Interactive comment on “Chemical characteristics of size resolved atmospheric aerosols in Iasi, north-eastern Romania. Nitrogen-containing inorganic compounds controlling aerosols chemistry in the area” by Alina

This study analyzed the ionic components of size-resolved atmospheric aerosols sampled in northeastern Romania including ion balance, aerosol acidity, formation of ammonium and nitrate, and influence of relative humidity and air mass origins. Measurements of aerosol chemical composition in Eastern Europe are scarce; however, similar measurements at this particular site have been previously reported. The paper is long and I feel the discussions in the paper need to be written more concisely, e.g. especially the discussion of the formation of ammonium and nitrate. The tendency for ammonium C1


Introduction
Despite dramatic progress made to improve air quality, global air pollution continues harming 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), are mainly originating from anthropogenic and natural sources 5 . 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, 2008Sicard et al., 2011;Ostro et al., 2014) as well as air quality (Directive 2008/50/EC, 2008Freney et al., 2014), visibility (Tsai and Cheng, 1999;Directive 2008/50/EC, 2008, ecosystems, weather and climate (Ramanathan şi colab., 2001;Directive 2008/50/EC, 2008IPCC, 2013). Aerosols are also known to play a significant role within the 10 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 Europe. With regard to the PM 2.5 fraction, the EEA Report 5 (2015) indicates that in 2013 the EU daily limit values for PM 10 and PM 2.5 were exceeded in, respectively, 22 and 7 out of 15 the 28 EU 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 13 th leading cause of mortality (Brook, 2008).
Up to now the chemical composition of atmospheric aerosols is reported for various urban European sites (Bardouki et al., 20 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 , 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.) 25 (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 30 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 10 from combustion, condensation of precursor gases) and the complexity of interactions with other atmospheric constituents (Zhang et al., 2015). Nowadays, sources, distribution and behaviour of natural and anthropogenic aerosols are still a matter of debate, exacerbate by the scarcity of aerosols-related work for eastern EU countries (EEA Report 5, 2015) but also by the existent 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 15 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 water-soluble ionic constituents of aerosols (Arsene et al., 2011) and 20 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 (north-eastern 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 of 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 25 factors (relative humidity, temperature), mixing layer depth and emission sources intensity. As a first attempt to assess particles 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 on acid rain. Moreover, the potential importance of gaseous precursors (e.g., NH 3 , HNO 3 , HCl) on 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 Region, "Alexandru Ioan Cuza" University of Iasi, CERNESIM-UAIC, Romania. AMOS is located north-east from the city centre, 5 on the rooftop of the highest University building (~ 35 m above the ground level), in a totally open area that characterises the site as a 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). 10

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 15 50 % calibrated aerodynamic cut-point diameters and at 21.7 o 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 dissembled in a Labguard Class II Safety 20 Biological 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 weekend and working days, for a total of 84 sampling events (41 during the cold seasons from October to March, and 43 during the warm seasons 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. It was collected an average volume of 64.33 ± 0.85 m 3 per sampling. At least two field blanks (consisting of loaded sampler taken 25 to and from the field but never removed from its tightened bag) were generated and simultaneously analysed together with the laboratory blanks, in order to assess possible contamination during sampler loading, transport, or analysis.
Meteorological parameters including atmospheric temperature (AT), relative humidity (RH), wind speed (WS), 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 30 site was obtained from the NOAA Air Resources Laboratory (ARL) website (Stein et al., 2015;Rolph et al., 2017).
After sampling and all other required preparative steps, one half of each collected filter was ultrasonically extracted for 45 minutes in 5 mL deionized water (resistivity of 18.2 MΩ.cm) produced by a Milli-Q Advantage A10 system (Millipore).
Traceable standard solutions (Dionex Seven Anions II and Dionex Six Cations II) were used to generate calibration curves for each species of interest, all having correlation coefficients R 2 well above 0.995. The detection limits of the major species 20 (defined as 3 times the standard deviation of blank measurements relative to the methods sensitivity) on a 36 h measurement period were: 0.0003 µg m -3 for NH 4 + (3.2 µg L -1 ), 0.001 µg m -3 for Na + (13.4 µg L -1 ), 0.0015 µg m -3 for K + (19.8 µg L -1 ), 0.0002 µg m -3 for Mg 2+ (3.2 µg L -1 ), 0.0016 µg m -3 for Ca 2+ (20.6 µg L -1 ), 0.0010 µg m -3 for SO 4 2-(13.2 µg L -1 ), 0.0012 µg m -3 for NO 3 -(15.1 µg L -1 ), and 0.0019 µg m -3 for Cl -(24.9 µg L -1 ). Ions concentrations in analysed blank filters (laboratory and field) were subtracted from their corresponding concentrations in the aerosol samples. 25 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 aerosols particles or of the analysed ionic components (individual or as a sum) were determined by fitting lognormal distributions.

Estimation of the aerosols acidity 30
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. ISORROPIA-II thermodynamic equilibrium model calculates the gas/liquid/solid equilibrium partitioning of K + , Ca 2+ , Mg 2+ , NH 4 + , Na + , SO 4 2-, NO 3 -, Cland aerosol water content, and it can predict particles 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 aerosols pH, ISORROPIA-II was run in the "forward mode" for metastable aerosol state, as preliminary runs of the experimental data in the "reverse mode" did not supply suitable information. In the "metastable" 5 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 aerosol-phase ion concentrations measured by IC, along with relative humidity 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 Hennigan et al., 2015), which seem to be less 10 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 HYSPLIT 4 model of the NOAA Air Resources Laboratory (Stein et al., 2015;Rolph et al., 2017). Forty-eight hour back trajectories, arriving at the investigated site at 18:00 local time (15:00 UTC), 15 were computed at 500, 1000 and 2000 m altitudes above the ground level. Four major sectors of air masses 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 is mainly prevailing during winter, while the N-E 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, 20 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/continentally driven buoyancy (Holton, 1979;Seinfeld and Pandis, 1998), travelling above large (long range transport) or short (local) continental areas were also indentified. 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 meters (brushing the ground 25 surface), with strong impact on the chemical composition of the collected particles (vide infra).

Results and discussion
3.1 Variability of 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 30 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 data-base was normally distributed (detailed statistics are given in Table S 1 from Supplement Material, SM). 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 5 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 ± stdev). 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 10 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 transport phenomena, and air masses buoyancy is presented in detail in Section S 1 of the SM. 15 result of seasonal emissions variations, local-and long-range air transport and dispersion, chemical processes, and deposition . The maxima of PM 2.5 /PM 10 during the cold seasons are most probably caused by combustion processes as the burning of coal/petroleum for heating purposes enhances secondary aerosols generation (Li et al., 2012). The lower PM 2.5 /PM 10 values during the warm seasons 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 both cause higher loading of coarse 25 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 seasons and hence their fine/coarse ratio.
Another clear seasonal pattern was observed for the mass concentration size distribution (FIG. 3). Over the cold seasons, it had a clear monomodal feature with maximum at 381 nm. In contrast, the warm seasons were characterized by the same 30 dominant fine mode, but also by the occurrence of a super-micron mode between 1.60-2.39 µm. Again, changes in sources contributions and meteorological conditions could account for the observed differences (details in Section S 2 of the SM).
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 aerosols acidity
The completeness of the ionic balance was checked for the identified and quantified species (F -, Cl - 10 and PM 2.5 . The slopes in the raw ion chromatography data related to ∑ cations vs. ∑ anions were < 1 in both PM 2.5 and PM 10 (detailed statistics in Table S 2), pointing 5 to an important cation deficit that was higher in the cold compared to the warm seasons. 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, it was even observed RH ~ 82 % during the cold seasons.
Predicting pH is suggested as the best method to analyse particle acidity . The ion balance method is usually based upon the principle of electroneutrality, and any deficit in measured cationic compared to anionic charge is 10 assigned to the presence of unmeasured protons (H + ). The reverse occurs for unmeasured hydroxyl (OH -) (Hennigan et al., 2015)  In an attempt to investigate whether or not the H + species would bring an important contribution within the ionic balance, ISORROPIA-II thermodynamic equilibrium model proposed by Fountoukis and Nenes (2007) has been used in the present work (more details about ISORROPIA-II runs in Section S 5 of the SM). We investigated the relationship between ISORROPIA-II predicted aerosol pH and the ionic balance for the present data-base, 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 20 ∼ neutral (dashed lines at 0). A very important fraction of the investigated samples was in the acidic range (pH < 3 if samples were in cation deficit mode), while the remaining fraction was alkaline (pH slightly above 7 with an 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 ∼ 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 25 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 toward NH 4 + concentrations are given in Section S 6.
In Iasi, north-eastern Romania, an important fraction of the total analysed samples was alkaline while the remaining part was acidic. A more detailed view of the samples pH distribution can be obtained from the data presented in FIGS. 4c,d. It seems 30 that over the warm seasons 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-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 aerosols pH in the 0-3 range. Note that strongly acidic aerosols affect air quality, health of aquatic and terrestrial ecosystems (especially through acid deposition), as well as 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 5 Section S 7. Moreover, aerosols 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 prevents 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) 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 20 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) is expected to be more important in all other remaining fraction, and especially in the 25 27.6-94.5 nm particles size range (vide infra). Of course, a higher confidence in the estimate of particles 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 there is suggestion 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 the results, either due to exclusion of minor ionic species or 30 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 molar ratio NH 4 + /(Cl -+ NO 3 -+ 2×SO 4 2-) < 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 particles 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 5 submicron particles (seconds to minutes; Meng et al., 1995). As previously mentioned, in the present work the number of samples with pH < 0 was significantly lower compared to those with pH > 0. Moreover, ISORROPIA-II predicted that in the 94.  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 4 2and NO 3 concentrations, and by temperature affecting both SO 4 2and 10 NO 3 production, is given in Section S 8 of the SM. 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. In other studies, thermodynamic equilibrium calculations predicted that all of the NH 3 was mainly susceptible of partitioning 25 to the particle phase at the equilibrium, and also that > 44 or 51 % of the investigated samples presented aerosols 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 on 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. S 1 (Section S 9). The same section reports considerations on possible interrelated emission factors governing the distribution in the NH 3 concentration levels in 30 Iasi.
The ISORROPIA-II data referred to the 155-612 nm size range (regardless of RH) suggested that the aerosol ammonium fraction (NH 4 + /(NH 3 + NH 4 + )) was over 0.20 irrespective of the calculation procedure (raw IC or NH 4 + (total)). Clear (NH 4 + /(NH 3 + NH 4 + )) maxima were observed at 381 nm (raw IC data: 0.71 at RH < 40 %, 0.63 at RH = 40-60 %, and 0.76 at RH > 60 %; NH 4 + (total), defined as the sum between that derived from raw IC data and the part estimated by using the rationale of Arsene et al. (2011): 0.66 at RH < 40 %, 0.56 at RH = 40-60 %, 0.65 at RH > 60 %). As seen in FIGS. 5c,d, the aerosol pH is very low in the 155-612 nm size range, while it often approaches 8 in the 27.6-94.5 and 612-9940 nm size ranges. Although not presented, in the 155-612 nm size range the aerosol ammonium fraction (NH 4 + /(NH 3 + NH 4 + )) approaches 1, while in other two size ranges it is very low or close to 0 thereby suggesting the occurrence of gaseous NH 3 . 5 Table 3 shows monthly-based 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)):   (inducing lower mixing heights or even temperature inversion), or due to different chemical/photochemical processing.

Seasonality of the major water-soluble ions
Although lowering mixing heights over the cold seasons 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 4 2-, 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 20 elevated occurrence of oxidant species such as ozone, hydroxyl and nitrate radicals. These conditions favour the oxidation of SO 2 to particulate SO 4 2- . Also particulate C 2 O 4 2was maximum in summer, possibly due to enhanced photochemical processing. Moreover, the maxima observed for SO 4 2during the cold seasons might be due to the intensification of coal burning for heating purposes. Higher abundances of particulate NO 3 -, SO 4 2-, 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). 25 Sharma et al. (2007) also suggest a potential role of CaCO 3 in controlling particulate NO 3 abundance in Kanpur, India. 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 Clmass concentrations show a clear seasonal pattern, with higher values during the cold than during the warm seasons (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 S 4). The 5 chloride maxima during the cold seasons might be the result of increased coal burning for heating purposes or of the use of NaCl in winter on icy/snowy roads. These observations are in agreement with other studies in eastern European sites (Arsene et al., 2011;Alastuey et al., 2016). However, the Clmass concentration follows a similar pattern as that of K + (tracer of biomass burning), thereby suggesting that that over the cold seasons wood burning might become an important heating source (Christian et al., 2010;Akagi et al., 2011). 10 Also nitrate 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 seasons (3.62 ± 1.10 µg m -3 in January, 15 February, December as the coldest months of the year, and 2.65 ± 0.38 µg m -3 over January, February, March, October, November, and December) seem to be in reasonably good agreement with those predicted for Europe by Backes et al. the NH 4 + case, this might reflect the susceptibility of NO 3 to be transferred to the gas phase over the warm seasons.
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 particles loading (Table 3, detailed statistics in Table S 4). 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 - 25 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 particulate 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 seasons might also be caused by higher NH 3 atmospheric levels from yet unaccounted sources, 30 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/spreading) is most likely an important NH 3 source.
Moreover, the high relative humidity observed in the cold seasons 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., , 2011Ianniello 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 4 2maxima are observable during both cold and warm seasons. However, the particulate SO 4 2mass concentrations showed statistically significant correlation with the measured meteorological parameters only at a 68 % 5 confidence level (detailed statistics in Table S 4). The data presented in FIG. 6c show that the seasonal trend of the monthly SO 4 2mean mass concentrations is not as clear as that of NO 3 and NH 4 + , which might suggest the occurrence of regional SO 4 2sources as well . Moreover, high RH (in Iasi, especially during cold months) may aid the conversion of SO 2 to SO 4 2- (Kadowaki, 1986), with a significant enhancement of SO 4 2production rate in 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 10 also by gas-phase radical hydroxyl (Vione et al., 2003). This issue can explain the PM sulfate maxima during the warm seasons 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 seasons, in agreement with reports at other European (Schwarz et al., 2012;Bressi et al., 15 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 to 9.04 µg m -3 among 20 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 seasons 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 25 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 S 4). The PM 2.5 fraction showed statistically significant correlation also with the mixing depth (Pearson coefficient higher than 0.67, p = 0.016). The seasonal variation of particulate NH 4 + (total) follows especially those of particulate NO 3 and Cl -, which would indicate that most probably NH 4 + (total) largely originates from neutralization between NH 3 , HNO 3 and HCl (Wang et al., 30 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, independently of the RH values ([NH 3 ]/([HNO 3 ] + [HCl]) >> 1).
At the investigated site, C 2 O 4 2and SO 4 2show similar behaviour and the C 2 O 4 2maxima 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 compared to the 5 warm seasons (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 similar pattern as Cl -(Pearson coefficient of 0.79, p = 0.002) but a different one compared to SO 4 2-. Therefore, we suggest that intense wood burning may be a common source for K + and Clspecies in the study area (Christian et al., 2010;Akagi et al., 2011). 10 High mass concentrations of Ca 2+ and Mg 2+ (with Mg 2+ shown in Table 3 but not in FIG. 6), as soil/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 seasons, 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. 15 likely explanation would rather be the use of both coal and wood for burning, as well as the effect of the mixing layer depth.

Stoichiometry of (NH 4 ) 2 SO 4 , NH 4 NO 3 and NH 4 Cl
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). FIGURES 7a,b present the relationship between the 30 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 4 2for both the cold (FIG. 7a) and warm (FIG. 7b)  ]) molar ratio was either ~ 1 (raw IC data), or (NH 4 + (total) values) equal to 1.76 (cold seasons) and 1.02 (warm seasons). 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 Unfortunately, at present, no measured NH 3 values are available for the interest site 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 15 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 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 20 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 wintertime atmosphere, and also that the fine collected particulates were almost fully neutralized by NH 3 . 7c,d,e,f it is quite clear that at the study site, if NH 4 + derived from raw IC data is used, the neutralisation ratios in the 30 investigated particles are less than unity. This suggests that the atmospheric particles are most likely acidic, and also that a more complex chemistry is ongoing involving the HNO 3 and HCl species. However, when NH 4 + (total) is used, the neutralisation ratio approaches 1 and suggests a possibly complete neutralization of particles acidity. From details in FIGS.
7e,f it can be easily seen that actually the available NH 4 + is not enough to compensate for other species. However, it should be noted that Clis not significantly influencing the neutralization of particles acidity. In a study performed by Zhao et al.  . 7c). In contrast, during the warm seasons (FIG. 7d) the molar ratio is slightly lower than 1 (i.e., 0.95). According to Seinfeld and Pandis 5 (1998), during the warm seasons the high temperature and the low relative humidity would be favourable for NH 4 + to reach a minimum concentration, because it is mainly transformed into NH 3 . Actually, temperature values above 25 °C, such as those often encountered at the investigated site over the warm seasons, are known to prevent formation of particulate NH 4 NO 3 (Adams et al., 1999). Under these circumstances, a considerable decrease in the [NH 4 + ](total)/([NO 3 -] + 2×[SO 4 2-]) molar ratio is to be expected during warm seasons. The cold-season temperature and relative humidity at the investigated site were, 10 respectively, 5.3 ± 3.9 °C and (65.3 ± 12.8) %. Coherently, the data presented in Table 4 show that the (NO 3 -, NH 4 + (total)) pair has significant correlation (Pearson coefficient of 0.98, p < 0.001) only during the cold seasons, while during the warm seasons (temperature and relative humidity of, respectively 18.9 ± 3.8 °C and 40.5 ± 7.7 %) the correlation is very poor as increasing temperature and decreasing relative humidity limit the production of the NH 4 NO 3 aerosol (Matsumoto and Tanaka 1996; Utsunomiya and Wakamatsu 1996; Alastuey et al., 2004). 15 In the atmosphere, both H 2 SO 4 and HNO 3 are known to compete for the reaction with NH 3 to form (NH 4 ) 2 SO 4 and NH 4 NO 3 .
As presented in Sect. 3.2.2, at the investigated site a reduction in both in NO 3 and SO 4 2was observed over the warm seasons and this may cause a further increase in gas-phase NH 3 . It is also interesting to observe that the reaction rate constant of (NH 4 ) 2 SO 4 aerosol formation is similar to the rate constant of NH 4 NO 3 formation (Harrison and Kitto, 1992;Pandolfi et al., 2012;Behera et al., 2013), and both are much higher than the rate constant between NH 3 and HCl (Behera and Sharma, 20 2012 During the warm seasons, 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 30 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 northwest 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 gas-phase HNO 3 during winter. In our study, results from ISORROPIA-II thermodynamic model foresee an increase of gas-phase HNO 3 at higher RH values during cold seasons. Higher levels of gas-phase HNO 3 may increase its condensation onto existing particles such as sodium chloride (NaCl), and the replacement of Clwith NO 3 may enhance the concentration of HCl in the atmosphere (similar processes are described in Arsene et al., 2011). 5 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 Clto be chemically bound to Ca 2+ , Mg 2+ , K + or Na + . The free NO concentrations, defined as the fractions of excess nitrate and chloride that are not bound to alkali or alkaline earth metals, were estimated for both the cold and warm seasons according to the concept described in Ianniello et al. (2011). Zero or negative values of free NO 3 and Climply that NH 4 NO 3 and NH 4 Cl are not present. Estimated free NO 3 and Cl -10 concentrations showed similar contributions in both the PM 2.5 and PM 10 fractions. Over the cold seasons we calculated (6.5 ± 7.9)×10 -3 µmol m -3 (0.4 ± 0.5 µg m -3 ) for NO 3 -, and negative values for free Cl -; over the warm seasons we had (1.6 ± 1.8)×10 -3 µmol m -3 (0.1 ± 0.1 µg m -3 ) for NO 3 -, and again negative values for free Cl -(data provided as mean ± stdev). This result suggests the potential presence of NH 4 NO 3 , especially during the cold seasons, but it excludes the occurrence of NH 4 Cl. During the cold seasons, particulate NO 3 didn't show correlation with either Ca 2+ or Mg 2+ , but it showed significant 15 correlation with K + (r = 0.85, p < 0.001, see  , 1997). The (NH 4 ) 2 SO 4 aerosol particles might 25 remain in "metastable" or supersaturated state liquid phase until a very low RH (crystallization point, RH around 33%) is reached (only thereafter a solid might be formed). In contrast, for NH 4 HSO 4 it has been shown that the solid phase is difficult to be formed. In the present work, it is very likely that only over July (RH of 32.95 %), August (RH of 37.12 %) and September (RH of 31.56 %) the formation of solid (NH 4 ) 2 SO 4 or NH 4 HSO 4 could occur.
Pure NH 4 NO 3 deliquesces at 62 % RH and there is suggestion that sometime even at 8 % RH the crystallization point is not 30 reached (Dougle et al., 1998). According to suggestions from the literature, only during months when the ambient RH is lower than the relative humidity at deliquescence (RHD), the NH 4 NO 3 is considered as a solid (Seinfeld and Pandis, 1998).
In the present work, from March to October the ambient RH was always lower than the RHD and NH 4 NO 3 could thus be assumed to be in equilibrium with the solid phase. Within all the other months, from the estimated RHD values there is the possibility that NH 4 NO 3 is in equilibrium with the aqueous phase and deliquescent particles. However, in the investigated site, solid NH 4 NO 3 could be formed almost all over the year either due to the very complex chemical composition of the collected particles, or due to abundant contribution of organic carbon to the particles mass concentration (Dougle et al., 1998). Formation of NH 4 NO 3 over the warm seasons has been reported also by Ianniello et al. (2011) but under different conditions. For deliquesced particles there is suggestion that most of the fine particulate NO 3 occurs as an internal mixture 5 with SO 4 2-, and also that HNO 3 can easily be absorbed into the droplets (Huang et al., 2010). In specific circumstances, the fine particulate NO 3 can be formed from HNO 3 and NH 3 through heterogeneous reactions on fully neutralized fine particulate SO 4 2-, which is abundant in urban areas (Stockwell et al., 2000). In the present work, a statistically significant correlation between SO 4 2and NO 3 was observed during the cold seasons in the PM 2.5 fraction (r = 0.91, p < 0.001). High concentrations of NO 3 were found in the presence of elevated RH levels (significant correlation: r = 0.84, p < 0.001), while 10 SO 4 2concentrations were high over the entire RH range (non-significant correlation: r = 0.44, p = 0.177). These results can be interpreted as nitrate being produced on pre-existing sulfate aerosols, which could provide sufficient surface area and water content for the heterogeneous reactions to occur. Although the formation of fine particulate NO 3 can take place via reaction (R2), in particular circumstances and especially at high RH values, the amounts of the gaseous precursors such as NH 3 and HNO 3 may have relatively little influence on the fine-particulate NO 3 formation (Markovic et al., 2011). 15 The salt NH 4 Cl is known to be 2-3 times more volatile than NH 4 NO 3 , as HCl is more volatile than HNO 3 . Moreover, at RH < [75][76][77][78][79][80][81][82][83][84][85]particulate NH 4 Cl is in equilibrium with the gaseous compounds (Ianniello et al., 2011 and references therein apart from alkali or alkaline earth metals, thereby excluding a significant occurrence of NH 4 Cl (especially over the warm seasons).

Relative ionic contribution in size resolved aerosol particles from Iasi and potential influence of long-range transport phenomena on particles size distribution
FIGURES 8a,b,c,d present, as monthly based averages, the relative contributions of identified and quantified water-soluble 30 ions to total detected components in fractions grouped in four stages, i.e., 0.0276-0.0945 µm size range (FIG. 8a), 0.155-0.612 µm size range (FIG. 8b), 0.946-2.39 µm size range (FIG. 8c) and 3. .94 µm size range (FIG. 8d). From details presented in FIG. 8a, for the 0.0276-0.0945 µm fraction there is important contribution of formate, acetate and oxalate that may actually indicate a possible important role of organic acids in secondary organic aerosols formation. Higher values of these components over the warm seasons may suggest an enhancement in the role of biogenic emission sources. Important contributions are brought by SO 4 2-, NH 4 + (total), K + and even by the (unexpectedly high) HCO 3 -. However, high particulate HCO 3 is also evident for the 0.946-2.39 µm (FIG. 8c) and 3. .94 µm size range (FIG. 8d) fractions. The 0. 155-0.612 µm fraction (FIG. 8b) seems to be mainly constituted by SO 4 2-, NO 3 and NH 4 + (total), with very small contributions from 5 other ions.
The seasonal variation observed mainly for SO 4 2and NO 3 might suggest an enhancement of photo-oxidative processes over the warm seasons. The 0.946-2.39 µm (FIG. 8c) and 3. .94 µm size range (FIG. 8d) 10 and 18.2 ± 5.1 % in PM 2.5 in July 2016, much higher than that reported by Arsene et al., 2011, for Iasi). The difference might reflect either an inversion of the photochemistry taking place at the investigated location, or differences in the sampling efficiency between the two studies. , a larger modal diameter in the cold compared to the warm seasons is most likely due to hygroscopic growth under high RH, and/or to increased secondary aerosol production (lower temperatures facilitate condensation). Secondary aerosol mass from aqueousphase reactions may also play a role (Wonaschutz et al., 2015). The maxima observed in the coarse mode could also be explained considering that heterogeneous chemistry on dust particles might act as a source for some particulate species 5 (Wang et al., 2012).
While particulate NO 3 over the cold seasons presented monomodal distributions in the sub-micron size range (maxima at 381 nm), over the warm seasons a second mode was observed with maxima in the 1.60-2.39 µm size range. Such a size distribution suggests that NO 3 during the warm seasons is produced by adsorption of HNO 3 on sea salt and soil particles (Park et al., 2004). According to Karydis et al. (2016), particulate NO 3 is not associated only with NH 4 + in the fine mode. 10 Light metal ions such as Ca 2+ , Mg 2+ , Na + , and K + , which mainly occur in the coarse mode, can be associated with NO 3 and affect its partitioning into the aerosol phase. Dust effects on the distribution of particulate species might include a decrease of fine-mode NH 4 + and a shift of particulate NO 3 from the fine to the coarse mode (Wang et al., 2012). In addition, the presence of significant levels of NO 3 -, Cl -, Mg 2+ , Ca 2+ and Na + in the coarse fraction might suggest that NO 3 possibly originates upon reaction of HNO 3 with MgCO 3 , CaCO 3 or NaCl. Similar patterns were identified in Vienna, Austria 15 (Wonaschutz et al., 2015) and in Prague, Czech Republic (Schwarz et al., 2012). Significant amounts of particulate NO 3 formed upon reaction of HNO 3 with CaCO 3 on soil-derived particles have also been observed and reported (Yao et al., 2003;Sharma et al., 2007). Moreover, sea-salt aerosols may also undergo chemical transformation of NaCl into NaNO 3 during transport (Schwarz et al., 2012).
The size distributions of particulate K + reflect the occurrence of one dominant fine mode (with maxima at 381 nm) during 20 both the cold and the warm seasons, and of a second less important mode during the warm seasons (with maxima in the 0.946-1.6 µm range). Such behaviour most likely reflects contributions from biomass burning all over the 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 seasons, 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 seasons it accounts for only (3.0 ± 0.6) % of 25 the PM 10 and (2.2 ± 0.5) % of the PM 2.5 fraction. These observations indicate that the impact from soil dust re-suspension could be more important during the warm (dry) seasons. Mineral dust may also explain the higher coarse fraction of Mg 2+ (mineral source being MgCO 3 ).
Clear evidences were obtained in this work that air-mass origin highly influences the aerosol chemical composition at the  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 seasons 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 5 (10-60 µg m -3 ) (Karydis et al., 2016). In the present work, particulate Na + and Clions 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 S-SE directions. However, one of the most interesting collected events was that of 9 th -11 th 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 two 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 10 Mediterranean Sea. As shown in FIG. 10a, the size distributions of particulate Na + , Ca 2+ , Mg 2+ , Clions 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 Clmake a very significant contribution to the total aerosol mass in the super -15 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 4 2and Mg 2+ (FIG. 10b). This month is highly affected by the atmospheric air masses buoyancy phenomenon, as shown by trajectories analysis for selected events and, while particulate NH 4 + and SO 4 2were mainly residing in the fine mode with clear maxima at 381 nm, NO 3 and Mg 2+ also presented a predominant mode in the 1.6-2.39 µm fraction. Such distributions, 20 corroborated with meteorological conditions, would actually suggest a possible heterogeneous formation route for SO 4 2- (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, 25 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 by means of a cascade Dekati Low-Pressure Impactor (DLPI), performing aerosol size classification in 13 specific fractions evenly distributed over the 0. .94 µm size range.
The entire data set was analysed to investigate the seasonal variations in fine particulate species and the meteorological used to estimate the pH values of the collected atmospheric particles, as on the present data-base it was the best method to analyse particles acidity.
Within the aerosol mass concentration, the identified ions mass brings contributions as high as 40.6 % with the rest being unaccounted yet. Fine particulate Cl -, NO 3 -, NH 4 + and K + exhibited clear seasonal variations with minima during the warm seasons, mainly due to cold-season enhancement in the emission sources, changes in the mixing layer depth and specific 5 meteorological conditions (e.g., higher RH values prevailing in Iasi during the cold seasons). Fine particulate SO 4 2did 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 seasons but not for the warm seasons. 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 4 2and NO 3 -. However, indirect 10 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 neutralise 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 seasons. 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 relative humidity conditions in the cold seasons (likely leading to dissolution of a 15 significant fraction of atmospheric HNO 3 and NH 3 ) are among the most important driving forces enhancing the fineparticulate NO 3 and NH 4 + distribution in 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 Clin PM 2.5 via neutralisation processes. The chemical composition data-base 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 20 present in sufficiently high concentration to promote fine particle acidity neutralisation, 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.