Snow scavenging and phase partitioning of nitrated and oxygenated aromatic hydrocarbons in polluted and remote environments in central Europe and the European Arctic

Nitrated and oxygenated polycyclic aromatic hydrocarbons (N/OPAHs) are emitted in combustion processes and formed in polluted air. Their environmental cycling through wet deposition has hardly been studied. Fresh snow samples at urban and rural sites in central Europe, as well as surface snow from a remote site in Svalbard were analysed for 17 NPAHs, 9 OPAHs, and 11 nitrated mono-aromatic hydrocarbons (NMAHs), of which most N/OPAHs as well 5 as nitrocatechols, nitrosalicylic acids, and 4-nitroguaiacol are studied for the first time in precipitation. In order to better understand the scavenging mechanisms, the particulate mass fractions (Ө) at 273K were predicted using a multiphase gas-particle partitioning model based on polyparameter linear free energy relationships. ∑NPAH concentrations were 1.2-17.6 and 8.8-19.1 ng L at urban and rural sites, whereas ∑OPAHs were 269.5 – 1114.1 and 478.7 2384.4 ng L at these sites, respectively. Acenaphthenequinone and 9,10-anthraquinone were predominant in 10 snow aqueous and particulate phases, respectively. NPAHs were only found in the particulate phase with 9nitroanthracene being predominant followed by 2-nitrofluoranthene. Among NMAHs, 4-nitrophenol showed the highest abundance in both phases. The levels found for nitrophenols were in the same range or lower than those reported in the 1980s and 1990s. The lowest levels of ∑N/OPAHs and ∑NMAHs were found at the remote site (9.2 and 390.5 ng L, respectively). N/OPAHs preferentially partitioned in snow particulate phase in accordance with 15 predicted Ө, whereas NMAHs were predominant in the aqueous phase, regardless of Ө. It is concluded that the phase distribution of non-polar N/OPAHs in snow is determined by their gas-particle partitioning prior to snow scavenging, whereas that for polar particulate phase substances, i.e. NMAHs, is determined by an interplay between gas-particle partitioning in the aerosol and dissolution during inor below-cloud scavenging.

NMAHs are ecotoxic (Pflieger and Kroflič, 2017) while little is known about human toxicity (Huang et al., 1995;Harrison et al., 2005;Kovacic and Somanathan, 2014). Due to nitrophenol phytotoxic potential, the research on them in precipitation was fostered during the late 1980s and early 1990s in relation with research on forest decline in central Europe (Rippen et al., 1987;Leuenberger et al., 1988;Herterich and Herrmann, 1990). Last but not least, NMAHs 15 significantly contribute to the light absorptive properties of PM organic carbon (Mohr et al., 2013;Hinrichs et al., 2016;Bluvshtein et al., 2017;Teich et al., 2017), and can influence climate by altering Earth's albedo.
Semi-volatile organic compounds (SVOCs) in the atmosphere are subject to removal by dry particle deposition and wet scavenging. The latter consists of two processesi.e. particle scavenging and gas scavenging. The particle scavenging is relevant for SVOCs that show higher affinity towards particulate phase (Bidleman, 1988;Shahpoury et 20 al., 2015). This is affected by SVOC gas-particle partitioning (GPP) in the aerosol, a process controlled mainly by the substance molecular structure, PM chemical composition, and ambient temperature . The magnitude of SVOC sorption to PM is defined by its particulate mass fraction, Ө Eq. (1): where cp and cg are concentrations of SVOC (ng m -3 ) in the particulate and gas phases, respectively. The gas scavenging is relevant for substances which demonstrate minimum interaction with PM and therefore remain to 25 various extents in the gas phase. Although gas scavenging is affected by GPP, the actual removal from the atmosphere is due to substance dissolution in cloud or rain droplets or sorption to snowflakes or other ice hydrometeors (Hoff et al., 1995;Bartels-Rausch et al., 2014). Following wet scavenging, SVOCs in the gas and particulate phases in the atmosphere accumulate in precipitation aqueous and particulate phases, respectively. The fraction of SVOCs removed by particle scavenging is given by Өw Eq. (2): 30 where cpp and cpa are analyte concentrations (ng L -1 ) in precipitation particulate and aqueous phases, respectively. It has been shown that the magnitude of particle scavenging is generally higher than gas scavenging for hydrophobic SVOCs such as PAHs (Atlas and Giam, 1988;Bidleman, 1988;Shahpoury et al., 2015). In principle, for such substances one could apply Ө as an indicator for predicting the SVOC wet particle scavengingi.e. the higher the Ө at a given temperature, the more efficient is the scavenging, and the magnitude of ӨW is expected to be close to Ө.
However, it is anticipated that for hydrophilic SVOCs, which may demonstrate low or high Ө, water solubility plays an additional role in the substance wet scavenging pathways and the comparability of ӨW with Ө. For instance, a NMAH with high Ө in the atmosphere may demonstrate low ӨW due to substance dissolution in hydrometeors following particle scavenging. Conversely, a NMAH with low Ө is expected to show low ӨW, as it undergoes gas 5 scavenging process.
The aims of the present study were (1) to investigate the presence of N/OPAHs and NMAHs in the aqueous and particulate phases of fresh snow, (2) estimate the substance particulate mass fractions in the atmosphere using a 20 multiphase GPP model, based on poly-parameter linear free energy relationships (ppLFER), and (3) determine the substance fractions removed by wet particle scavenging, and explore the effect of GPP vs. water solubility on target compound wet scavenging. For calculating Ө, our method took into account the interaction of atmospheric SVOCs with PM liquid organic and polymeric phases, as well as the interaction with PM black carbon and salts, while disregarding the partitioning into PM aqueous phase, particle-liquid interactions, and liquid-liquid phase separation 25 within PM (Sect. 2.6).

Sampling
Snow samples were collected between winter 2015 and 2017 from three locations in Germany, i.e. Mainz (MZ15 and MZ17; urban-residential, ≈200000 inhabitants), Winterberg (WB) and Altenberg (AB; rural, >10 km from small 30 towns), two locations in Inn Valley, Austria, i.e. Götzens (GS; urban-residential of a mid-sized city, Innsbruck, ≈140000 inhabitants) and Kolsassberg (KB; rural, 10-20 km from city and towns), two locations in the Czech Republic, i.e. Ostrava (OS; urban, conurbation with ≈700000 inhabitants) and Pustá Polom (PP1 and PP2; rural, 20 km upwind from Ostrava), and one location in the Arctic, Tempelfjorden, Svalbard (TF, remote, 50-80 km from small towns). The sample site details are shown in Table 1. Fresh snow samples (all sites, except TF) were collected by placing several polypropylene trays (0.25 m 2 each) on the ground prior to snowfall. To this end, the snow forecast for a number of pre-selected sites was monitored on daily basis. Both collection trays and bottles were pre-cleaned prior to sample collection in the lab using detergent, tap water, deionized water, and high-purity ethanol. The snow was 5 transferred with compaction in amber 2-L bottles and kept frozen at -18°C until analysis. Dry and light surface snow (0-5 cm deep, somewhat harder at the surface), which had fallen 3-2 days before, was collected at TF and stored in pre-cleaned amber bottles. Following this sample collection and storage procedure, we rely on exclusion of significant phase change during storage and prior to sample processing.

Sample processing 10
Our sample processing was performed in such way to minimize the analyte phase change prior to analysis, namely the samples were thawed at room temperature in the lab and, immediately after thawing, when the samples were near freezing point, the meltwater was passed through a pre-assembled filtration-extraction setup ( Fig. S1), which allowed simultaneous separation of meltwater particulate phase and extraction of aqueous phase. This made it possible to minimize the time that particles were in contact with the aqueous phase of meltwater. The setup consisted of a pre-15 assembled sterile analytical filter funnel (250 mL, graduated, Nalgene, Thermo Scientific, Waltham, MA, USA), connected to a solid-phase extraction disk (Bakerbond Speedisk, J.T. Baker, the Netherlands) using a Teflon adaptor designed in-house (Fig. S1). This was assembled on a J.T. Baker extraction station connected to a vacuum pump. A steady sample flow (10 mL min -1 ) was established between the filter funnel and Speedisk throughout sample processing, by occasionally applying vacuum, where needed. The volume of meltwater for each sample was 20 determined during filtration using the graduated filter funnel (Fig. S1), and used for calculating the final analyte concentrations in aqueous and particulate phases. The pH of meltwater samples was 4.5-5.
Briefly, each filter paper was placed in a glass centrifuge tube (Duran, Schott, Mainz, Germany) and added with 7 mL of dichloromethane (DCM). The centrifuge tubes were capped with screw caps containing PTFE lining. Each sample was vortexed for 1.5 min, passed through a glass funnel plugged with a small amount of deactivated glass wool (in order to remove residual sample matrix), and concentrated to 0.5 mL using a Turbovap II (Biotage, Uppsala, Sweden).
The extracts were later loaded on a pre-conditioned 500 mg SiO2 cartridge (Macherey-Nagel, Weilmünster, Germany), and eluted with 9 mL of 65:35 n-hexane-DCM. The aqueous phase samples enriched on C18 Speedisks were eluted with 40 mL of 1:1 n-hexane-DCM. All extracts from particulate and aqueous phases were concentrated to 0.5 mL and 5 the solvent was exchanged to ethyl acetate (3×5 mL). The sample volumes were further adjusted to 0.3 mL and transferred to 2-mL vials containing pre-baked 0.4-mL glass inserts for further analysis. All other glassware used for sample analysis were washed with lab-grade detergent and deionized water, and baked at 300°C for 12hrs. All solvents used for N/OPAH analysis were Suprasolv grade (Merck, Darmstadt, Germany).
The samples were analysed using a Trace 1310 gas chromatograph (GC, Thermo Scientific, Waltham, MA, USA) 10 coupled to a TSQ8000 Evo triple-quadrupole mass selective detector (MS/MS, Thermo Scientific) in negative chemical ionization and selected ion monitoring (SIM) modes. The analyte separation was achieved on a J&W DB-5ms column (30 m + 10 m integrated guard, 0.25 mm ID, 0.25 µm film thickness, Santa Clara, CA, USA) with helium (99.9999%; Westfalen AG, Münster, Germany) as carrier gas at 1 mL min -1 flow rate. The GC operating conditions were as follows: the GC oven was held at 60°C for 2 min, then ramped to 180°C at 15°C min -1 , followed by a 5°C 15 min -1 ramp to 280°C and final hold time of 15 min. The injection port temperature was set to 250°C and operated in pulsed splitless mode (30 psi pulsed pressure for 1.5 min, and splitless time of 1.8 min). MS transfer line and ion source temperature were set to 290 and 230°C, respectively. Methane (>99.9995, Messer, Bad Soden, Germany) was used as ionization gas with 1.5 mL min -1 flow rate. Emission current and electron energy were set to 100 µA and -70 eV, respectively. The samples were analysed for N/OPAHs listed in Table 2. Each target analyte was identified using 20 its retention time and quantification ion ( Table 2). The analyte quantification was done using the internal calibration method with 11-point calibration curves ranging from 0.25-1000 pg µL -1 .

NMAH extraction and chemical analysis
0.22-micron cellulose acetate filter in the 250 mL analytical filter funnel and divinylbenzene hydrophilic Speedisk were used for NMAH extraction. The detailed analytical method is described in a companion paper (Kitanovski and 25 Naumoska,in preparation). Briefly, the pre-conditioned Speedisk was spiked with 100 ng of 4-nitrophenol-d4, the sample (250 mL) was acidified with 2 mL of formic acid, and passed through the disk. The elution of NMAHs from the disks was done using a mixture of acetonitrile-methanol containing 3.4 µM ethylenediaminetetraacetic acid (EDTA). The presence of EDTA in the elution solvent was necessary for complete recovery of the NMAHs from the polymeric disks. The SPE extracts were further concentrated to near dryness using TurboVap II and later dissolved in 30 1 mL mixture of methanol and 7.15 mM ammonium formate buffer pH 3 (3:7, v/v) containing 400 µM EDTA.
The PM retained on the cellulose acetate filters was spiked with the same quantity of 4-nitrophenol-d4 and extracted using a previously published procedure (Kitanovski et al., 2012). Briefly, the particles were extracted using methanol containing 3.4 µM EDTA with agitation (3×5 min) in an ultrasonic bath. The final extracts were concentrated to near dryness, and dissolved in 1 mL mixture of methanol and 7.15 mM ammonium formate buffer pH 3 (3/7, v/v), 35 containing 400 µM EDTA. All samples were analyzed using a 1200 Series liquid chromatograph (LC; Agilent Technologies, Santa Clara, CA, USA) coupled to a 6130 single-quadrupole MS (Agilent Technologies) with an electrospray ionization (ESI) source. Separation was achieved on an Atlantis T3 column (150 × 2.1 mm ID, 3 µm; Waters, Milford, MA, USA), thermostated at 30°C during sample analysis. The NMAH elution was done using 30:15:55 methanol-tetrahydrofuran-aqueous ammonium formate buffer (5 mM, pH 3) mobile phase in isocratic mode. 5 The deprotonated NMAHs [M-H] ͞ listed in Table 2 were detected in negative ion ESI and SIM modes. The analyte quantification was done using the internal calibration method in concentration range 1-500 pg µL -1 .

Quality control
Field blanks were prepared during sample collection by exposing the pre-cleaned sample bottles with open cap to air for 5 min at the sites. The inner wall of the bottles was rinsed with 500 mL of deionized water in the lab and processed 10 as field blank along with the rest of samples. Limits of quantification (LOQ) for analytes were calculated based on instrument detection limits (IDL), which in turn are determined using 3 times the chromatogram baseline noise level.
LOQ values were determined as mean concentration of each analyte in blanks + 3 standard deviations. For this purpose, blank values that were <IDL were replaced with IDL values. Where analyte concentrations in samples exceeded the LOQ, mean blank concentrations were subtracted from those in the corresponding samples. 15

Estimation of particulate mass fractions
The Ө for target analytes were estimated using modelled GPP constants, KP (m 3 g -1 at 273.15 K), which were calculated using a multiphase ppLFER model . The model differentiates between various organic and inorganic phases of PM, and accounts for absorption into water soluble organic matter (WSOM) and organic polymers (OP), as well as adsorption onto black carbon, (NH4)2SO4 and NaCl, Eq. (4) where KBC, K (NH4)2SO4 and K NaCl are the target substance partitioning coefficients (mol m -2 surface/mol m -3 air) for black carbon/diesel soot, (NH4)2SO4 and NaCl (the last two represent the contributions of secondary inorganic aerosols and sea salt), respectively, aBC, a (NH4)2SO4 and a NaCl are the adsorbent specific surface areas (m 2 surface g -1 adsorbent), and fBC, f (NH4)2SO4 and f NaCl are their mass mixing ratios in PM (gadsorbent g -1 PM). For aBC, the geometric mean of 18.21 m 2 g -1 was calculated from the values reported for traffic, wood, coal, and diesel soot (Jonker and Koelmans, 2002), whereas, 25 a (NH4)2SO4 and a NaCl of 0.13 and 0.10 m 2 g -1 were taken from Goss et al., (2003). KDMSO (m 3 air m -3 DMSO) and KPU (m 3 air g -1 PU) are the substance partitioning (absorption) coefficients for dimethyl sulfoxide-air and polyurethane-air partitioning systems; ρDMSO is dimethyl sulfoxide density (g m -3 ); fWSOM and fOP, are mass mixing ratios of absorbing phases (gabsorbent g -1 PM), corresponding to fOM (the mixing ratio of total organic matter in PM) × 0.60 and fOM × 0.40, respectively. The correction factors of 0.60 and 0.40 were estimated based on the data from Rogge et al., (1993). These factors were previously verified through a sensitivity study . We assumed two scenarios for model calculations: fBC = 0.03 and fOM = 0.30, and fBC = 0.06 and fOM = 0.60. This resulted in fWSOM and fOP of 0.18 and 0.12, and 0.36 and 0.24 for the two scenarios, respectively. The contributions of inorganic salts and PM aqueous phase to the overall sorption process were neglected; we acknowledge that under high relative humidity, salts may be 5 present in aqueous phase, and subject to liquid-liquid phase separation with PM organic matter (You et al., 2014). The individual partitioning constants used in the multi-phase model were calculated using substance-specific Abraham descriptors listed in Table S2 and ppLFER models listed in Table S3 (Abraham et al., 2010;Kamprad and Goss, 2007;Roth et al., 2005;Goss et al., 2003).
One must note that the ppLFER model used here predicts KP of a substance in neutral form. In particulate phase, depending on pH of the PM aqueous phase and pKa of the target substance, NMAHs may partly or completely deprotonate, resulting in enhanced solubility of the substance in the aqueous phase (Ahrens et al., 2012). This implies 15 that the actual partitioning could be under-predicted for such substances. The effect is expected to be relevant for 5nitrosalicylic acid (5-NSA; see Table 2

Air mass history analysis
The HYSPLIT (Draxler and Rolph, 2003) model was used to identify the air masses' three-day histories related to the snowfall events. The meteorological data (1°x1° resolution) used were from the Global Data Assimilation System 25 (GDAS, NCEP, USA). Air mass changes were identified based on weather charts (Berliner Wetterkarte, 2015), except for sample site in Svalbard, TF, as the snow fell 2-3 days prior to sample collection. The results are presented in the Supplement, Text S1 and Fig. S2.
Potential source areas for such emissions are reflected in the air mass histories of all samples (see Text S1 and Fig.  10 S2). Overall, TF and MZ17 were the least and most polluted sites, with ∑ OPAH aqueous concentrations of 7.3 and 834 ng L -1 , respectively. NPAHs were not found in the snow aqueous phase. Our GPP model suggests that at nearzero temperatures the targeted NPAHs would be completely sorbed to the particulate phase in the atmosphere, with the exception 1-NNAP, 2-NNAP, 5-NACE, and 2-NFLU, which would partition between gas and particulate phases.
Regardless, relatively low water solubility of NPAHs, indicated by their octanol-water partitioning coefficients (log 15 KOW; Fig. S3), may limit their gas scavenging from the atmosphere and subsequent presence in the snow aqueous phase.
The NMAHs were less abundant in the snow particulate phase (Fig. 4B and 5B). 4-NP was found in all samples, with 10 concentrations up to 106.9 ng L -1 (MZ17), followed by 2-M-4-NP (≤39.2 ng L -1 ), 3-M-4-NP (21.8 ng L -1 ), and DNOC (≤11.8 ng L -1 ; Fig. 4B and 5B). Interestingly, 4-NP was the only NMAH found at the remote site TF. The rest of NMAHs were more sporadic in the particulate phase, with 3-NSA, 4-NG, and 5-NSA being found exclusively at the urban site MZ17 (Fig. 4B). Overall in the particulate phase, the remote site TF was the cleanest (∑NMAHs: 3.5 ng L -1 ) and the urban site MZ17 was the most polluted (∑NMAHs: 276.6 ng L -1 ) in our study. 15 The measured nitrophenol concentrations are in the same range as those reported previously for snow from central Europe and Antarctica (Alber et al., 1989;Vanni et al., 2001; Table S1), but lower (up to two orders of magnitude lower) than those reported in rainwater from central and northern Europe (Leuenberger et al., 1988;Herterich and Herrmann, 1990;Levsen et al., 1991;Bossi et al., 2002;Asman et al., 2005;Schummer et al., 2009; Table S1) and the USA (Ganranoo et al., 2010; Table S1). It must be noted that here we report the NMAH concentrations in both 20 aqueous and particulate phases, which was not reported in the previous studies (Table S1). When detected, the aqueous phase concentrations of 4-NG were comparable in urban and rural samples (~ 155-285 ng L -1 ; Fig. 4A and 5A), indicating polluted and aged air masses with biomass burning origin (Kitanovski et al., 2014;Kroflič et al., 2015;Yang et al., 2016). These values are one order of magnitude higher than that measured for the remote sample TF (Fig.   5A). In urban samples with the exception of OS, nitrocatechols (4-NC and isomeric methyl-nitrocatechols) were the 25 second most abundant NMAHs, following nitrophenols (Table S4A). This is in agreement with our current unpublished data and previous studies on winter PM samples (Kitanovski et al., 2012). In contrast, in rural samples KB and PP1, and the remote sample TF, nitrosalicylic acids are the second most abundant NMAH species (Table   S4A), which might indicate either higher stability or higher formation of these compounds during the transport of polluted air to the rural environments and the Arctic. 30

Modelled particulate mass fractions and actual fractions removed by particle scavenging
The target compound Ө was estimated following the method explained in Sect. 2.6. We examined three scenarios i.e., a lower, middle and upper scenario with regard to pollution: (1) with fBC and fOM of 0.03 and 0.30 and cPM of 25 µg m -fact, the 24-h mean cPM values for MZ17 and OS sampling events were 25 and 27 µg m -3 , respectively. No data are available from the other sites. As can be seen from Fig. 6A-C, Ө at 273 K is ~ 1 (complete sorption to PM) for 70% of the target compounds, independent of the above scenarios. The calculations showed that <20% of 1-NNAP, 2-NNAP, 1,4-O2NAP, 9-OFLN, 2,4-DNP, and DNOC, and between 20 and 80% of 5-NACE, 2-NFLN, O2ACE, 3-NSA, and 4-NG can be expected to be in the particulate phase at 273 K (Fig. 6). For these substances, Ө increases by up to 5 two fold when increasing fBC, fOM, or cPM. An interesting trend was seen for 2,4-DNP and 4-NP: the presence of two nitro groups on 2,4-DNP is expected to promote stronger H-bonding with PM, compared to 4-NP which has one nitro group (compare the Abraham descriptor B for the two compounds in Table S2). However, our model calculations suggest the opposite pattern, i.e. complete sorption of 4-NP to PM, but <5% sorption for 2,4-DNP. This is largely related to intramolecular H-bonding between the H-atom of the hydroxy group and the O-atom of the nitro group in 10 ortho position. This reduces the H-bonding ability of 2,4-DNP compared to 4-NP (compare the Abraham descriptor A in Table S2).
For substances which demonstrate complete sorption to PM, particle scavenging is expected to be the dominant source of wet deposition and, consequently, such substances will be enriched in precipitation particulate phase. Our observations depicted in Fig. 7A support this assumption, namely, N/OPAHs with high Ө (Fig. 6A-B) were largely 15 associated with precipitation particulate phase. The only exception was 9,10-O2ANT in MZ15 and MZ17 samples.
For substances that distribute between gas and particulate phases, both gas and particle scavenging are relevant; however, the substance water solubility is a factor that could enhance or limit the gas scavenging mechanism, regardless of the compound phase partitioning in the atmosphere. For instance, 1-and 2-NNAP are expected to be ≥90% in the gas phase under our experimental conditions; nonetheless, they were found in the precipitation particulate 20 phase. A similar effect was seen for 9-OFLN. This is explained by the substances' relatively low water solubility (see estimated KOW in Fig. S3) which limits their dissolution in hydrometeors. On the contrary, 1,4-O2NAP (Ө: 0.001-0.003) and O2ACE (Ө: 0.24-0.56) were more enriched in the aqueous phase, which is explained by their relatively low log KOW (1.95-2.13; Fig. S3). Overall, our findings support the implied assumption that phase partitioning in air is preserved in snow, provided that the substance solubility in water is not a limiting factor. 25 The observation for NMAHs, however, are considerably different (Fig. 7B). These substances were mainly found in the aqueous phase, including the seven analytes for which complete sorption to PM was predicted under various scenarios. Among all, 4-NP showed the highest fraction in the precipitation particulate phase and, as predicted, higher than 2,4-DNP; this was more pronounced in urban and rural samples OS, PP1 and PP2. The opposite pattern observed for NMAHs may suggest that besides GPP another process influenced the substance wet scavenging and consequently 30 their distribution in precipitation phases. We rule out post-sampling dissolution of NMAHs from particulate to aqueous phase for the reasons mentioned in Sect. 2.2. In addition, looking at the Czech samples, 3-M-4-NP and 2-M-4-NP, which have lower water solubility (log KOW: 2.27-2.43; Fig. S3) than 4-NP (log KOW: 1.68) and comparable pKa values (7. 15-7.33), demonstrated much higher partitioning in the aqueous phase than 4-NP did. If a post-sampling phase change had occurred, we would have observed an opposite pattern. We hypothesize that following both in-and below-35 cloud scavenging into sub-cooled droplets, particulate-phase NMAHs would dissolve into the bulk water or water layer surrounding ice hydrometeors (Hoff et al., 1995). This process is, at least partly, affected by the substance water solubility and, therefore, is anticipated to be more efficient for highly water soluble substances such as NMAHs (s = 0.3-10 g L -1 , log KOW = 0.64-2.54; Fig. S3). This hypothesis is consistent with the results of a previous study where positive correlation was found between in-cloud scavenging efficiencies of PM-bound polar organics and the substance polarity (Limbeck and Puxbaum, 2000). This process (which potentially results in predominance of NMAHs in the aqueous phase), however, must be distinguished from the gas scavenging which was proposed by Leuenberger 5 et al., (1985) as a dominant process for removal of volatile methylated phenols, where the predominance of these substances in the aqueous phase corresponded to their abundance in the gas phase.
For the urban site OS, the analyte concentrations in the gas and particulate phases of the air have been determined (sample collected over 24h preceding snowfall onset; Table S4B) in addition to concentrations in precipitation. The scavenging ratios WT calculated for the target N/OPAHs and NMAHs (see Eq. 3) were 10 3 -10 4 and 10 3 -10 5 , 10 respectively (Table S4B), which fall within the range suggested for removal of polyaromatic compounds through wet particle scavenging (Shahpoury et al., 2015). With the exception of 1,4-O2NAP and 9-OFLN, the range of calculated WT is consistent with that of modelled Ө at 273 K (Table S4B), meaning that the particle scavenging was the dominant removal mechanism. The difference in WT between 3-and 5-NSA (~1.4 times higher for 5-NSA) and between 9,10-O2ANT and OBAT (~1.1 times higher for OBAT) closely resembled that of estimated Ө at 273 K (Table S4B). 15 However, we found differences in WT between the NMAH subgroups, namely WT values were higher for nitrophenols (1.3×10 4 -1.6×10 5 ) and nitrosalicylic acids (5.7×10 4 -8.2×10 4 ) than nitrocatechols (1.1×10 3 -2.8×10 3 ; Table S4B), which cannot be explained by the substances' GPP alone (compare Ө in Fig. 6C). Although based on a single precipitation event, these results are indication of additional atmospheric processes which NMAHs could undergo and which may affect their wet scavenging. For instance, PM size-dependent mass distribution is a parameter which was 20 suggested to influence the snow scavenging efficiencies (Zhang et al., 2013)i.e. lower efficiency in the PM size range 0.01-1 µm than for coarse PM (≥ 1 µm). This parameter should to be taken into account in future studies, and more precipitation episodes need to be considered in order to draw a full picture.

Final remarks
The phase-separated concentrations of nitrated and oxygenated aromatic compounds were measured in snow from 25 several locations in central Europe and European Arctic. For the first time, we have reported the snow concentrations for several classes of nitrated and oxygenated aromatics, such as nitrocatechols, nitrosalicylic acids, nitrated and oxygenated PAHs, and 4-nitroguaiacol. Our results showed that a model-predicted particulate mass fraction, Ө, can be reliably applied in order to predict the scavenging mechanism (gas or particle scavenging) and efficiency of hydrophobic N/OPAHs and, hence, their distribution in precipitation phases i.e., aqueous and particulate. This 30 suggests that the atmospheric lifetime of N/OPAHs in relation to removal processes follows the same pattern as that of parent PAHs (Sharma and McBean, 2002). On the contrary, Ө is not a good measure for predicting phase distribution of hydrophilic NMAHs in precipitation. Our data suggests that the phase distribution of polar particulate phase substances, such as NMAHs in snow is determined by an interplay between GPP in the aerosol and dissolution in cloud or rain droplets, or the liquid water phase on the surface of ice hydrometeors during in-or below-cloud 35 scavenging. This behaviour is in line with what was found for in-cloud scavenging of other polar mono-and difunctional organics, such as aliphatic alcohols, and aliphatic and monoaromatic aldehydes and carboxylic acids (Limbeck and Puxbaum, 2000), namely the polarity of these substances is a significant parameter for their scavenging when solubility is > 1 g L -1 . The experimental data on GPP of NMAHs is scarce in the literature, which was also not addressed here. The contribution of gaseous polar NMAHs to the total scavenging, beyond the scope of this study, 5 should be addressed, similar to other polar aliphatic and aromatic organics (Limbeck and Puxbaum, 2000).

Data availability
The dataset used in this manuscript is included in the Supplement, and further information is available from the corresponding author (p.shahpoury@mpic.de).

Competing interests
The authors declare that they have no conflict of interest. A (100) P (78) Abbreviations, retention times (RT), and quantification ions (Q1) of target analytes; a analyte detection in aqueous (A) or particulate (P) phase; numbers in brackets indicate percentage of detection across the samples in each phase; n.d.: not detected