Nitro-polycyclic aromatic hydrocarbons-gas-particle partitioning , mass size 1 distribution , and formation along transport in marine and continental background air 2 3

14 Nitro-polycyclic aromatic hydrocarbons (NPAH) are ubiquitous in polluted air but little is 15 known about their abundance in background air. NPAHs were studied at one marine and one 16 continental background site i.e., a coastal site in the southern Aegean Sea (summer 2012) and 17 a site in the central Great Hungarian Plain (summer 2013), together with the parent 18 compounds, PAHs. A Lagrangian particle dispersion model was used to track air mass history. 19 Based on Lagrangian particle statistics, the urban influence on samples was quantified for the 20 first time as a fractional dose to which the collected volume of air had been exposed to. 21 At the remote marine site, the 3-4 ring NPAH (sum of 11 targeted species) concentration was 22 23.7 pg m -3 while the concentration of 4-ring PAHs (6 species) was 426 pg m -3 . 223 nitrofluoranthene (2NFLT) and 3-nitrophenanthrene were the most abundant NPAHs. Urban 24 fractional doses in the range <0.002–5.4% were calculated. At the continental site, the Σ11 325 4rNPAH and Σ6 4rPAH were 58 and 663 pg m -3 , respectively, with 9-nitroanthracene and 26 2NFLT being highest concentrated amongst the targeted NPAHs. The NPAH levels observed 27 in the marine background are the lowest ever reported and remarkably lower, by more than 28 one order of magnitude, than one decade before. Day-night variation of NPAHs at the 29 continental site reflected shorter lifetime during the day, possibly because of photolysis of 30 some NPAHs. The yields of formation of 2NFLT and 2-nitropyrene (2NPYR) in marine air 31 seem to be close to the yields for OH-initiated photochemistry observed in laboratory 32 experiments under high NOx conditions. Good agreement is found for prediction of NPAH 33 gas-particle partitioning using a multi-phase poly-parameter linear free energy relationship. 34 Sorption to soot is found less significant for gas-particle partitioning of NPAHs than for 35 PAHs. The NPAH levels determined in the southeastern outflow of Europe confirm 36 intercontinental transport potential. 37 38

Secondary formation of NPAH from PAHs is thought to occur on short timescales (hours).This has been observed Published by Copernicus Publications on behalf of the European Geosciences Union.
Like their precursors, NPAHs are semi-volatile organic compounds (SVOCs), partitioning between the phases of the atmospheric aerosol.Similar to other SVOCs, the NPAHs' phase distribution was found to depend on temperature (summer and winter campaigns in the Alps; Albinet et al., 2008b) and results from both absorptive and adsorptive contributions (Tomaz et al., 2016).NPAHs have primarily been observed in polluted areas (e.g.Pitts et al., 1985;Ramdahl et al., 1986;Garner et al., 1986;Albinet et al., 2007Albinet et al., , 2008a;;Ringuet et al., 2012a, b;Zimmermann et al., 2012;Barrado et al., 2013;Li et al., 2016), though there are a few studies in rural environments, e.g. in Germany (Ciccioli et al., 1996), in the French Alps (100-1000 pg m −3 range for the sum of 10 NPAHs; Albinet et al., 2008a), and in northern China (Li et al., 2016).Very few measurements have been performed in a remote atmospheric environment, i.e. in the Mediterranean (Tsapakis and Stephanou, 2007), high-altitude sites in the Himalayas (single data; Ciccioli et al., 1996) and French Alps (Albinet et al., 2008a), and in the Arctic (with so-called Arctic haze; Masclet et al., 1988;Halsall et al., 2001).With regard to the long-range transport potential, the consensus is that at least some NPAHs are expected to go into intercontinental transport (Lafontaine et al., 2015) and might be ubiquitous in the global atmosphere (Ciccioli et al., 1996).
However, there are limited NPAH data from remote atmospheric environments and little is known about their longrange transport potential.The aim of this study was to characterise the long-range transport potential of NPAHs by measurements at remote sites in Europe, addressing the continental background and the outflow of the continent.

Sampling
High-volume air sampling was conducted at a marine background site, Finokalia (35.3 • N, 25.7 • E; 250 m a.s.l.), in the context of a coordinated field experiment from 2 to 13 July 2012 (Lammel et al., 2015) and at a continental background site in central Europe, K-puszta (46 • 58 N, 19 • 33 E; 125 m a.s.l.; Degrendele et al., 2016), from 5 to 16 August 2013.The Finokalia site is located on a cliff at the northern coast of Crete, some 70 km east of major significant anthropogenic emissions (Iraklion, a city of 100 000 inhabitants with airport and industries; Mihalopoulos et al., 1997;Kouvarakis et al., 2000).The K-puszta site is located in a clearing, characterised by uncultivated grassland, in a mostly coniferous forest in the Hungarian (Pannonian) Great Plain, ca.70 and 270 km south-east of Budapest and Vienna, respectively (≈ 2 mn inhabitants each).The background site character of both observatories has been previously demonstrated (Borbély-Kiss et al., 1988;Kouvarakis et al., 2000;Vrekoussis et al., 2005).Meteorological parameters and trace gases are measured at both observatories, which are stations of the EMEP network (EMEP, 2015).
Particle number concentration, N, was determined by an optical particle counter (Grimm model 107, Ainring, 31 channels between 0.25 and 32 mm of aerodynamic particle diameter).Aerosol surface concentration, S (cm −1 ), was derived as S = π i N i D 2 i assuming sphericity.Hereby, true S will be underestimated, in particular if particles of irregular form were abundant (e.g.Jaenicke, 1988).Comparisons with absolute methods (e.g.Pandis, et al., 1991) suggest that the discrepancy may reach up to a factor of 2-3.The mass median diameter (D m , µm) was derived as log D m = i m i log D i / i m i , with m i denoting the mass in size class i and D i being the geometric mean diameter collected on stage i of the cascade impactor.

Chemical analysis
All air samples were extracted with DCM using an automatic warm Soxhlet extractor Switzerland).Deuterated PAHs (D8-naphthalene, D10-phenanthrene, D12perylene; Wellington Laboratories, Canada) were used as surrogate standards for both PAHs and NPAHs.Deuterated PAHs proved to be suitable surrogate standards for NPAHs.These were spiked on each PUF prior to extraction.The extract was split in two parts: 1/10 for PAHs and nitro-PAHs analysis and 9/10 for PBDEs, PCBs, and OCPs.The PAHs and nitro-PAHs aliquot were subject to open column chromatography clean-up.A glass column (1 cm i.d.) was filled with 5 g activated silica (150 • C for 12 h) and the sample was loaded and eluted with 10 mL n-hexane, followed by 40 mL DCM.The cleaned sample was evaporated under a stream of nitrogen in a TurboVap II apparatus (Biotage, Sweden), transferred into a conical GC vial, and spiked with recovery standard, terphenyl; the volume was reduced to 100 µL.
Recovery of native analytes varied 72-102 % for PAHs and deuterated PAHs and 70-110 % for NPAHs (for details see Supplement, Table S1a).The results were not recovery corrected.The mean of field blank values was subtracted from the sample values.Values below the mean + 3 standard deviations of the field blank values were considered to be < LOQ (limit of quantification).Field blank values of some analytes in air samples were below the instrument limit of quantification (ILOQ), which corresponded to 0.004-0.069pg m −3 for NPAHs (except for 1NNAP, for which it ranged from 0.60 to 0.87 pg m −3 ) and 0.010-0.126pg m −3 for 4-ring PAHs (except for FLT and PYR, for which it ranged from 0.17 to 0.59 pg m −3 ) (Supplement Table S1).
Particulate matter mass (PM 10 ) was determined by gravimetry (microbalance, filters accommodated to stable temperature and humidity, three replicate weighings) and organic matter (OM) and elemental carbon (EC) contents of PM by a thermal-optical method (Sunset Laboratory, USA; EUSAAR protocol).

Gas-particle partitioning
Gas-particle partitioning was studied by applying a multiphase ppLFER model, which was recently introduced (Shahpoury et al., 2016).In brief, partitioning of semi-volatile compounds in air can be described (Yamasaki et al., 1982) by where K p (m 3 air (g PM) −1 ) is the temperature-dependent partitioning coefficient, c PM (g m −3 ) is the concentration of particulate matter in air, and c ip and c ig are the analyte (i) concentrations (ng m −3 ) in the particulate and gas phase, respectively.K p can be predicted using models based on single-parameter and poly-parameter linear free-energy relationships (spLFER, ppLFER).spLFERs relate the partitioning coefficient to one physicochemical property, i.e. assume one process to determine the sorption process, while ppLFERs in principle account for all types of molecular interactions between solute and matrix (Goss and Schwarzenbach, 2001).The observed particulate mass fraction data, θ = c p /(c g + c p ) (Table 2), were tested with both a spLFER and a ppLFER model.The spLFER chosen is the widely used K oa model of Finizio et al. (1997) (results presented in the Supplement, Sect.S2.3).The ppLFER is a multi-phase model recently presented (Shahpoury et al., 2016) and applied for NPAHs (Tomaz et al., 2016).It is based on linear solvation energy relationships (Abraham, 1993;Goss, 2005): where E, S, A, B, L, and V are solute-specific Abraham solvation parameters for excess molar refraction (describes interactions between π and lone (n) electron pairs), polarizability/dipolarity, solute H-bond acidity, solute H-bond basicity, logarithm of solute hexadecane-air partitioning coefficient (unitless), and McGowan molar volume (cm 3 mol −1 )/100, respectively (Endo and Goss, 2014).The corresponding parameters e, s, a, b, l, and v reflect matrix-specific soluteindependent contribution to K p .Due to the lack of experimental data, the solute descriptors for NPAHs were taken from M. H. Abraham (personal communication, 2015).
The multi-phase ppLFER considers adsorption onto soot, (NH 4 ) 2 SO 4 , and NH 4 Cl and absorption into particulate OM.
OM is assumed to be constituted of two separate phases.For these, ppLFER equations for dimethyl sulfoxide-air (representing the low molecular mass range of both organicsoluble and water-soluble OM) and for polyurethane etherair (representing the high-molecular-mass organic-soluble OM) are used (Shahpoury et al., 2016).A conventional single-parameter LFER (K oa ) model is applied, too.

Air mass history analysis
The HYSPLIT (Draxler and Rolph, 2003) and FLEXPART (Stohl et al., 1998(Stohl et al., , 2005) ) models were used to identify air mass histories over 10 and 2 days, respectively.The possible influence of polluted air on samples was quantified using a novel method of applying Lagrangian particle statistics (FLEXPART, see Sects.2.5 and S2.1.2).To this end, for the entire sampling period, one particle per second was released.The model output is generated at 0.062 • (≈ 7 km), every 30 min, and expressed as "residence time", i.e. a measure of the time particles resided in grid cells.ECMWF meteorological data (0.125 • × 0.125 • resolution, hourly) were used as input.

Quantification of urban influence on samples
The potential urban influence for individual samples collected at the marine site was based on the fraction of released Lagrangian particles that travelled through an urban boundary layer.A backward run from the sampling site was performed with Lagrangian particles (i.e.air parcels) being released during the entire sampling period.Three urban areas were considered: Izmir (≈ 300 km direct distance; 38.2-38.8• N, 26.2-27.3• E), Athens (≈ 300 km; 37.8-38.1 • N, 23.5-23.8• E), and Istanbul (≈ 500 km away; 40.8-41.1 • N, 28.6-29.5 • E).
The urban fractional dose, D u i , which an air mass collected in sample i had received during a given simulation period t, can be derived as where N blua (t) is the number of virtual particles within the urban boundary layer during the specific time step, model output time resolution t Rblua is 0.5 h, and N tot (t) is the number of virtual particles present during the specific time step.Under the given flow conditions in the region, a 2-day time horizon is considered here.Hence, the simulation period is given as with t sample being the sampling time.D u i takes values between 0 and 1, corresponding to none or all, respectively, of the entire sample of air that crossed the urban boundary layer.
The D u time series with allocation to three urban areas is shown in Fig. S3.The comparison of urban influence in samples of various sample volume, V , requires normalisation to V , a relative dose D ru i (Eq.6, with n being total number of samples collected).Values of D ru i may exceed 1.
The accuracy of the urban fractional dose, D u i , is limited by the meteorological input data (here 0.125 • × 0.125 • resolution, hourly) and boundary layer depth calculation.In the FLEXPART model, the latter is done according to Vogelezang and Holtslag (1996).

Results and discussion
The NPAH levels are distinctly lower at the marine than at the continental site, 11 3-4-ring NPAH = 22.5 and 58.5 pg m −3 , respectively (Table 1).The NPAHs showing the highest concentrations were 2NFLT and 3NPHE at the marine site (Fig. 1b) and 9NANT and 2NFLT at the continental site (Fig. 1d, Table 2).The substance patterns (composition of NPAH mixture) at both sites are similar, however (R 2 = 0.76, P > 0.99, t test).At the marine site, advection was northerly, with air masses originating (time horizon of 10 days) in eastern and central Europe and, towards the end of the campaign, in the western Mediterranean.The site was placed into the south-eastern outflow of Europe.NO x (0.2-0.6 ppbv), EC (0.2-0.8 µg m −3 ), and PM 10 (18.3-39.3µg m −3 ) reflect background conditions.Air mass history analysis suggests that the somewhat elevated concentration in the first sample collected at the marine site (Fig. 1a) is related to long-range transport influenced by passage over the urban areas of Izmir and Istanbul (urban fractional dose D u = 5.0 %, in contrast to the mean of 1.6 %; Fig. S3).Overall, urban fractional dose in the range < 0.002-5.4% was received at the marine site.Across all samples at the marine (a) (b) 2.7.12 3.7.124.7.12 5.7.12 6.7.12 7.7.128.7.12 9.7.12 10.7.2.7.12 3.7.124.7.12 5.7.12 6.7.12 7.7.128.7.12 9.7.12 10.7.1211.site, D u is found to be significantly correlated with the pollutant sum concentrations 6 4-ring PAH and 11 3-4-ring NPAH (R = 0.61 and 0.69, respectively; both P > 0.99).From the marine site data set, subsets of each two samples are formed, representing the minimum, i.e. almost no influence from industrialised area 48 h prior to arrival (henceforth called "marine background", urban fractional dose D u = 0.4 %), and maximum observed influence (henceforth called "background with urban influence", D u = 3.1 %; Table S2, Fig. S3).The results for these subsets are listed in Tables 1-3.Such classification was not deemed meaningful for the samples collected at the continental site, as the relevant source distribution in central Europe was too homogeneous during this episode.Advection was mostly from the northwest and partly from easterly directions, with air mass origin (time horizon of 10 days) mostly in central Europe and, to a Table 2. Total (g + p) time-weighted concentrations, c tot (pg m −3 ), particulate mass fraction, θ = c p /c tot , and mass median diameter (MMD, µm) of 2-4-ring NPAHs and 4-ring PAHs at the marine (as "mean (background mean/urban influence mean)", n = 8(2 a /2 b )), and continental (as "mean (day mean/night mean)", n = 22(11/11)) sites, together with temperature and supporting aerosol parameters (PM 10 and carbonaceous mass fractions).LOQ is the limit of quantification and nd indicates no data.a Two filter and four PUF samples, i.e. nos.9, 10, 19, and 22 in Fig. S3.b One filter and one PUF sample, i.e. nos. 1 and 2 in Fig. S3.c Co-eluted with 3NFLT, assuming c 3NFLT = 0. lesser extent, in eastern Europe and the western Balkans.The NO 2 (1.2-2.6 ppbv), total carbon (3-6 µg m −3 ), and PM 10 (10.7-46.3µg m −3 ) levels during the campaign reflect continental background conditions.The 4-ring PAH concentrations in samples from the continental site and in background air with urban influence collected at the marine site (urban areas 300-500 km away) are similar (Table 2).Also, the substance patterns are more similar than when relating all samples at the marine site, i.e.R 2 = 0.88 (P > 0.999, t test) instead of R 2 = 0.76.The investigation of the diffusive air-surface exchange processes during the measurements presented here showed that 4ring PAHs were in fact influenced by secondary emissions, namely throughout day and night from the soil at the continental site (on average 16.3 and 9.3 pg m −2 h −1 for FLT and PYR, respectively; Degrendele et al., 2016) or occasionally from surface seawater at the marine site (during at least one daytime interval out of in total three of this data subset; Lammel et al., 2016).In the data set from the continental site, we study day-night effects (subsets listed in Tables 1-3): PAH concentrations (c tot ) were ≈ 60 % higher during the day than during the night, while c tot of NPAH were by average ≈ 5 % lower during the day (Table 2).NPAHs are subject to photolysis, while PAHs are not.At the site, the PAH concentrations were driven by re-volatilisation from soil, de-termined by temperature variation (Degrendele et al., 2016).For NPAHs (partly primary emitted) this indicates that the higher emissions during the day (due to re-volatilisation and road traffic) were compensated by shorter lifetime.NPAH lifetimes may be limited by heterogeneous photolysis, but available kinetic data are scarce and limited to a few aerosol types (Fan et al., 1996;Feilberg andNielsen, 2000, 2001;García-Berríos et al., 2017).Also, different NPAH / PAH ratios (the potential NPAH yields), which were 5.6 and 8.9 % at the marine and continental sites, respectively, reflect the combination of emission sources and photochemical sinks.The NPAH / PAH ratios at the two sites were influenced by similar substance patterns upon emission, similar irradiation (summer, no or almost no clouds), and similar deposition velocities (θ in the range of 0.05-0.20 for 11 3-4-ring NPAH and 6 4-ring PAH, no precipitation), but different re-volatilisation fluxes and different characteristic transport times elapsed.The distance to major urban source areas was 300 to > 1000 km at the marine and 100-500 km at the continental site.The NPAH / PAH ratios being lower at the more distant receptor site, the marine site, may suggest that photochemical degradation of NPAHs along transport was on average faster than degradation of the precursors.
The NPAH levels observed in marine background air are the lowest ever reported.Remarkably, the concentrations are much lower, by more than 1 order of magnitude, than 1 decade before at the same site during the same season (Tsapakis and Stephanou, 2007).The concentrations observed now are a factor of 4-10 lower than in a forest site in Amazonia 2 decades before (which might have been influenced by biomass burning emissions), a factor of 3 lower (for 2NPYR) than observed at an extremely remote site in the Himalayas 2 decades before (Ciccioli et al., 1996), and comparable to a high-altitude site in the Alps (with the exception of 2NPYR, which was observed 1 order of magnitude higher there in winter; Albinet et al., 2008a; Table 3).The NPAH levels observed at the marine site with influence of pollution and at the continental site are comparable but are also at the lower end of the range spanned by previous observations at rural and remote sites (Table 3).

Gas-particle partitioning
The time-weighted mean NPAH phase distributions ( 11 3-4-ring NPAH) differ, corresponding to θ = 0.05 and 0.17 at the marine and continental sites, respectively -despite similar temperatures (Table 1).In contrast, and despite similar temperature ranges, the 4-ring PAHs' ( 6 4-ring PAH) particulate mass fraction was higher at the marine than at the continental site (θ = 0.42 and 0.20, respectively).Both 4ring PAHs and 3-4-ring NPAHs were more associated with PM in polluted air than in clean air.This trend is weak for PAHs with θ = 0.02 for 6 4-ring PAH in marine background air and 0.07 in background air with urban influence (and θ = 0.09 and 0.20 for CHR; Table 2) but is obviously strong for NPAHs, namely θ = 0.19 for 2NPYR in marine background air and 0.69 in background air with urban influence, ≈ 0.93 in polluted continental air, and θ = 0.01 for 11 3-4-ring NPAH in marine background air but 0.22 in background air with urban influence (Table 2).The urbaninfluenced air at the marine site is also reflected in a much higher organic carbon (OC) (a factor of 3 higher than the allcampaign mean) and elevated EC (less prominent, ≈ 50 % above mean).This confirms the understanding that gasparticle partitioning of both PAHs (Lohmann and Lammel, 2004;Shahpoury et al., 2016) and NPAHs (Tomaz et al., 2016) is mostly determined by absorption in particulate OM (POM) and adsorption to soot.When comparing polluted air at the continental site and background air with urban influence at the marine site, a strong shift of 6 4-ring PAH towards the particulate phase, θ ≈ 0.21 vs. 0.07, respectively, is found; for 11 3-4-ring NPAH, θ is similar, i.e. ≈ 0.16 vs. 0.22, respectively.This phase partitioning trend of the 4-ring PAHs could be explained by sorption to EC, which is a factor of ≈ 2 higher, but not by OC (only ≈ 20 % higher).In conclusion, these observations consistently indicate that sorption to soot is less significant for gas-particle partitioning of NPAHs than for PAHs.
While NPAHs were significantly phase-shifted (θ = 0.24 during daytime but θ = 0.58 during night time), this was not the case for 4-ring PAHs (θ = 0.18 during daytime and θ = 0.23 during night time).This is in line with the perception that the temperature sensitivity of phase change is stronger for the substance class with stronger molecular interactions in the condensed-phase NPAHs.For example, the enthalpies of phase change between air and OC of FLT and NFLT are −98 and −75 kJ mol −1 , respectively (OC represented by DMSO;ACD, 2015).
Good agreement is found for the prediction of NPAH partitioning using the multi-phase (three-phase) ppLFER with most values predicted within 1 order of magnitude of the observed values (Fig. 2; quantification of deviations in Sect.S2.3.1).While the sensitivity of assumptions regarding PM phase composition made in the model does not contribute significantly to the deviations ( 1, log K p units), a significant part can be attributed to the usage of estimated solute-specific Abraham solvation parameters (taken from ACD, 2015), due to the lack of experimentally based descriptors.For example, for an urban site (Tomaz et al., 2016) it was found that experimentally based descriptors used for 9NPAH lead to better predictions than the estimated descriptors; i.e. root-mean-square errors differed by 0.43 log units.The agreement of the ppLFER prediction is better than assuming absorption (into OM) to be the only relevant process (K oa model; see Sect.S2.3.2,Fig. S5).The same was found when studying gas-particle partitioning of NPAHs in urban air (Tomaz et al., 2016).This supports the perception that gas-particle partitioning of NPAHs is governed by various molecular interactions with OM, with its polarity being well represented by DMSO, better than by octanol.Earlier, it had been found for eight 3-4-ring NPAHs at urban and rural sites (Li et al., 2016) that the dual model, assuming adsorption (to soot) and absorption (into OM), predicts better than single adsorption (to the total aerosol surface, i.e.Junge-Pankow) or single absorption (K oa ) models do.
The interactions with the aerosol matrix of 9NPHE (continental site) and 5NACE, 2NFLN, 2NFLT, and 1NPYR (marine site) are less well represented than other NPAHs by the model, as suggested by the low slopes of their log K p experimental / log K p predicted relationships.The reason is unknown.Moreover, sampling or sample handling artefacts cannot be excluded, even when the same temperature range, sampler, and sampling protocols are applied across sites with both satisfactory and deficient agreement between predicted and observed K p .Further conclusions are not supported by the limited amount of data and uncertainties on both the model (estimated ppLFER parameters) and experimental (concentrations close to LOQ) sides.

Mass size distribution
The NPAH mass size distribution had its maximum in the < 0.49 µm size range at both sites.The 4-ring PAHs mass size distribution had two maxima, < 0.49 µm and between 0.95 and 1.5 µ m, at the marine site but only one, at < 0.49 µm, Table 3.Comparison of total (g + p) concentrations in air, c tot (pg m −3 ), with other measurements at remote and rural sites.a Co-eluted with 3NFLT, assuming c 3NFLT = 0. b Samples 9, 10, 19, and 22 in Fig. S3.c Particulate-phase concentration only.d Val de Maurienne sites (Albinet et al., 2008a).e Plan de l'Aiguille site (Albinet et al., 2008a).
at the continental site (Table 1).This is probably related to the presence of aged aerosol at the marine site vs. a larger contribution of fresh aerosols at the continental site.This is, furthermore, supported by the analysis of air mass origins that shows significant influence of urban areas for only few samples at the marine and for all samples at the continental site (Sect.S2).Sums of NPAHs' and PAHs' mass size distributions are found unimodal with the maximum in particles < 0.49 µm except PAHs at the marine site, which shows a second maximum between 1.5 and 3.0 µm (Fig. 3).At the marine site, 50 and 69 % of 1NPYR and 2NFLT, respectively, was found associated with particles < 0.45µm; 68 and 86 % with particles < 0.95 µm; and even more, 83 and 100 % with particles < 0.45 µm at the continental site.6 4-ring PAH mass size distributions are shifted to larger particles in background air with urban influence as compared to marine background air (both collected at the marine site), i.e.MMD = 0.19 and 0.28, respectively.However, such a trend is not apparent for NPAHs (Table 2).The size shift of PAHs does not correspond to the PM 10 mass size distribution: the MMD of PM 10 for all samples collected at the marine site was 0.58 µm, while it was 1.13 and 0.62 µm in the marine background air and background air with urban influence data subsets, respectively.The PM 10 as well as the OC mass size distributions were bimodal with maxima corresponding to < 0.49 and 3.0-7.2µm particles (MMDs listed in Table 2), while the EC mass size distribution was unimodal, with the maximum concentration in the finest fraction.At the continental site, the 11 3-4-ring NPAH mass size distribu-tion was bimodal with maxima corresponding to < 0.49 and 7.2-10 µm particles, while the 6 4-ring PAH mass size distribution was unimodal, with the maximum concentration in the finest fraction (for all samples as well as for day and night data subsets; Table 1).
The formation of a second maximum, at larger particles than emitted, reflects the redistribution of semi-volatile organics in an aged aerosol and, hence, is expected at receptor sites such as the marine site.This was also observed in polluted air at rural and suburban sites but not at traffic sites or in winter at a rural site, when primary emissions dominated (unimodal; Albinet et al., 2008b;Ringuet et al., 2012b).

Substance patterns and NPAH formation during long-range atmospheric transport
Among the targeted NPAHs and apart from NNAPs, which were the most concentrated, 2NFLT and 3NPHE prevailed at the marine site (accounting together for ≈ 60 % of the NPAH mass, excluding the NNAPs), while at the continental site 9NANT and 2NFLT prevailed (accounting for ≈ 65 % together) (Fig. 1, summarised in Fig. S4).The analytical method did not separate the isomers 2NFLT and 3NFLT, but at receptor sites, far from diesel emissions, it appears justified to assume c 2NFLT c 3NFLT (Finlayson-Pitts and Pitts, 2000;Zimmermann et al., 2012).The ratio 1NPYR / 2NPYR is higher, ≈ 1, at the continental site than at the marine site (≈ 0.25), which reflects the significance of primary sources for polluted air (Atkinson and Arey, 1994;Finlayson-Pitts and Pitts, 2000;Zimmermann et al., 2012).This ratio was found similarly high or even higher at urban log K p experimental (m 3 g -1 ) sites (Ringuet et al., 2012c;Tomaz et al., 2016).Similarly, the ratio 2NFLT / 1NPYR, the concentration of a secondarily formed over a primary emitted NPAH, has been used as indicator for fresh emissions (if < 5) vs. photochemically aged air mass (Keyte et al., 2013).These values were 5 in 21 out of 22 and 7 out of 8 samples at the marine and continental sites, respectively.The only sample collected at the continental site with elevated primary NPAH (2NFLT / 1NPYR = 4.3) was possibly influenced by emissions from Budapest, which was passed by the advected air within the last hours before arrival.The only sample collected at the marine site with elevated primary NPAH (2NFLT / 1NPYR = 5.9) was indeed directly influenced by emissions into the boundary layer above the Izmir and Istanbul metropolitan areas (urban fractional dose D u = 5.0% for samples 1 and 2 in Fig. S3).
In conclusion, these results from receptor/background sites confirm the existing knowledge about primary emitted and secondarily formed NPAHs.
The ratio of two secondarily formed NPAHs, 2NFLT / 2NPYR, indicative for daytime vs. night-time formation paths (Atkinson and Arey, 1994;Ciccioli et al., 1996), is found ≈ 2 at the marine and ≈ 8 at the continental site (normalised to the precursor ratio, i.e. 2NFLT / 2NPYR / (FLT / PYR); Table 4).Such low values point to daytime (OH-initiated) formation, while night-time (NO 3 -initiated) formation was negligible, practically excluded, at the marine site.This is in line with the perception that NO 3 must have been very low in this remote environment.(NO x levels at the marine site were in the range 0.2-0.6 ppbv.)A similar conclusion has been previously drawn in a semi-rural environment (Feilberg et al., 2001).
For 2NFLT and 2NPYR (secondary sources only) and for 1NPYR, which has mostly primary sources (Finlayson-Pitts and Pitts, 2000;Ringuet et al., 2012a;Jariyasopit et al., 2014a, b), we infer the potential yields (Table 4).Here, yield is defined as c NPAH /c PAH (total concentrations).This yield is called "potential" as it reflects an upper estimate, as other PAH photochemical sinks, such as formation of oxy-PAHs, are neglected.The yield of 2NFLT in polluted air exceeds the one in background air only slightly, while the yield of 2NPYR in polluted air exceeds the one in background air much more (a factor of 3 higher).
As expected, the highest potential yield of 1NPYR is found in polluted air (both sites), reflecting the dominance of primary emissions of 1NPYR.Similarly, higher yields of secondary NPAHs are found for marine background air compared to background air with urban influence (marine site), reflecting the longer reaction times elapsed since PAH emission.The yield for 2NFLT, c 2NFLT /c FLT , which is ≈ 2-4 % at both sites, ranges higher than the one for 2NPYR, c 2NPYR /c PYR , which is found ≈ 0.5-2 %.Note that because of the co-elution of 2NFLT and 3NFLT, and neglect of 3NFLT, the so-derived values of c 2NFLT /c FLT actually represent upper estimates.Apart from sites which were immediately influenced by PAH sources (road traffic, power plant, biomass burning), only very few studies reported NPAH together with precursor data in both phases of ambient air.c 2NPYR /c PYR = 1.0 %, similar to our finding at remote sites, but a very high c 2NFLT /c FLT of 12.9 % was reported from a suburban site in France in summer during daytime (corresponding values for night time were 2.0 and 9.4 %, respectively; Ringuet et al., 2012c).2NFLT was not separated from 3NFLT (similar to our data set).A suburban site will be influenced by direct 3NFLT emissions, such that c 2NFLT /c FLT is an upper estimate.Many lower ratios, c 2NFLT /c FLT = 0.20 % and c 2NPYR /c PYR = 0.08 %, were reported as the median values for 90 sites of various categories, rural and urban, in northern China in summer (Lin et al., 2015).These yields are somewhat higher for the subset of the rural sites.The potential yields found at the marine site in our study are close to the yields for OH-initiated photochemistry observed in laboratory experiments under high NO x conditions, i.e. 3 % for c 2NFLT /c FLT and 0.5 % for c 2NPYR /c PYR (Atkinson and Arey, 1994).

Conclusions
For the first time, pollution contained in individual background air samples was quantified by means of a fractional dose.The fractional dose indicated how much the collected volume of air had been exposed to an urban boundary layer within a given time horizon.This is found suitable to discriminate among samples and discuss results, clearly beyond qualitative reasoning on back trajectories alone.The concept could be applied to any type of georeferenced origin and might be useful to track the influence of land use of various kinds or ship and aircraft routes.
Our measurements confirmed the occurrence of mutagenic NPAHs, earlier reported from polluted atmospheric environments of America, Europe, and Asia, for the European background atmosphere and the outflow of Europe as well.These substances are obviously subject to intercontinental transport and might indeed be distributed ubiquitously.The mass size distribution is determined by the particle size upon emission (primary NPAHs), and condensation and redistribution in the aerosol along transport, hence, do not include the short-lived coarse mass fraction.This indicates a high long-range transport potential.However, the observation of 3.8 and 0.92 pg m −3 of 2NFLT and 2NPYR, respectively, measured at the south-eastern outflow of Europe (the lowest ever reported concentrations; Table 3), may indicate that their abundance in the remote global environment could be less than anticipated.Earlier, this was based on a single measurement of 2NPYR, 3 pg m −3 , at an extremely remote site in central Asia 2 decades before (Ciccioli et al., 1996).Moreover, this air, classified as marine background, was not completely clean but had been exposed to a non-zero fractional urban pollution dose (0.4 % of the total, time horizon of 2 days).More measurements at remote sites should verify NPAH levels globally.PAHs have been abated significantly in Europe during the last decades (EEA, 2014), which should also be reflected in long-term trends of their derivatives.However, a temporal trend for the Aegean or the south-eastern outflow of Europe in summer cannot be inferred based on current and earlier (2002; Tsapakis and Stephanou, 2007) campaign data.NPAHs should be included in monitoring programmes to better assess the exposure of human health hazards of atmospheric pollution, even in remote areas.
Our understanding of NPAH formation in ambient air is still rudimentary.Both the kinetics of NPAH formation and photolysis remain to be quantitatively studied under conditions of the background atmosphere, i.e. low NO x , and on various aerosol matrices, including sea salt.More studies into NPAH atmospheric fate, both field observations and kinetic data, are needed in order to assess and quantify spatial and temporal trends, the long-range transport potential, and persistence.
Data availability.Data used in this study are available upon request to the corresponding author.
The Supplement related to this article is available online at doi:10.5194/acp-17-6257-2017-supplement.

Figure 1 .
Figure 1.Time series of absolute (a, c; pg m −3 ) and relative (b, d) total (gas + particulate) NPAH concentrations at the (a, b) marine (24 h means shown) and(c, d) continental site (day-night means).At the marine site, the gas phase (PUF) was sampled for individual daytimes and nighttimes while the particulate phase (filter) was sampled over 24 or 48 h (48 h during the period of 7-12 July 2012).

Figure 2 .
Figure 2. Predicted versus experimental log K p (m 3 air g −1 PM) for NPAHs using the multi-phase ppLFER model at the (a) marine and (b) continental site.

Figure 3 .
Figure 3. Time-weighted mean 6 4-ring PAH and 11 3-4-ring NPAH mass size distributions (pg m −3 ) at the marine and continental sites.The error bars show the standard deviation from the campaign mean.

Table 1 .
Overview of time-weighted mean concentrations in the particulate and gas phases and ambient temperature.Data subsets (B is background, P is polluted, D is day mean, N is night mean) and mass size distribution (< 0.