Biogenic, urban, and wildfire influences on the molecular composition of dissolved organic compounds in cloud water

Organic aerosol formation and transformation occurs within aqueous aerosol and cloud droplets, yet little is known about the composition of high molecular weight organic compounds in cloud water. Cloud water samples collected at Whiteface Mountain, New York during August-September 2014 were analyzed by ultrahigh-resolution mass spectrometry to investigate the molecular composition of 25 dissolved organic carbon, with a focus on sulfurand nitrogen-containing compounds. Organic molecular composition was evaluated in the context of cloud water inorganic ion concentrations, pH, and total organic carbon concentrations to gain insights into the sources and aqueous phase processes of the observed high molecular weight organic compounds. Cloud water acidity was positively correlated with the average oxygen:carbon ratio of the organic constituents, suggesting the possibility for aqueous acid30 catalyzed (prior to cloud droplet activation, or during/after cloud droplet evaporation) and/or radical (within cloud droplets) oxidation processes. Many tracer compounds recently identified in laboratory studies of bulk aqueous-phase reactions were identified in the cloud water. Organosulfate compounds, with both biogenic and anthropogenic volatile organic compound precursors, were detected for cloud water samples influenced by air masses that had traveled over forested and populated areas. Oxidation 35 products of long-chain (C10-12) alkane precursors were detected during urban influence. Influence of Canadian wildfires resulted in increased numbers of identified sulfur-containing compounds and oligomeric species, including those formed through aqueous-phase reactions involving methylglyoxal. Light absorbing aqueous-phase products of syringol and guaiacol oxidation were observed in the wildfireinfluenced samples, and dinitroaromatic compounds were observed in all cloud water samples (wildfire, 40

biogenic/urban, and urban-influenced). Overall, the cloud water molecular composition depended on air mass source influence and reflected aqueous-phase reactions involving biogenic, urban, and biomass burning precursors.

Introduction
Aqueous reactions have been suggested as an important source of high molecular weight organic matter 5 in the atmosphere (Ervens et al., 2015;Herrmann et al., 2015). Similar to sulfate formation upon dissolution and oxidation of SO2 in cloud droplets (Mohnen and Kadlecek, 1989), water-soluble organic gases can also be incorporated into cloud droplets and undergo aqueous reactions (Blando and Turpin, 2000), depending on their solubility (Ervens et al., 2011). Modeling has shown that including cloud processing reactions improves prediction of secondary organic aerosol (SOA) mass concentrations 10 . However, there is substantial inconsistency between the chemical composition of laboratory generated and ambient organic aerosol, in part likely due to aqueous processing (Chen et al., 2015). Potential products of aqueous processing include carboxylic acids, esters, organosulfur compounds, polyols, amines, amino acids, and highly oxygenated oligomeric species (Blando and Turpin, 2000;Ervens et al., 2011). However, only limited ambient evidence of cloud processing involving high 15 molecular weight organic compounds has been reported (Boone et al., 2015;Feng and Möller, 2004;Lee et al., 2011Lee et al., , 2012Pratt et al., 2013;Sagona et al., 2014;van Pinxteren et al., 2016;Zhao et al., 2013). A recent review found that more than 50% of organic matter in fog and cloud water remains unspeciated (Herckes et al., 2013). Therefore, examination of high molecular weight organic compounds present in cloud water is crucial to improve our current understanding of the complex multiphase chemical processes 20 leading to SOA formation and transformation.
Laboratory studies have shown that aqueous-phase reactions can lead to the formation of organonitrate and organosulfate compounds Minerath et al., 2008;Minerath and Elrod, 2009;Perri et al., 2010). Organosulfates are suggested to be formed from acid-catalyzed reactive uptake of epoxides onto sulfate aerosol (Surratt et al., 2010), as well as in-cloud radical-radical reactions 25 involving sulfate radicals (Nozière et al., 2010;Perri et al., 2010;Schindelka et al., 2013). Isoprene and monoterpene-derived organosulfates have been observed in cloud water over the southeastern United States (Boone et al., 2015;Pratt et al., 2013). Modeling suggests that organosulfate formation most commonly occurs in aqueous aerosol rather than in cloud droplets (McNeill et al., 2012). Evidence of tertiary organosulfate hydrolysis in cloud water was observed in the southeast US when cloud water 30 composition was compared to below-cloud aqueous aerosol, showing another cloud processing pathway (Boone et al., 2015).
Aqueous-phase processes, particularly in aqueous aerosol, may also produce high molecular weight compounds, such as oligomers (Ervens et al., 2011). Laboratory studies have shown the production of oligomers in bulk aqueous solutions from the photochemical oxidation of glyoxal, methylglyoxal, 35 pyruvic acid, and methacrolein (Ervens et al., 2011). Oligomers have been observed in ambient cloud water, suggesting the influences of these aqueous-phase reactions (Boone et al., 2015;Pratt et al., 2013;Zhao et al., 2013). However, with few studies reporting the molecular composition of high molecular weight compounds, it is difficult to assess the potential importance of these aqueous-phase derived compounds and their impacts on SOA composition and wet deposition to ecosystems (Hallquist et al., 5 2009).
In this work, cloud water samples were collected at Whiteface Mountain, NY during August-September 2014 to investigate the influence of various sources and aqueous-phase reactions on the molecular composition of dissolved organic compounds. An in-depth summary of Whiteface Mountain cloud water studies has been published elsewhere (Schwab et al., 2016a). Whiteface Mountain cloud 10 water is more acidic than rainwater, with higher sulfate and nitrate concentrations (Aleksic et al., 2009), which are attributed to fossil fuel combustion (Dukett et al., 2011). Cloud deposition of total soluble sulfur accounts for 80-90% of total sulfur deposition during June-September (Baumgardner et al., 2003). The cloud water samples were analyzed by electrospray ionization (ESI) coupled to Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS) in negative ion mode to characterize high molecular 15 weight oxidized organic compounds. Nitrogen-containing, sulfur-containing, and oxygenated organic compounds (CHNO, CHNOS, CHOS, and CHO) were identified in the cloud water and examined in the context of inorganic ion concentrations, pH, and total organic carbon (TOC) concentrations to gain further insights into the aqueous-phase transformations of atmospheric organic aerosol.

Sampling collection and analysis for inorganic ions, pH, and TOC
Cloud water samples were collected during August-September 2014 at Whiteface Mountain (44.3659° N, 73.9026° W; summit observatory 1483 m above mean sea level (AMSL)) in the Adirondack Mountains in New York. An in-depth site description is published elsewhere (Schwab et al., 2016b). Sampling times are provided in Table 1. Cloud water collection and analysis efforts by the Adirondack Lake Survey 25 Corporation (ALSC) closely follow the original guidelines established by the EPA Clean Air Status and Trends Network (CASTnet) program in 1994. An omni-direction passive collector (Aleksic and Dukett, 2010) collected cloud water samples when the following conditions were met: cloud liquid water content > 0.05 g m -3 (to indicate the presence of clouds), temperature > 2°C (to avoid freezing of the sampler), wind speed > 2 m s -1 (to assist with cloud water collection by the sampler), and no precipitation (to isolate 30 cloud droplets only). A field blank was collected at the summit of Whiteface Mountain by pouring deionized water into a sample container. Routine analyses of inorganic ions (SO4 2-, NO3 -, Cl -, Ca 2+ , Mg 2+ , Na + , K + , NH4 + ), pH, and TOC are performed by the ALSC, as described elsewhere (Baumgardner et al., 2003). Average August-September values from 2010-2015 are presented in (Table S1), and the data can be accessed at http://www.adirondacklakessurvey.org/.

ESI-FTICR-MS analysis
Cloud water samples and a field blank for mass spectrometric analysis were pre-concentrated using solid phase extraction (Strata™-X polymeric SPE sorbents), according to the method of Zhao et al. (2013) to 5 enhance detection and to reduce potential matrix effects from inorganic ions. TOC concentrations of each cloud water sample after SPE ranged from 15-398 mg L -1 (Table S2), with original TOC concentrations shown in Table 1. It should be noted that lower molecular weight organic compounds (i.e. most isoprene oxidation products), including some organosulfates, are expected to be removed during extraction (Zhao et al., 2013). 10 A 12-tesla Bruker SolariX FTICR mass spectrometer (Billerica, MA) was used to collect ultrahigh-resolution mass spectra of the cloud water samples. The detailed instrument parameters and analysis procedures have been described previously (Kujawinski and Behn, 2006;Minor et al., 2012;Tfaily et al., 2015). A standard Bruker electrospray ionization (ESI) source was used to generate negatively charged molecular ions to target analysis of oxidized organic compounds. Samples were first 15 diluted in LC-MS grade acetonitrile to similar TOC concentrations 15 to 45 mg L -1 and then introduced through a syringe pump at a flow rate of 3.0 μL min -1 through a 200 μm i.d. fused silica transfer tube.
Experimental conditions were as follows: needle voltage, +4.5 kV; Q1 set to 100 m/z; heated glass coated metal capillary operated at 180ºC, and the ion accumulation time was adjusted for best spectral quality.
These parameters were chosen based on previous experiments that optimized complex organic matter 20 characterization (Tfaily et al., 2011(Tfaily et al., , 2012(Tfaily et al., , 2013. Two hundred individual scans were co-added, and the average resolving power (m/Δm) was >450,000 at 451 Da. Mass spectra were calibrated by two dissolved organic matter homologous series separated by 14 Da (-CH2 groups), and the mass accuracy was calculated to be <1 ppm for singly charged ions ranging across the mass spectral distribution (m/z 100-1000). A solvent blank of acetonitrile was also analyzed. A threshold was set to assign peaks with S/N > 25 15 from the spectra of all samples. Only sample ion peaks with intensities at least 100 times higher than those in either blank or acetonitrile were retained for further data analysis. Assignments were made with elements limited to C, H, O, N, and S, with S=0-1, N=0-2, and the minimum number of O atoms per detected S or N was set to be 3. Assignments were limited to mass error of ± 0.5 ppm. All 13 C isotopic peaks were removed. These conservative limits of data processing and assignments resulted in 30 approximately 430-2300 peaks identified per sample (up to ~60% of the total number of detected peaks, after removal of 13 C and blank peaks). A complete list of assigned peaks is provided in Table S3. For individual elemental composition CcHhNnOoSs, double bond equivalents (DBE) were determined by Eq.
(1), where c, h, and n are the number of carbon, hydrogen, and nitrogen atoms in the molecular formula, respectively. 35 Sulfur and oxygen are divalent, and nitrogen is trivalent here. Therefore, the DBE calculation does not account for tetravalent and hexavalent S or pentavalent N (Mazzoleni et al., 2010).

Backward air mass trajectory analyses
Sources of the air masses arriving at the sampling site were evaluated using NOAA Hybrid Single Particle 5 Lagrangian Integrated Trajectory Model (HYSPLIT) analysis (Stein et al., 2015). Six day backward air mass trajectories were run from the beginning of the sample collection time at a starting height of 1500 m amsl, based on the Whiteface Mountain Observatory altitude (1483 m

Results and Discussion
Overall, Whiteface Mountain cloud water samples show a correlation between water acidity and the average oxygen-to-carbon atomic ratios ( sources. The average O:C ratios of the grouped biogenic, urban, and wildfire-influenced cloud water samples were 0.460 ± 0.008 (95% confidence interval), 0.489 ± 0.007, and 0.534 ± 0.005, respectively. 25 Average H:C ratios and DBE values corresponding to each of the air masses are listed in Table S4. The following sections discuss the cloud water chemical composition as a function of air mass influence.

Biogenic influence
For the cloud water samples collected on August 16-17, 2014 (A1-A3, Table 1), the air mass had traveled from the Canadian boreal forest before passing over Detroit, MI and Buffalo, NY prior to arrival at 30 Whiteface Mountain (Fig. S1). The measured cloud water pH was 4.79-5.25, less acidic than samples collected under greater urban or biomass burning influence, as well as the Aug.-Sept. 2014 average pH (4.8 ± 0.1, 95% confidence interval) ( Table 1). These cloud water samples had the lowest nitrate, sulfate, chloride, and ammonium concentrations compared to the other samples. Notably, the measured TOC and sulfate concentrations were 0.73-2.16 mg L -1 and 3.6-9.7 μM, respectively, which are below the Aug.-Sept. 2014 averages of 3.1 ± 0.8 mg L -1 and 32 ± 9 µM (Table 1) (Table S1) Table S5. For the biogenic-influenced cloud water samples (A1-A3), CHO 10 compounds compose 51-56%, by number, of the dissolved organic species identified (Table S5). CHO compounds potentially arising from aqueous reactions were identified by comparison to laboratory studies of the aqueous photooxidation of α-pinene oxidation products (cis-pinonic acid and the surrogate tricarballylic acid) (Aljawhary et al., 2016) and isoprene oxidation products (methyl vinyl ketone and isoprene SOA) (Nguyen et al., 2012a;Renard et al., 2015) (Table 3). CHOS and CHNOS compounds 15 account for 14% and 7-9%, respectively, of the measured organic compounds in the A1 and A2 samples; lower percentages were observed for sample A3 (Table S5), consistent with the lower sulfate concentration in A3 (Table 1). These biogenic-influenced cloud water samples contained several unique CHOS and CHNOS compounds with relatively low O:C (0.05-0.4) and H:C (0.7-1.1) ratios ( Fig. 3 and   S4), compared to the sulfur-containing organic compounds in other cloud water samples. These unique 20 compounds all contained high DBE values (9 or greater for C16-C25 molecules), suggesting the presence of multiple double bonds or ring structures. In the Amazon, Kourtchev et al (2016) previously found a similar subset of these unidentified CHOS compounds in SOA samples, suggesting a biogenic influence.
The mass spectra of the B1 and B2 cloud water samples are shown in Figures 2 and S2. A total of 20 2276 oxidized organic compounds were identified in the B1 sample, highlighting the diversity and complexity of the cloud water molecular composition. The CHO compounds were the most prevalent at 44-53%, by number, similar to the other air mass influences (Table S5). CHO compounds with assignments consistent with aqueous-phase reaction products, based on laboratory studies, are noted in Table 3. Notably, CHNO compounds account for 36-43% of the identified compounds, generally much 25 higher than the number fractions (as well as absolute numbers) observed in the biogenic-influenced cloud water samples (Table S5). This is likely due to the influence of elevated NOx emissions in urban locations due to fossil fuel combustion. The CHNO compounds in these samples ranged from containing 7 to 32 carbon atoms, with a median of C18 (Fig. 2). 24% of the CHNO compounds in the urban-influenced samples were composed of more than 23 carbon atoms, compared to only 3% and 12% of the CHNO in 30 the biogenic and wildfire-influenced samples, respectively, suggesting the influence of long-chain fossil fuel-derived organic precursors in these urban-influenced samples. suggested aqueous-phase formation of the CHNO compounds, and due to wintertime sampling, Zhao et al. (2013) suggested contribution from residential wood burning.
CHOS and CHNOS account for 7-10% and 3-4%, by number, respectively, of the identified organic compounds in the urban-influenced samples (Table S5). A fraction of the CHOS compounds, unique to sample B1, were characterized by low O:C (0.05-0.25) and H:C (0.6-1.2) ratios, similar to 5 previous cloud water observations by Zhao et al. (2013). Notably, aliphatic organosulfate species derived from photooxidation of long-chain alkane precursors (C10-12), including dodecane, cyclodecane, and decalin, recently observed in laboratory and urban ambient aerosol studies (Riva et al., 2016;Tao et al., 2014), were detected in the B1 cloud water sample (Table 2), consistent with the urban air mass influence.
In a recent study by Boone et al. (2015) in the southeastern US, several of these aliphatic organosulfates 10 were observed in below-cloud atmospheric particles, but not in the cloud water. Similar to the northern biogenic-influenced cloud water, many likely monoterpene-derived organosulfates and nitrooxyorganosulfates (Nguyen et al., 2012b;Surratt et al., 2008), as well as corresponding hydrolysis products (or precursors), were observed in the B1 and B2 cloud water samples (Table 2 and 3), due to the time the air mass spent over the forested areas of the southeast and midwest United States (Fig. S1). 15 inorganic ion concentrations were the highest observed in this study, although sample C3 had lower inorganic ion concentrations than the C1 and C2 samples (Table 1). Elevated K + concentrations, commonly used as a tracer of biomass burning (Artaxo et al., 1994), were present in the wildfire samples (1. 30-5.38 µM), compared to the non-wildfire conditions (biogenic and urban samples: 0.02-0.87 µM) (Table 1). A positive correlation between K + and TOC concentrations was observed (Fig. S6), with 25 increased TOC mass concentrations (7.86-16.6 mg L -1 ) in wildfire-influenced cloud water, compared to non-wildfire conditions (0.73-2.16 mg L -1 ), as well as the Aug.-Sept. 2010-2015 averages (Table S1).

Canadian wildfire influence
While the contribution of aqueous SOA formation to the measured TOC cannot be determined in this study, Gilardoni et al. (2016) reported production of light-absorbing SOA and an increase in O:C ratio from aqueous-phase processing of biomass burning emissions. Notably, there appear to be a greater 30 diversity of oligomeric compounds present in the wildfire samples (Fig. 2, Fig. S2), which is likely a combination of the identities of the specific organic compounds present at high concentrations in the smoke, as well as the potential role of acidity (as observed through cloud water acidity), resulting in the production of the observed oligomeric species.
Previously, Nguyen et al. (2012b) observed the formation of organosulfates following evaporation and re-dissolution of an aqueous solution (pH 2-4) of sulfuric acid mixed with SOA formed from dlimonene ozonolysis. In this study, numerous possible organosulfates corresponding to the formulas observed by Nguyen et al. (2012b) were identified in urban and wildfire-influenced cloud water samples, but not in the biogenic samples (Table 4). Notably, unlike the common organosulfates, likely formed 5 through reactive uptake of epoxides on aqueous aerosol (Surratt et al., 2010) and listed in Table 2, these organosulfates were primarily observed in the B1, C1, and C2 samples, which were characterized by the highest sulfate cloud water concentrations (30.7, 69.4, and 103.2 µM, respectively) and lowest pH (4. 05-4.18), compared to the other cloud water samples (SO4 2-<10.1 µM, pH >4.5) (Table 1). Since these organosulfates were observed in cloud water, the compounds are likely relatively stable toward hydrolysis. 10 However, possible hydrolysis products of many of the organosulfates were also observed (Table 4).
Overall, these observations are consistent with suggested acid-catalyzed organosulfate formation via the formation of aldol condensation products, followed by reaction with sulfuric acid, during cloud droplet evaporation (Nguyen et al. 2012b).
Biomass burning is one of the largest sources of methylglyoxal (Fu et al., 2008), and aqueous-15 phase reactive uptake of methylglyoxal has been suggested to be a significant source of organic aerosol (Fu et al., 2009;Zhao et al., 2006) through the formation of carboxylic acids, organosulfates, and oligomers Ervens et al., 2011). During wildfire influence, oligomeric species were observed in cloud water samples, with the addition of up to six C3H4O2 (methylglyoxal) monomers to glycolic acid, pyruvic acid, malonic acid, malic acid, glyoxylic acid, and succinic acid based compounds 20 (Fig. 5). The masses corresponding to these oligomeric series were also observed by Altieri et al. (2008) in bulk aqueous-phase laboratory studies of the photooxidation of methylglyoxal, suggesting the importance of methylglyoxal-hydroxyl radical reactions for oligomer formation. Notably, pyruvic acid is one of the most abundant photooxidation products of methylglyoxal (Altieri et al., 2006;Lim et al., 2013).
Modeling suggests that methylglyoxal oligomerization primarily occurs in aqueous aerosols, rather than 25 during cloud processing (Lim et al., 2013;Tan et al., 2010). Aqueous-phase methylglyoxal reaction products are expected to remain largely in the aerosol phase upon cloud droplet evaporation (Heaton et al., 2009;Loeffler et al., 2006). Syringol (2,6-dimethoxyphenol) and guaiacol (2-methoxyphenol) are also emitted in significant quantities from biomass burning, and previous studies have examined the bulk aqueous-phase photooxidation of syringol and guaiacol (Yu et al., 2014(Yu et al., , 2016, showing production of 30 several CHO compounds that were also observed in the wildfire-influenced cloud water studied here (Table 3). Notably, aqueous SOA formation from phenolic compounds has been shown to enhance light absorption in the UV-visible region (Yu et al., 2014), suggesting that these brown carbon compounds in the cloud water may be important upon cloud droplet evaporation.

Conclusions and atmospheric implications
This study represents only the second determination of high molecular weight organic molecular composition in cloud water at Whiteface Mountain, NY (Sagona et al., 2014) and the first using ultrahighresolution mass spectrometry. The average O:C ratios of the oxygenated high mass organic compounds (negative ion mode, m/z 100-1000) increased with decreasing cloud water pH, suggesting the influence 5 of aqueous acid-catalyzed and acid-dependent (radical) oxidation reactions contributing to the dissolved organic compounds. However, without knowledge of the organic aerosol composition prior to cloud formation, it is not possible to distinguish between aqueous aerosol and cloud droplet processes in this study. Cloud water samples from different air masses featured notable differences in the detected compounds. A higher absolute number and number fraction of CHNO compounds were observed during 10 the wildfire and urban air mass periods, compared to biogenic influence, likely due to greater NOx influence. In addition, observations of likely dinitroaromatics during all air mass influences, potentially from aqueous nitration of nitroaromatic compounds, have important implications because of their lightabsorbing and mutagenic properties (Purohit and Basu, 2000;Zhang et al., 2011Zhang et al., , 2013. During wildfire influence, the cloud water showed evidence of aqueous SOA formation, including oligomer formation 15 involving methylglyoxal Yasmeen et al., 2010) and aqueous-phase reactions of syringol and guaiacol (Yu et al., 2014(Yu et al., , 2016. Monoterpene-derived organosulfates and organonitrates (Surratt et al., 2008) were observed in the cloud water during all air mass influences, similar to previous cloud water studies (Boone et al., 2015;Pratt et al., 2013). Notably, long chain alkane-derived organosulfates (Riva et al., 2016) were observed in cloud water when urban influence was present. Future 20 work comparing cloud water composition with aqueous aerosol is needed to isolate in-cloud processes from aqueous-aerosol processes.
Competing Interests: The authors declare that they have no competing financial interests.         Figure   S2 and S3, respectively.  with the oligomeric series previously reported by Altieri et al. (2008) from the aqueous reactions of methylglyoxal + ·OH.