at the investigated site (i).The inset distribution presented within NO − 3 seasonal variation reflects the contribution of the coarse fraction over the warm season.The horizontal black line represents the mean, the horizontal coloured line the median, the box the 25-75 % percentiles, the length of the whiskers plot 10 and 90 % of observed concentrations, and circles the outliers.

Figure 10 .
Figure 10.Evidence of long-range transport contributions from Saharan dust within the size distribution of particulate Na + , Ca 2+ , Mg 2+ , Cl − ions, and aerosol mass (a) and of air mass buoyancy phenomena within the size distribution of particulate NH + 4 , NO − 3 , SO 2− 4 , Mg 2+ ions, and aerosol mass (b).

Table 1 .
Basic statistics for the PM 10 and PM 2.
5 fractions' mass concentrations determined over the investigated period (n = 84 sampling events) in Iasi, north-eastern Romania.

Table 2 .
Annual and/or seasonal arithmetic means of the PM 10 and PM 2.5 fraction mass concentrations in Iasi, north-eastern Romania, and other various European sites (mean ± SD).
might give a significant pH contribution to ∼ 7 % of the warm season samples and ∼ 10 % of the cold season ones.

Table 3 .
Monthly averages of meteorological variables, PM