Supplement of Characterization of aerosol composition , aerosol acidity , and organic acid partitioning at an agriculturally intensive rural southeastern US site

The implementation of stringent emission regulations has resulted in the decline of anthropogenic pollutants including sulfur dioxide (SO2), nitrogen oxides (NOx), and carbon monoxide (CO). In contrast, ammonia (NH3) emissions are largely unregulated, with emissions projected to increase in the future. We present real-time aerosol and gas measurements from a field study conducted in an agriculturally intensive region in the southeastern US during the fall of 2016 to investigate how NH3 affects particle acidity and secondary organic aerosol (SOA) formation via the gas–particle partitioning of semi-volatile organic acids. Particle water and pH were determined using the ISORROPIA II thermodynamic model and validated by comparing predicted inorganic HNO3-NO−3 and NH3-NH + 4 gas–particle partitioning ratios with measured values. Our results showed that despite the high NH3 concentrations (average 8.1± 5.2 ppb), PM1 was highly acidic with pH values ranging from 0.9 to 3.8, and an average pH of 2.2± 0.6. PM1 pH varied by approximately 1.4 units diurnally. Formic and acetic acids were the most abundant gas-phase organic acids, and oxalate was the most abundant particle-phase water-soluble organic acid anion. Measured particle-phase water-soluble organic acids were on average 6 % of the total non-refractory PM1 organic aerosol mass. The measured molar fraction of oxalic acid in the particle phase (i.e., particle-phase oxalic acid molar concentration divided by the total oxalic acid molar concentration) ranged between 47 % and 90 % for a PM1 pH of 1.2 to 3.4. The measured oxalic acid gas–particle partitioning ratios were in good agreement with their corresponding thermodynamic predictions, calculated based on oxalic acid’s physicochemical properties, ambient temperature, particle water, and pH. In contrast, gas–particle partitioning ratios of formic and acetic acids were not well predicted for reasons currently unknown. For this study, higher NH3 concentrations relative to what has been measured in the region in previous studies had minor effects on PM1 organic acids and their influence on the overall organic aerosol and PM1 mass concentrations.

Abstract. The implementation of stringent emission regulations has resulted in the decline of anthropogenic pollutants including sulfur dioxide (SO 2 ), nitrogen oxides (NO x ), and carbon monoxide (CO). In contrast, ammonia (NH 3 ) emissions are largely unregulated, with emissions projected to increase in the future. We present real-time aerosol and gas measurements from a field study conducted in an agriculturally intensive region in the southeastern US during the fall of 2016 to investigate how NH 3 affects particle acidity and secondary organic aerosol (SOA) formation via the gas-particle partitioning of semi-volatile organic acids. Particle water and pH were determined using the ISORROPIA II thermodynamic model and validated by comparing predicted inorganic HNO 3 -NO − 3 and NH 3 -NH + 4 gas-particle partitioning ratios with measured values. Our results showed that despite the high NH 3 concentrations (average 8.1 ± 5.2 ppb), PM 1 was highly acidic with pH values ranging from 0.9 to 3.8, and an average pH of 2.2 ± 0.6. PM 1 pH varied by approximately 1.4 units diurnally. Formic and acetic acids were the most abundant gas-phase organic acids, and oxalate was the most abundant particle-phase water-soluble organic acid anion. Measured particle-phase water-soluble organic acids were on average 6 % of the total non-refractory PM 1 organic aerosol mass. The measured molar fraction of oxalic acid in the particle phase (i.e., particle-phase oxalic acid molar con-centration divided by the total oxalic acid molar concentration) ranged between 47 % and 90 % for a PM 1 pH of 1.2 to 3.4. The measured oxalic acid gas-particle partitioning ratios were in good agreement with their corresponding thermodynamic predictions, calculated based on oxalic acid's physicochemical properties, ambient temperature, particle water, and pH. In contrast, gas-particle partitioning ratios of formic and acetic acids were not well predicted for reasons currently unknown. For this study, higher NH 3 concentrations relative to what has been measured in the region in previous studies had minor effects on PM 1 organic acids and their influence on the overall organic aerosol and PM 1 mass concentrations.

Introduction
Ammonia (NH 3 ) is the most abundant basic gas in the troposphere and plays an important role in many atmospheric processes. It is a major neutralizer of atmospheric acidic species, reacting readily with sulfuric acid (H 2 SO 4 ) and nitric acid (HNO 3 ) to form ammonium sulfate and nitrate salts (e.g., (NH 4 ) 2 SO 4 , and other forms such as NH 4 HSO 4 , (NH 4 ) 3 H(SO 4 ) 2 , and NH 4 NO 3 ), which are often the main inorganic components of atmospheric aerosols. The forma-Published by Copernicus Publications on behalf of the European Geosciences Union.
tion of particle-phase ammonium sulfate and nitrate salts in the aerosol phase depends on the thermodynamic states of their precursors and the environmental conditions, which can consequently affect aerosol pH. For example, Guo et al. (2017b) showed that for southeastern US summertime conditions, as aerosol pH increases, the relative fractions of SO 2− 4 and HSO − 4 increase and decrease, respectively. Wet and dry deposition are the principle NH 3 sinks (Dentener and Crutzen, 1994). NH 3 is spatially heterogeneous, with the highest concentrations typically found near emission sources (Seinfeld and Pandis, 2016). The dominant NH 3 sources in rural areas are agricultural in nature, and include the application of fertilizers and volatilization of livestock waste (Reis et al., 2009;Ellis et al., 2013;Van Damme et al., 2014). Biomass burning, either from wildfires or from controlled burning during land-clearing operations, is also a significant source of NH 3 in rural environments. The primary source of NH 3 in urban areas is industrial emissions (e.g., NH 3 synthesis, manufacture of ammonium nitrate and urea, fluid, and thermal catalytic cracking processes in petroleum refineries), though vehicular emissions can be a significant NH 3 source in some heavily populated cities (Reis et al., 2009;Lamarque et al., 2010;Yao et al., 2013;Sun et al., 2017). Vehicular NH 3 emissions are thought to be produced primarily from the reaction of nitrogen oxide with hydrogen in the presence of carbon monoxide in three-way catalysts of gasoline lightduty vehicles (Barbier Jr. and Duprez, 1994;Whittington et al., 1995;Livingston et al., 2009;Suarez-Bertoa et al., 2014).
In the US, implementation of stringent emission controls on traditional anthropogenic air pollutants, such as sulfur dioxide (SO 2 ), nitrogen oxides (NO x ), and carbon monoxide (CO), have led to steady decreases in their emissions, and consequently their concentrations (Blanchard et al., 2013b;Xing et al., 2013). In contrast, NH 3 emissions are largely unregulated and are projected to increase due to increased agricultural operations to feed a growing world population (Reis et al., 2009;Ellis et al., 2013). Satellite observations showed that gas-phase NH 3 concentrations have increased substantially in US agricultural areas from 2002 to 2014 (Warner et al., 2017). More wildfires from a changing climate, or from controlled burning for land clearing for agricultural use, may also lead to increased NH 3 emissions (Reis et al., 2009;Pechony and Shindell, 2010;Warner et al., 2016). These trends suggest that NH 3 could play an increasingly important role in atmospheric chemistry.
Previous laboratory studies have shown that NH 3 can influence secondary organic aerosol (SOA) formation and processing. For example, NH 3 increases SOA mass yields in the α-pinene ozonolysis system and is hypothesized to be due to the formation of ammonium salts from the reaction of NH 3 with organic acids (Na et al., 2007). The heterogeneous uptake of NH 3 by SOA can also lead to the formation of particulate organonitrogen compounds, a class of brown carbon species that can reduce visibility and impact climate (J. Laskin et al., 2010;Updyke et al., 2012;Lee et al., 2013;A. Laskin et al., 2015).
The southeastern US is a natural outdoor laboratory for studying the effects of biogenic-anthropogenic interactions on atmospheric aerosol formation and processing. Subtropical vegetation composed mainly of mixed conifer and deciduous forests emits large quantities of biogenic volatile organic compounds (BVOCs) that can act as precursors for SOA formation (Blanchard et al., 2011(Blanchard et al., , 2013aGuenther et al., 2012). Large urban centers and small towns are surrounded by large expanses of forests and widespread rural areas with agricultural activities. Scattered within the southeastern US are also coal-burning power plants and industrial facilities. Anthropogenic activities in this region emit large concentrations of VOCs, SO 2 , NO x , CO, NH 3 , and aerosols (Blanchard et al., 2013c). Similar to other parts of the US, SO 2 , CO, and NO x concentrations have decreased steadily in the southeastern US due to the implementation of emission controls (Blanchard et al., 2013b). In contrast, gas-phase NH 3 concentrations have increased in the southeastern US over the same time period (Saylor et al., 2015). These factors make the southeastern US an intriguing place to study the influence of NH 3 on atmospheric aerosol chemistry.
We performed aerosol and gas measurements during a field study conducted in Yorkville, Georgia, US, in the fall of 2016, with the goal of understanding how NH 3 affects aerosol acidity and SOA formation. The field site is surrounded by forest and agricultural land, affording an opportunity to make ambient observations in an area impacted by local emissions of BVOCs and NH 3 . In this paper, we present gas and aerosol composition measurements that include a suite of organic acids. The thermodynamic equilibrium model, ISORROPIA II, is used to calculate particle water and pH based on measured inorganic aerosol and gas composition (Nenes et al., 1998;Fountoukis and Nenes, 2007), and these predictions are compared to observed gas-particle partitioning of NH 3 , HNO 3 , and organic acids. Together, these measurements are used to determine how aerosol acidity affects the mass concentration of particle-phase organic acids at this site.

Field site
Aerosol and gas measurements were conducted at the Yorkville, Georgia (33.929 • N, 85.046 • W) SouthEastern Aerosol Research and Characterization (SEARCH) field site from mid-August to mid-October 2016. This is one of the sampling sites for the Southeastern Center for Air Pollution and Epidemiology (SCAPE) study in which aerosol characterization measurements were conducted in the summer and winter of 2012 (Xu et al., 2015a, b). A detailed description of the field site can be found in Hansen et al. (2003). This Atmos. Chem. Phys., 18,2018 www.atmos-chem-phys.net/18/11471/2018/ rural site is situated in a mixed forest-agriculture area approximately 55 km northwest and generally upwind of Atlanta. The immediate surrounding area is used for cattle grazing and poultry concentrated animal feeding operations (CAFOs) (Fig. S1 in the Supplement). There are no major roads near the field site and nearby traffic emissions were negligible. A large coal-fired power plant (Plant Bowen) is situated approximately 25 km north of the site. Hence, the field site is impacted mainly by BVOC and NH 3 emissions, with occasional spikes in SO 2 and minimal influence from urban anthropogenic pollutants such as HNO 3 , O 3 , NO x , and CO (Fig. S2). The sampling period was characterized by moderate temperatures (24.0 • C average, 32.6 • C maximum, 9.5 • C minimum) and high relative humidities (RHs) (68.9 % RH average, 100 % RH maximum, 21.6 % RH minimum). Meteorological data are shown in Fig. S3. Data reported are displayed in eastern daylight time (EDT).

Instrumentation
Instruments were housed in a temperature-controlled (∼ 20 • C) trailer during the field study. Gas-phase HNO 3 , SO 2 , and organic acids (formic, acetic, oxalic, butyric, glycolic, propionic, valeric, malonic, and succinic acids) were measured with a custom-built chemical ionization mass spectrometer (CIMS) using sulfur hexafluoride ions (SF − 6 ) as reagent ions. SO 2 and HNO 3 were detected as fluoride adducts (F 2 SO − 2 and NO − 3 q HF, respectively) while the organic acids (HX) were detected primarily as conjugated anions (X − ) using the quadrupole mass spectrometer (Huey et al., 1995(Huey et al., , 2004Nah et al., 2018). This CIMS is referred to hereafter as the SF 6 -CIMS. Gas-phase NH 3 was measured with an additional custom-built CIMS using protonated ethanol clusters ((C 2 H 5 OH) + n ) as reagent ions. NH 3 was detected primarily as NH + 4 ions with the quadrupole mass spectrometer (Nowak et al., 2002;Yu and Lee, 2012;You et al., 2014a). This CIMS is referred to hereafter as the NH 3 -CIMS.
Since HNO 3 , NH 3 , and organic acids may condense on surfaces, both SF 6 -CIMS and NH 3 -CIMS used inlet configurations that minimized wall interactions (Huey et al., 2004;Nowak et al., 2006). Each CIMS was connected to an inlet (a 7.6 cm ID aluminum pipe) that protruded beyond the trailer's wall by ∼ 40 cm into the ambient air. Both inlets were ∼ 2 m above the ground. A donut-shaped ring was attached to the ambient sampling port of each pipe to curtail the influence of crosswinds on the pipe's flow dynamics. Both rings were wrapped with a fine wire mesh to prevent ingestion of insects. A flow of ∼ 2800 L min −1 was maintained in each pipe using regenerative blowers (Ametek Windjammer 116637-03). Part of this flow (7 L min −1 for the SF 6 -CIMS and 4.6 L min −1 for the NH 3 -CIMS) was sampled through a custom-made three-way PFA Teflon valve, which connected the pipe's center to the CIMS sampling orifice and could be switched automatically between ambient and background measurements.
Background measurements were performed every 25 min for 4 min for both the SF 6 -CIMS and NH 3 -CIMS. During each background measurement, the sampled air flow was passed through an activated charcoal scrubber (Sigma Aldrich) that removed SO 2 , HNO 3 , and organic acids prior to delivery into the SF 6 -CIMS, and through a silicon phosphate scrubber (Perma Pure LLC) that removed NH 3 prior to delivery into the NH 3 -CIMS. More than 99 % of the targeted species were removed during background measurements for both the SF 6 -CIMS and NH 3 -CIMS. Standard addition calibrations were performed every 5 h for the SF 6 -CIMS using the outputs of a 1.12 ppm 34 SO 2 gas cylinder (Scott Marrin Inc.) and a formic or acetic acid permeation device (VICI Metronics). Calibrations for the other gases measured by the SF 6 -CIMS were performed in post-field laboratory work, details of which can be found in Nah et al. (2018) and Supplement Sect. S1. Standard addition calibrations were performed hourly for the NH 3 -CIMS using the output of a NH 3 permeation device (KIN-TEK). The outputs of the formic and acetic acid permeation devices were measured periodically by scrubbing the output of the permeation tube in deionized water, followed by ion chromatography analysis for formate and acetate. The emission rate of the NH 3 permeation device was measured using UV optical absorption (Neuman et al., 2003).
The detection limits for species measured by the SF 6 -CIMS and NH 3 -CIMS were approximated from 3 times the standard deviation values (3σ ) of the ion signals measured during background mode. The detection limits for HNO 3 , SO 2 , and the various organic acids measured by the SF 6 -CIMS ranged from 1 to 60 ppt for 2.5 min integration periods, which corresponded to the length of a background measurement with a ∼ 4 % duty cycle for each m/z (Table S1 in the Supplement). Measurement uncertainties for the concentrations of HNO 3 , SO 2 , and the various organic acids originate mainly from calibration measurements and were between 12 % and 25 % (Table S1). The detection limit for NH 3 measured by the NH 3 -CIMS was 1 ppb for 2.3 min integration periods, which corresponded to the length of a background measurement with a ∼ 29 % duty cycle for the NH + 4 ion. Measurement uncertainties for NH 3 concentrations were 13 %.
A high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS, Aerodyne Research Inc.) was used to measure the elemental composition of ambient nonrefractory PM 1 (particles with aerodynamic diameters < 1 µm). Ambient air was sampled at 16.7 L min −1 through a URG PM 1 cyclone and then through a Nafion dryer prior to delivery into the HR-ToF-AMS. Aerosols were dried to RH < 20 % to eliminate the influence of RH on the HR-ToF-AMS's particle collection efficiency. A detailed description of the HR-ToF-AMS can be found in the literature (DeCarlo et al., 2006;Canagaratna et al., 2007Canagaratna et al., , 2015. Briefly, the aerodynamic lens of the HR-ToF-AMS focused the dried submicron aerosols into a narrow beam. The aerosols were then im-pacted onto a heated tungsten surface (∼ 600 • C) on which they were flash vaporized. The resulting vapors were ionized by electron impact ionization (70 eV), and the ions were detected by a time-of-flight mass spectrometer. Gas-phase interferences were accounted for by subtracting the signals obtained during daily measurements of filtered, particle-free sampling air. Ionization efficiency calibrations were performed weekly using 300 nm ammonium nitrate and ammonium sulfate particles. Composition-dependent collection efficiency (CDCE) values of 0.44 to 0.55 were determined using the procedure detailed by Middlebrook et al. (2012), in which CDCE values are derived based largely on aerosol inorganic species concentrations and the RH in the sampling line. In addition, a constant collection efficiency (CE) value of 0.9 was determined from the comparison of raw HR-ToF-AMS SO 2− 4 data with other particulate SO 2− 4 measurements performed during the study. Comparisons of aerosol mass concentrations obtained from the application of CDCE values (i.e., 0.44 to 0.55) vs. a constant CE value (i.e., 0.9) to the raw HR-ToF-AMS data are discussed in Sect. 3.2. Uncertainties in HR-ToF-AMS measurements were estimated to be approximately 25 % (Canagaratna et al., 2007).
Particle-phase water-soluble organic acids and inorganic cations and anions were measured using two particleinto-liquid sampler (PILS) systems coupled to ion chromatographs (ICs) (Orsini et al., 2003). Each PILS sampled ambient air at nominally 16.7 L min −1 through a URG PM 1 cyclone. Before PILS1, which was used to measure watersoluble inorganic cation and anions, two long (24 cm) URG glass annular denuders coated with sodium carbonate and phosphorous acid were used to remove acidic and basic gases. Before PILS2, which measured water-soluble organic acids, a 28 cm parallel plate carbon denuder (Sunset Lab) was used to remove organic gases (Eatough et al., 1993). In each PILS, aerosols were mixed with water vapor at ∼ 100 • C generated from heated ultrapure deionized water (Weber et al., 2001;Orsini et al., 2003). The resulting droplets were impacted onto a plate, with the resulting liquid sample analyzed using the ICs. Each IC system was calibrated at the beginning and end of the study using five multi-compound standards in order to create calibration curves. Periodically, a HEPA filter (Pall Life Sciences) was placed on the inlet to determine the background in near real time. The measurement uncertainty for each IC system was about 10 %.
PILS1 was connected to two Dionex ICS-1500 ICs (Thermo Fisher Scientific) to measure the water-soluble inorganic ions. These two IC systems include an isocratic pump, self-regenerating anion or cation suppressor, and conductivity detector. This system will be referred to hereafter as the PILS-IC. Anions were separated using a Dionex Ion-Pac AS15 guard and analytical column (4×250 mm, Thermo Fisher Scientific) employing an eluent of 38 mM sodium hydroxide at a flow rate of 1.5 mL min −1 . Cations were separated using a Dionex IonPac CS12A guard and analytical column (4 × 250 mm, Thermo Fisher Scientific) employ-ing an eluent of 18 mM methanesulfonic acid at a flow rate of 1 mL min −1 . A new chromatogram was obtained every 30 min with a sample loop fill time (i.e., ambient sample integration time) of 20 min. The limit of detection for the various anions and cations was approximately 0.01 µg m −3 .
PILS2 was coupled to a Dionex ICS-4000 capillary highpressure ion chromatography (HPIC) system to measure the water-soluble organic acids. The HPIC includes an eluent generator, isocratic pump, degausser, suppressor, carbonate removal device, and conductivity detector. This system will be referred to hereafter as the PILS-HPIC. The organic acids were separated using a Dionex AS11-HC-4 µm capillary guard and analytical column (0.4 × 250 mm, Thermo Fisher Scientific), which used a potassium hydroxide gradient separation method at a flow rate of 0.015 mL min −1 . A new chromatogram was obtained every 60 min with a sample loop fill time of 2 min. The limit of detection for the various organic acids was approximately 0.001 µg m −3 .
Particle-and gas-phase water-soluble organic carbon (WSOC p and WSOC g , respectively) were measured using two Sievers 900 series total organic carbon (TOC) analyzers (GE Analytical Instruments), as described by Sullivan et al. (2004). For WSOC p measurements, ambient air was sampled at 15.2 L min −1 through a URG PM 1 cyclone and a parallel plate carbon denuder into a PILS coupled to the first TOC analyzer. For WSOC g measurements, ambient air was sampled at 20 L min −1 through a Teflon filter (45 mm diameter, 2.0 µm pore size, Pall Life Sciences) to remove particles in the air stream. This filter was changed every 3 to 4 days. The particle-free air was then directed to a MIST chamber filled with ultrapure deionized water, which scrubbed the soluble gases at an air flow rate of 20 L min −1 . Soluble gases with Henry's law constants greater than 10 3 mole L −1 atm −1 were scrubbed into deionized water in the MIST chamber (Spaulding et al., 2002). The resulting MIST chamber liquid sample was analyzed by the second TOC analyzer. The TOC analyzers converted the organic carbon in the liquid samples to carbon dioxide using UV radiation and chemical oxidation. The carbon dioxide formed was then measured by conductivity. The amount of organic carbon in the liquid samples is proportional to the measured increase in conductivity of the dissolved carbon dioxide. Each WSOC p and WSOC g measurement lasted 4 min. Background WSOC p and WSOC g measurements were performed for 45 min every 12 h by stopping the sample air flow and rinsing the system with deionized water. Both TOC analyzers were calibrated at the beginning and end of the study using varying concentrations of sucrose solutions to create calibration curves (as specified by the instrument manual). The limit of detections for WSOC p and WSOC g were 0.2 and 0.4 µg C m −3 , respectively. The measurement uncertainties for WSOC p and WSOC g were estimated to be 10 % based on uncertainties in the TOC analyzer, sample air, and liquid flows.
A suite of instruments operated by the SEARCH network provided supporting gas and aerosol measurements Atmos. Chem. Phys., 18, 11471-11491, 2018 www.atmos-chem-phys.net/18/11471/2018/ (Hansen et al., 2003;Edgerton et al., 2005Edgerton et al., , 2006Edgerton et al., , 2007. O 3 was measured with a UV absorption instrument (Thermo Fisher Scientific) with a temporal resolution of 1 min. NO and NO x were measured using a chemiluminescence instrument (Thermo Fisher Scientific) with a temporal resolution of 1 min. NO 2 was obtained from the difference between NO and NO x . CO was measured using a nondispersive infrared absorption instrument (Thermo Fisher Scientific) with a temporal resolution of 1 min. NH 3 was measured by a denuder-based instrument (ARA) with a temporal resolution of 5 min. Comparisons of measurements by the NH 3 -CIMS and denuder-based instrument will be presented in Sect. 3.1. A filter-based particle composition monitor (ARA) provided 24 h-integrated PM 2.5 measurements of particle mass and major inorganic ions measured offline by ion chromatography. Organic carbon (OC) and elemental carbon (EC) in PM 2.5 were measured by an OC-EC analyzer (Sunset Labs) with a temporal resolution of 1 h. This analyzer determined OC by thermal optical transmittance. VOCs were measured by a gas chromatography-flame ionization detector (GC-FID, Agilent Technologies) with a temporal resolution of 1 h.

Particle pH and water calculation
The thermodynamic equilibrium model ISORROPIA II was used to determine the phase state and composition of an -water inorganic aerosol in equilibrium with its corresponding gasphase species (Fountoukis and Nenes, 2007;Nenes et al., 1998). This approach was used in previous studies to determine particle water and pH in different parts of the world (Guo et al., 2015(Guo et al., , 2017aBougiatioti et al., 2016;Weber et al., 2016;Shi et al., 2017). The pH of an aqueous solution is defined as the negative logarithm of the hydronium ion (H 3 O + ) activity on a molality basis (http: //www.goldbook.iupac.org/html/P/P04524.html, last access: 6 July 2018): where a(H + ) is the hydronium ion activity in an aqueous solution, m(H + ) is the hydronium ion molality, γ m (H + ) is the molality-based hydronium ion activity coefficient, and m θ is the standard molality (1 mol kg −1 ). For simplicity, H 3 O + is denoted here as H + even though we recognize that the un-hydrated hydrogen ion is rare in aqueous solutions. Since most thermodynamic equilibrium models (e.g., ISOR-ROPIA II, E-AIM) do not report liquid concentrations, but instead report species in terms of concentration per volume of air (e.g., µg m −3 , µmol m −3 ), we have calculated the particle pH by pH = −log 10 γ H + H + aq = −log 10 where γ H + is the hydronium ion activity coefficient (assumed to be 1), H + aq is the concentration of hydronium ions in particle water in moles per liter (i.e., the density of water is assumed to be 1000 kg m −3 , and so pH is calculated in terms of molality), H + air (µg m −3 ) is the hydronium ion concentration per volume of air, and W i and W o (µg m −3 ) are the bulk particle water concentrations associated with inorganic and organic species per volume of air, respectively. In Eq. (1b), the molecular weight of H + is taken as 1 g mole −1 , and 1000 is the factor needed for unit conversion of grams per liter to micrograms per cubic meter. H + air and W i are outputs of the ISORROPIA II model. Previous studies have shown that particle pH values predicted using only W i are reasonably accurate since the sensitivity of particle pH to the effects of W o is small (Guo et al., 2015). For the southeastern US, Guo et al. (2015) reported that particle pH values predicted using only W i were systematically 0.15 to 0.23 units lower than those predicted using W i +W o during the 2013 Southern Oxidant Aerosol Study (SOAS) and SCAPE campaigns. Given this small deviation and that organic aerosol hygroscopicity was not measured in this field study, we report particle pH only considering W i .
ISORROPIA II was run in "forward" mode, which assumes that aerosols are "metastable" with no solid precipitates, to predict particle pH and the partitioning of semivolatile compounds. In forward mode, the model calculates the gas-particle equilibrium partitioning concentrations based on the input of the total concentration of a species (i.e., gas + particle). In reverse mode, the model calculates the gas-particle equilibrium partitioning concentrations based on the input of only the particle-phase concentration of a species. We used forward mode because the reverse mode is sensitive to measurement errors, which often result in large model biases in the predicted particle pH (Hennigan et al., 2015). The measured particle-phase inorganic NH + 4 , SO 2− 4 , and NO − 3 concentrations and gas-phase HNO 3 and NH 3 concentrations were used as model inputs. The metastable assumption is reasonable since the high RH (average RH 68.9 %) observed during the study indicated that the aerosols had likely deliquesced. We excluded data for periods when the RH was above 95 % since the exponential growth in particle liquid water with RH introduces large pH uncertainties (Malm and Day, 2001;Guo et al., 2015).
In using ISORROPIA II to predict particle pH and the partitioning of semi-volatile compounds, we also assumed that the aerosols are internally mixed and that the particle pH does not change with particle size (i.e., the overall particle pH is characterized by the particle's bulk properties). As long as some small fraction of sulfate is mixed with various aerosol components, (e.g., non-volatile cations), the assumption that aerosols are completely internally mixed has a small effect on the predicted pH (Guo et al., 2017b). However, the presence of multiple organic and inorganic species in ambient aerosols may lead to multiple phases within the particle (i.e., phase separation). Consequently, this may result in the unequal distribution of inorganic species among different phases, each with its own water activity and inorganic concentration. Previous studies have shown that liquid-liquid and solid-liquid phase separations may occur for mixed organic and inorganic aerosols at low RH and organic aerosol oxygen-to-carbon atomic ratios (O/C) (Bertram et al., 2011;Song et al., 2012;You et al., 2013You et al., , 2014bYou and Bertram, 2015). Phase separations were always observed at O/C ≤ 0.5, while no phase separation was observed at O/C ≥ 0.8. The probability for the occurrence of phase separation decreased at higher RH for 0.5 < O/C < 0.8. The average O/C for this field study is 0.69 ± 0.06. Organic acids were not included in the calculation of particle pH. This is reasonable since their total mass concentration was small compared to the total inorganic mass concentration. The average ratio of the organic acid mass concentration to the inorganic mass concentration is 0.25. Furthermore, Song et al. (2018) showed that including organic acid mass concentrations in thermodynamic model calculations had minor effects on particle pH if the system is in equilibrium. The validity of these assumptions and the resulting thermodynamic model predictions will be evaluated by comparing the predicted gas-particle partitioning ratios of semi-volatile inorganic compounds with measured values in Sect. 3.3.

NH 3 observations
Continuous measurements of NH 3 were made using the NH 3 -CIMS from 13 September to 12 October. Figure 1a and b show the time series and average diurnal profile of NH 3 , respectively. NH 3 concentrations ranged from 0.7 to 39.0 ppb (0.5 to 28.5 µg m −3 ) and exhibited consistent diurnal cycles. NH 3 was generally higher in the late mornings and early afternoons. Concentrations started to increase at 07:30, which coincided with an increase in temperature at sunrise (Fig. S3). Possible reasons for the morning increase include volatilization of particulate ammonium and animal waste, entrainment from the residual layer where NH 3 may not have been depleted, evaporation of dew or fog that contained dissolved NH 3 , and emission from plant stomata (Ellis et al., 2011). NH 3 decreased at 14:30, approximately 1 h before temperature decreased, and may be due to changes in the boundary layer height. However, this hypothesis cannot be tested since the boundary layer height was not measured during the study. The diurnal plot does not account for dilution as the boundary layer expanded and only indicates that if emissions were solely from the surface and lower concentrations aloft, these NH 3 sources were of significant magnitude. (c) Average NH 3 concentration normalized to wind speed (i.e., NH 3 concentration (ppb) × wind speed (m s −1 )) in each 10 • bin (red line). The average normalized NH 3 concentration is shown as a grey line.
The average NH 3 concentration measured by the NH 3 -CIMS is 8.1 ± 5.2 ppb. This is approximately 2 times higher than the average NH 3 concentration (3.8±2.9 ppb) measured by the denuder-based instrument operated by the SEARCH network over the same time period (Fig. S4). Differences in Atmos. Chem. Phys., 18, 11471-11491, 2018 www.atmos-chem-phys.net/18/11471/2018/ NH 3 concentrations measured by the two instruments may be due to positive and negative sampling artifacts caused by differences in sampling inlets (e.g., inlet length and location), frequency of calibration and background measurements, and (in the case of the denuder-based instrument) possible sample contamination during chemical analysis. Discussions on how differences in measured NH 3 concentrations affect PM 1 pH predictions will be presented in Sect. 3.3. Nevertheless, there is a record of NH 3 concentrations measured by the denuder-based instrument at this site since 2008. Just prior to and during this study, NH 3 concentrations are generally the highest observed since 2011 (Fig. S5). These elevated NH 3 concentrations may be due to sporadic biomass burning episodes caused by elevated temperatures and widespread drought across the southeastern US in 2016 (Park Williams et al., 2017;Case and Zavodsky, 2018). The NH 3 -CIMS measurements are examined with the meteorological data to gain insights into the primary NH 3 sources during the sampling period. To account for wind speed, the 1 h averaged NH 3 concentrations are first multiplied by their corresponding 1 h averaged wind speeds. These normalized NH 3 concentrations are then used to construct a wind direction polar plot showing the average normalized NH 3 concentration per 10 • bin (Fig. 1c). The wind direction polar plot shows that the normalized NH 3 is approximately 2 times greater than the average when air masses are transported from the southeast, the general direction of the poultry CAFOs located approximately 2 km from the field site ( Fig. S1), which are known for having high NH 3 emissions. This conclusion is reaffirmed by NH 3 measurements from the SEARCH network's denuder-based instrument. NH 3 concentrations measured by the two instruments in this study are substantially higher than those measured in three recent field studies conducted in the continental US: the 2010 California Nexus (CalNex) study, 2013 Southeast Nexus (SENEX) study, and 2013 SOAS study (see Table 1). The differences in NH 3 may be attributed to differences in land use, proximity to CAFOs, and meteorological conditions. The high NH 3 concentrations in this study allow us to make ambient observations of the effect of NH 3 on particle acidity and the gas-particle partitioning of semi-volatile inorganic and organic compounds, and compare them with previous studies.

PM 1 composition
The aerosol inorganic chemical composition was measured by several instruments during this study. The HR-ToF-AMS, PILS-IC and PILS-HPIC measured the composition of PM 1 , while a filter-based particle composition monitor measured the composition of PM 2.5 . Comparisons of aerosol SO 2− 4 , NO − 3 , and NH + 4 mass concentrations obtained from the application of CDCE values to the raw HR-ToF-AMS data are compared to those measured by the other three instruments in Fig. S6. NH + 4 measurements by the PILS-IC are not avail-able for comparison due to denuder breakthrough that occurred during the study. SO 2− 4 measurements by the various instruments are generally well correlated with each other, with R 2 values ranging from 0.64 to 0.92. Although PM 1 SO 2− 4 measurements by the two PILS systems show good agreement with each other, HR-ToF-AMS CDCE-applied SO 2− 4 measurements are approximately 2 times higher than the PILS and filter measurements. Similar systematic differences are also observed for NO − 3 and NH + 4 measurements. NO − 3 and NH + 4 measurements from the four instruments are moderately correlated (R 2 = 0.54 to 0.79 and R 2 = 0.94, respectively). NO − 3 measurements from the PILS and filter systems are mostly similar; however, HR-ToF-AMS CDCE-applied PM 1 NO − 3 and NH + 4 measurements are approximately 3 times and 2 times higher than the PILS and filter measurements. One possible reason is that the calculated CDCE is lower due to organics dominating the aerosol composition during the study (average of 74.2 ± 7.9 % of the non-refractory PM 1 mass concentration). Lee et al. (2015) suggested that a high organic mass fraction may impede the complete efflorescence of aerosols when they are passed through the drier prior to delivery into the HR-ToF-AMS, thus reducing the particle bounce and increasing the CE value. Hence, we estimated HR-ToF-AMS PM 1 mass concentrations that would be consistent with PILS and filter measurements by multiplying all the raw HR-ToF-AMS data by a constant CE value of 0.9, which was obtained from comparisons of the raw HR-ToF-AMS SO 2− 4 data with PILS-IC and PILS-HPIC SO 2− 4 measurements. The constant CE-applied HR-ToF-AMS data are used in all our subsequent analyses. Figure 2 shows the time series and average diurnal profiles of non-refractory PM 1 species. The average non-refractory PM 1 organics, SO 2− 4 , NO − 3 , and NH + 4 mass concentrations are 5.0 ± 2.3, 1.6 ± 0.4, 0.2 ± 0.1, and 0.4 ± 0.2 µg m −3 , respectively. Organics are the dominant non-refractory PM 1 species, accounting for 74.2 ± 7.9 % of the non-refractory PM 1 mass concentration during the field study. Organic aerosol mass concentration was slightly higher at night, which is likely caused by changes in the boundary layer height, emission sources, and SOA formation processes (Xu et al., 2015b). Previous studies have shown that nighttime SOA production in the southeastern US is largely attributed to nitrate radical oxidation and ozonolysis of monoterpenes, which are abundant at night (Pye et al., 2015;Xu et al., 2015a, b;Lee et al., 2016;Zhang et al., 2018). Specifically, the nitrate radical oxidation of some monoterpenes (e.g., βpinene) could form low-volatility organic nitrates that are condensable and could contribute substantially to the nocturnal organic aerosol mass (Boyd et al., , 2017. Apportionment of organic aerosol sources will be discussed in an upcoming publication. SO 2− 4 is the second most abundant non-refractory PM 1 species (16.3±5.7 % mass fraction), followed by NH + 4 (5.9 ± 2 % mass fraction) and NO − 3 (3.6 ± 2.2 % mass fraction). SO 2− 4 mass concentration peaked in the afternoon due to enhanced SO 2 photooxidation . The NO − 3 mass concentration measured by the HR-ToF-AMS is the nitrate functional group (-ONO 2 ) present on organic and inorganic nitrates. Hence, the diurnal profile of the NO − 3 mass concentration in Fig. 2 has contributions from both organic and inorganic nitrates. The mass concentrations of organic and inorganic nitrates increased after sunset and peaked at sunrise (Fig. S7), likely due to the formation of organic nitrates from nighttime NO 3 chemistry and increased gas-to-particle partitioning of organic and inorganic nitrates as temperature decreased (Xu et al., 2015a, b). Quantification and characterization of organic nitrates based on HR-ToF-AMS and PILS-IC PM 1 NO − 3 measurements will be discussed in a future publication. NH + 4 mass concentration has moderate diurnal variations with marginally higher concentrations in the afternoon, likely due to the contrasting day-night phases of ammonium sulfate and ammonium nitrate formation. SO 2− 4 , NO − 3 , and NH + 4 molar concentrations indicated that NH + 4 is mainly associated with SO 2− 4 in PM 1 .

PM 1 pH predictions
CIMS HNO 3 and NH 3 data, HR-ToF-AMS PM 1 SO 2− 4 and NH + 4 data, PILS-IC PM 1 NO − 3 and non-volatile cation (Cl − , Na + , Ca 2+ , K + , and Mg 2+ ) data, measured temperature, and RH are used as ISORROPIA II model inputs to predict PM 1 W i and pH from 13 September to 6 October. Figure 3 shows the time series and average diurnal profiles of ISORROPIApredicted PM 1 W i and pH. PM 1 is highly acidic with pH values ranging from 0.9 to 3.8 and an average pH of 2.2 ± 0.6. The average PM 1 pH is 2.5 ± 0.6 during periods when the NH 3 concentration is higher than 13.3 ppb (i.e., average NH 3 concentration +1 standard deviation = 8.1+5.2 = 13.3 ppb). The PM 1 pH values in this study are generally similar to those reported by Guo et al. (2015) at the same field site during winter 2012. Our observation that PM 1 is acidic despite the high NH 3 concentrations in this study is consistent with previous studies showing that particle pH has weak sensitivities to wide NH 3 and SO 2− 4 mass concentration ranges due to pH buffering caused by the partitioning of NH 3 between the gas and particle phases Guo et al., 2017c). This weak particle pH sensitivity also explains the small changes in PM 1 pH values (about 10 % lower, Fig. S8) when NH 3 measurements by the SEARCH network denuderbased instrument are used in ISORROPIA II calculations (instead of NH 3 -CIMS measurements). PM 1 pH varied by approximately 1.4 units throughout the day. W i has an average value of 1.6 ± 1.7 µg m −3 . PM 1 W i and pH showed similar diurnal profiles, with both peaking in the midmorning and reaching their minima in the midafternoon. These diurnal trends are consistent with those previously reported by Guo et al. (2015) for PM 1 measured during the summer and winter in different parts of the southeastern US. Also shown in Fig. 3b  which peaked in the midafternoon. The W i and H + air maximum / minimum ratios are comparable (6.5 and 5.3, respectively), thus indicating that the diurnal variation in particle pH is driven by both W i and H + air . The average PM 1 pH for this study is about 1 unit higher than that for the SENEX and SOAS campaigns (Table 1) and is likely due to the much higher abundance of NH 3 in this study. The average NH 3 mass concentration in this study is approximately 49 times and 15 times higher than that in the SENEX and SOAS campaigns, respectively. The average PM 1 pH for this study is similar to that for the CalNex campaign even though the average NH 3 mass concentration in this study is only approximately 4 times higher than that in the CalNex campaign (Guo et al., 2017a). This may be due, in part, to PM 1 SO 2− 4 and NO − 3 mass concentrations at Cal-Nex being approximately 2 times and 18 times larger than those of this study, respectively. Aerosol inorganic SO 2− 4 and NO − 3 species are hygroscopic species. The much higher NO − 3 mass concentrations in the CalNex campaign (due, in part, to high NO x emissions) increased particle W i substantially, which diluted H + and raised particle pH, resulting in more gas-to-particle partitioning of NO − 3 , and eventually leading to pH levels similar to those observed in this study. This type of feedback does not happen in the southeastern US, where non-volatile SO 2− 4 dominates the uptake of particle water. It is also possible that the higher RH and lower temperatures during the CalNex campaign (relative to this study) contributed to high particle W i , which diluted H + and raised particle pH levels similar to those observed in this study.
The validity of this study's thermodynamic model predictions is evaluated by comparing the predicted gasparticle partitioning ratios of semi-volatile inorganic compounds (i.e., NO − 3 and NH + 4 ) with measured values (Fig. S9). CIMS HNO 3 and NH 3 data, PILS-IC NO − 3 , and HR-ToF-AMS NH + 4 data are used in this comparison. ε(NO − 3 ) and ε(NH + 4 ) are defined as the particle-phase molar concentration divided by the total molar concentration (gas + particle), i.e., ε(NO − 3 ) = NO − 3 /(HNO 3 + NO − 3 ) and ε(NH + 4 ) = NH + 4 /(NH 3 + NH + 4 ). Predicted NH 3 , NH + 4 , and ε(NH + 4 ) values are generally within 10 % of and are Figure 4. Analytically calculated S curves of ε(NH + 4 ) and ε(NO − 3 ) and ambient data plotted against ISORROPIA-predicted particle pH for this study, SENEX, SOAS, and CalNex. For the ambient data sets, a narrow range of W i (1 to 4 µg m −3 ) and temperature (15 to 25 • C) is selected to be close to the analytical calculation input (i.e., W i = 2.5 µg m −3 and temperature = 20 • C). Similar to Guo et al. (2017a), γ NH + 4 = 1 and γ H + −NO − 3 = γ H + γ NO − 3 = 0.28 are used for the analytically calculated S curves. highly correlated (R 2 = 0.96 to 0.99) with measured values (Fig. S9). While predicted HNO 3 values generally agreed with measurements, substantial scatter can be seen between the predicted and measured values for NO − 3 and ε(NO − 3 ). This scatter can be attributed, at least in part, to uncertainties brought about by the low PM 1 NO − 3 mass concentrations and effects of coarse-mode cations (e.g., Na + , Ca 2+ , K + , and Mg 2+ ) on fine-mode HNO 3 -NO − 3 gas-particle equilibrium (i.e., HNO 3 can partition to both fine and coarse modes, thereby affecting fine-mode NO − 3 concentrations; no such effect occurs for NH 3 -NH + 4 gas-particle equilibrium). In general, the overall good agreement between model predictions and measurements indicated that our assumptions that aerosols are metastable (i.e., aerosols are supersaturated aqueous droplets) with no phase separation for the thermodynamic calculations are reasonable for the conditions of this study and do not affect model predictions.
The molar fractions of NO − 3 and NH + 4 in the particle phase (i.e., ε(NO − 3 ) and ε(NH + 4 )) measured in this study are compared with those measured during the CalNex, SENEX, and SOAS campaigns. Figure 4 shows the measured ε(NO − 3 ) and ε(NH + 4 ) values as a function of their ISORROPIA-predicted particle pH for the various field studies. For each field study, only a subset of the data are chosen for this comparison (1 ≤ W i ≤ 4 µg m −3 and 15 • C ≤ temperature ≤ 25 • C) to reduce the effects of variability in W i and temperature on gas-particle partitioning for comparison with the calculated S (or sigmoidal) curves, which are calculated based on W i = 2.5 µg m −3 and temperature = 20 • C. The S curves for HNO 3 -NO − 3 and NH 3 -NH + 4 partitioning as a function of particle pH are also plotted as solid lines. The S curves are calculated based on the solubility and dissociation of NO − 3 and NH + 4 species in water: where H * HNO 3 and H * NH 3 (mole 2 kg −2 atm −1 ) are equilibrium constants and are the products of the Henry's law constant and the dissociation constant of HNO 3 and NH 3 , respectively; R is the gas constant (8.314 m 3 Pa K −1 mol −1 ); T is temperature (K); and γ i 's are activity coefficients. H *

HNO 3 and H *
NH 3 values at 20 • C are calculated using equations found in Clegg and Brimblecombe (1990) and Clegg et al. (1998), respectively. Activity coefficients predicted by ISORROPIA II are γ H + −NO − 3 = γ H + γ NO − 3 = 0.28, γ H + = 1 and γ NH + 4 = 1. Derivations of the analytically calculated S curves for ε(NO − 3 ) and ε(NH + 4 ) in Eqs. (2) and (3) can be found in Guo et al. (2017a). As shown in Fig. 4, the measured ε(NO − 3 ) and ε(NH + 4 ) values for the four field studies all generally converged on the calculated S curves. The higher particle pH values in this study and the CalNex campaign relative to those for the SENEX and SOAS campaigns resulted in less NH 3 and more HNO 3 partitioned to the particle phase, as predicted by these simple analytical expressions. A similar analysis will be performed for the organic acids in Sect. 3.5.

WSOC and water-soluble organic acids
The time series and average diurnal profiles of WSOC g and WSOC p are shown in Fig. S10. The average WSOC g mass concentration (3.6 ± 2.7 µg C m −3 ) is roughly 4 times higher than that of WSOC p (1.0 ± 0.6 µg C m −3 ). The diurnal profile of WSOC p is somewhat flat, likely due to various organic aerosol sources having different water solubility and diurnal cycles and compensating for each other throughout Atmos. Chem. Phys., 18, 11471-11491, 2018 www.atmos-chem-phys.net/18/11471/2018/ the day (Xu et al., 2015b. In contrast, WSOC g displayed strong diurnal variations. WSOC g increased at 07:30, which coincided with the sharp increase in solar irradiance (Fig. S3). WSOC g decreased at 21:30, approximately 2 h after sunset. Also shown in Fig. S10 are the time series and average diurnal profile of the mass fraction of total WSOC in the particle phase, i.e., F p = WSOC p /(WSOC p + WSOC g ).
The peak F p coincided with the minima of WSOC g at 07:30. The average WSOC g and WSOC p (3.6±2.7 µg C m −3 and 1.0 ± 0.6 µg C m −3 ) are slightly lower than those measured during the SOAS campaign (SOAS WSOC g = 4.9 µg C m −3 and WSOC p = 1.7 µg C m −3 ) . While the di-urnal profiles of WSOC p in both studies are flat, the diurnal profiles of WSOC g measured in the two studies are different. WSOC g measured in the SOAS study decreased at sunset, while WSOC g measured in this study decreased 2 h after sunset. Differences in WSOC g diurnal profiles in the two studies are likely due to differences in emission sources as a result of different sampling periods (SOAS was in early summer and this study was in early fall), land use, and/or land cover. The ratio of WSOC p to OC for this study was estimated at 30 %, but this comparison is imprecise because WSOC p was PM 1 and OC was PM 2.5 (refer to Fig. S11 and Supplement Sect. S2).  Figure 5 shows the time series of particle-and gas-phase concentrations of formic, acetic, oxalic, malonic, succinic, glutaric, and maleic acids. Their diurnal profiles are shown in Fig. 6. Gas-phase measurements of glutaric and maleic acids are not available. Gas-phase measurements of butyric, glycolic, propionic, and valeric acids were also measured during the study and have been presented in Nah et al. (2018), but will not be discussed here since their particle-phase measurements are not available.
Assuming that all the measured organic acids are completely water soluble, 30 % of the WSOC g is comprised of these organic acids (Nah et al., 2018). Formic and acetic acids are the most abundant gas-phase organic acids, with averages of 2.2±1.6 and 1.9±1.3 µg m −3 , respectively. The average carbon mass fraction of WSOC g comprised of formic and acetic acids is 7 % and 13 %, respectively. All the gasphase organic acids displayed strong and consistent diurnal cycles, with higher concentrations being measured during warm and sunny days. Their concentrations start to increase at sunrise (at 07:30), building to a peak between 15:30 and 19:30, then decrease overnight.
Atmos. Chem. Phys., 18, 11471-11491, 2018 www.atmos-chem-phys.net/18/11471/2018/ Nah et al. (2018) previously showed that the measured gasphase organic acids during the study, including oxalic acid, likely have the same or similar sources. Poor correlations between gas-phase organic acid concentrations and those of anthropogenic pollutants (HNO 3 , SO 2 , CO, and O 3 ) indicated that these organic acids are not due to anthropogenic emissions and are likely biogenic in nature. Biogenic emissions of gas-phase organic acids and/or their BVOC precursors are elevated at high temperatures, resulting in higher organic acid concentrations during warm and sunny days. For example, isoprene, which is the dominant BVOC in Yorkville, has a somewhat similar diurnal profile to the organic acids. In addition, the concentration of isoprene is moderately correlated with those of formic and acetic acids (Fig. S10 of Nah et al., 2018), which are known products of isoprene photooxidation. Some of these gas-phase organic acids may also be formed in the particle phase during organic aerosol photochemical aging, with subsequent volatilization into the gas phase. The gas-particle partitioning of organic acids likely depends on thermodynamic conditions, which are controlled by particle pH and W i and meteorological conditions, as will be shown in Sect. 3.5.
The measured particle-phase water-soluble organic acids contributed on average 6 % to the HR-ToF-AMS-measured organic aerosol mass concentration. The average carbon mass fraction of WSOC p comprised of these organic acids is 4 %. Previous studies have shown that particle-phase organic acids found in rural environments are oxidation products of gas-phase aliphatic monocarboxylic acids, which are formed in the photochemical oxidation of biogenic unsaturated fatty acids and other BVOC precursors (Kawamura and Gagosian, 1987;Kawamura and Ikushima, 1993;Kerminen et al., 2000;Kawamura and Bikkina, 2016). These particlephase organic acids can also be produced during the multiphase photochemical aging of ambient organic aerosols (Ervens et al., 2004;Lim et al., 2005;Sorooshian et al., 2007Sorooshian et al., , 2010. Oxalate is the most abundant measured particle-phase water-soluble organic acid anion (contributing on average 26 % to the total particle-phase organic acid mass concentration), with mass concentrations ranging from 0.01 to 0.34 µg m −3 and an average of 0.07 ± 0.05 µg m −3 . Acetate (average of 0.06 ± 0.03 µg m −3 ) and formate (average of 0.05 ± 0.03 µg m −3 ) are the second and third most abundant measured particle-phase water-soluble organic acid anions, respectively. Particle-phase formate, acetate, and maleate showed weak diurnal variations, and may be due, in part, to various emission sources having different diurnal cycles and compensating for each other throughout the day. Particlephase oxalate, malonate, and succinate peaked in the middle to late afternoon, while glutarate generally peaked in the midmorning. This suggests that while the production of these organic acids is photochemically driven, they may have different BVOC precursors and/or different photochemical production pathways. In addition, since oxalic (C 2 ), malonic (C 3 ), succinic (C 4 ), and glutaric (C 5 ) acids belong to the same homologous series of organic diacids, it is possible that the photochemical aging of particle-phase glutaric acid resulted in the formation of its successive homologues via the cleavage of C-C bonds. Hence, organic aerosol photochemical aging may also have contributed to the diurnal profiles of particle-phase oxalate, malonate, succinate, and glutarate.
3.5 Gas-particle partitioning of organic acids The online and simultaneous measurements of gas-and particle-phase organic acid mass concentrations provided the opportunity to study gas-particle partitioning behavior of semi-volatile organic compounds with respect to particle pH, as is more commonly done with semi-volatile inorganic species (see Sect. 3.3). Since formic, acetic, and oxalic acids are the three most abundant measured organic acids present in the gas and particle phases, we focus on the gas-particle partitioning behaviors of these three organic acids. The average molar fractions (±1 standard deviation) of formic, acetic, and oxalic acid in the particle phase (i.e., ε(HCOO − ), ε(CH 3 CO − 2 ), and ε(C 2 O 2− 4 )) are 3.6 ± 3.6 %, 5.8 ± 5.0 %, and 73.7 ± 9.8 %, respectively. The uncertainties of these ratios for formic, acetic, and oxalic acids are 16 %, 16 %, and 17 %, respectively, which are obtained from the propagation of their SF 6 -CIMS and PILS-HPIC measurement uncertainties.

Oxalic acid
To investigate the factors affecting oxalic acid gas-particle partitioning, the equation for the S curve describing the dependence of oxalic acid gas-particle partitioning (i.e., ε(C 2 O 2− 4 ) = C 2 O 2− 4 /(C 2 H 2 O 4 + C 2 O 2− 4 )) on particle pH is derived. As shown in Supplement Sect. S3, the analytically calculated S curve for ε(C 2 O 2− 4 ) can be simplified to where H C 2 H 2 O 4 (mole L −1 atm −1 ) is the Henry's law constant for oxalic acid, K a1 (mole L −1 ) is the first acid dissociation constant for oxalic acid, R is the gas constant (8.314 m 3 Pa K −1 mol −1 ), T is temperature (K), and γ i 's are activity coefficients. We used the web version of AIOM-FAC (http://www.aiomfac.caltech.edu, last access: 6 December 2017) (Zuend et al., , 2011 to compute an average γ C 2 H 2 O 4 value of 0.0492. Since AIOMFAC does not predict γ H + γ C 2 HO − 4 , we assumed that γ H + γ C 2 HO − 4 = γ H + γ NO − Müller (2014), and Saxena and Hildemann (1996) (6.11 × 10 8 mole L −1 atm −1 at 25 • C) and accounted for the effect of temperature using Eq. (19) in Sander (2015). Although K a1 also depends on temperature, we used the K a1 value at 25 • C (5.62 × 10 −2 mole L −1 , Haynes, 2014) for all the oxalic acid S curve calculations since equations that compute K a1 values for pure aqueous oxalic acid solutions at different temperatures are not available in the literature. In addition, the temperatures observed in this study were close to 25 • C (study-average temperature = 23.4 ± 4.0 • C). Different S curves for ε(C 2 O 2− 4 ) are calculated using 1 h average values obtained from the diurnal profiles of temperature and W i (specifically at 00:30, 06:30, and 12:30). The shape of the S curve changes with the time of day due to the diurnal variations in temperature and W i (Fig. S12 and Supplement Sect. S3). The S curves for ε(C 2 O 2− 4 ) are very different from those of other acids, such as ε(NO − 3 ) (shown in Fig. 4b). From the S curves for ε(C 2 O 2− 4 ), which are calculated using conditions in this study, some molar fraction of oxalic acid is always expected to be present in the particle phase, even at low particle pH (i.e., the S curve does not go to zero at low pH). In contrast, HNO 3 is expected to be present primarily in the gas phase at low particle pH (i.e., pH < 1) under similar temperature and W i conditions. This is due primarily to differences in the Henry's law constants for the two acids. H HNO 3 (2.57 × 10 5 mole L −1 atm −1 ) at 23.4 • C is 3 orders of magnitude smaller than H C 2 H 2 O 4 (7.27 × 10 8 mole L −1 atm −1 ) (Clegg and Brimblecombe, 1990;Sander, 2015). This means that some undissociated form of oxalate can be found in the particle phase at any pH, and the molar fraction of this form increases with particle W i (see Fig. S12). Oxalic acid's very high Henry's law constant combined with the W i conditions in this study ensures that some fraction of the organic acid will be in the particle phase regardless of the particle pH. Figure 7 compares the measured ε(C 2 O 2− 4 ) vs. ISORROPIA-predicted PM 1 pH to the analytically calculated S curve(s). The S curve is calculated based on the average temperature and W i from 13 September to 6 October (23.4 ± 4.0 • C and 1.6 ± 1.7 µg m −3 , respectively). We also calculated the "upper" and "lower" bounds of this S curve based on 1 standard deviation from the average temperature and average W i . A temperature = 27.4 • C and W i = 0.5 µg m −3 are used for calculations of the lower bound, while a temperature = 19.4 • C and W i = 3.3 µg m −3 are used for calculations of the upper bound. For the ambient data, a range in W i (0.5 to 4 µg m −3 ) and temperature (15 to 31 • C) is chosen to be close to the analytical calculation. As shown in Fig. 7, the measured ε(C 2 O 2− 4 ) generally converged around the S curve calculated using the average temperature and W i values. Although there is some scatter, the measured ratios are mostly within the upper and lower bounds of the S curve. Figure 7. Analytically calculated S curve of ε(C 2 O 2− 4 ) and ambient data from 13 September to 6 October 2016 plotted against ISORROPIA-predicted particle pH. For the ambient data, a range in W i (0.5 to 4 µg m −3 ) and temperature (15 to 31 • C) is chosen to be close to the analytically calculated outputs. For the analytically calculated S curves, we used γ C 2 H 2 O 4 = 0.0492 (AIOMFAC predicted). We also assumed that γ H + γ C 2 HO − The black line is the S curve calculated using the selected time period's average temperature (23.4±4.0 • C) and W i (1.6±1.7 µg m −3 ). The grey lines are S curves calculated using 1 standard deviation from the average temperature and W i (i.e., temperature = 27.4 • C and W i = 0.5 µg m −3 for dotted grey line; temperature = 19.4 • C and W i = 3.3 µg m −3 for solid grey line).
Since the measured ε(C 2 O 2− 4 ) values are in general agreement with the analytically calculated S curve (Fig. 7), we can use the S curve to understand qualitatively how high NH 3 events at the site affect oxalic acid gas-particle partitioning. Here we define high NH 3 events as periods when the NH 3 concentration was higher than 13.3 ppb (which is the average NH 3 concentration +1 standard deviation). As discussed in Sect. 3.3, the PM 1 pH during high NH 3 events is 2.5 ± 0.6, which is slightly higher than the average PM 1 pH of 2.2±0.6. Based on the S curve calculated using the average temperature and W i values, ε(C 2 O 2− 4 ) increases from 81 % to 89 % when particle pH increases from 2.2 to 2.5. While this result indicates that high NH 3 concentrations can raise the particle pH sufficiently such that it can promote gas-to-particle partitioning of oxalic acid, this is not always the case. Specifically, increasing the particle pH from −2 (or lower) to 1 will not result in a significant increase in ε(C 2 O 2− 4 ). Therefore, whether or not particle pH, and consequently oxalic acid gas-particle partitioning, is sensitive to NH 3 concentration depends strongly on particle pH values.
We also examined how well the analytically calculated S curve for ε(C 2 O 2− 4 ) captures diurnal variations in the measured ε(C 2 O 2− 4 ). The ambient data are divided into two 12 h sets (08:00 to 19:59 and 20:00 to 07:59) based on the diurnal profile of solar irradiance. Two S curves and their cor- A number of inferences can be drawn from the overall good agreement between the predicted and measured molar fractions of oxalic acid in the particle phase in Figs. 7 and S13. Our assumptions regarding the activity coefficients, Henry's law constant, and acid dissociation constants used in the S curve calculations of ε(C 2 O 2− 4 ) are reasonable for the conditions of this study (or are at least self-consistent). Analytically calculated S curves are a simple way of exploring how the gas-particle partitioning of semi-volatile inorganic and organic compounds in the atmosphere are affected by the compound's physicochemical properties (e.g., Henry's law constants and acid dissociation constants), temperature, W i , and pH. Overall, these results indicate that particle-phase oxalate is in equilibrium with gas-phase oxalic acid and that particle pH can influence particle-phase oxalate concentrations. It also showed that particle-phase oxalate can be found over a broad pH range and that the presence of oxalate does not necessarily provide insights into the particle pH. Because of its high Henry's law constant, particle-phase oxalate can be found in aerosols even at extremely low pH values (i.e., the flat region in Fig. 7), although our data cannot be used to test this since ambient particle pH values in this study are too high.
S curves for ε(HCOO − ) and ε(CH 3 CO − 2 ) calculated based on temperature = 23.4 • C and W i = 1.6 µg m −3 can be seen in Fig. 8. Practically no formic or acetic acids are predicted to partition to the particle phase (relative to oxalic acid) for the range of PM 1 pH calculated in this study. This is due to significant differences in the Henry's law constants and acid dissociation constants for the three organic acids. H HCOOH and H CH 3 CO 2 H (9540 and 5370 mole L −1 atm −1 , respectively) at 23.4 • C are substantially smaller than H C 2 H 2 O 4 (7.27 × 10 8 mole L −1 atm −1 ) (Sander, 2015). The K a1 values for formic and acetic acids (1.78 × 10 −4 and 1.75 × 10 −5 mole L −1 , respectively) are also considerably smaller than the K a1 value for oxalic acid (5.62 × 10 −2 mole L −1 ) (Haynes, 2014). Note that H HNO 3 is between that of H C 2 H 2 O 4 and those of H HCOOH and H CH 3 CO 2 H (compare Fig. 4b with Figs. 7 and 8).
As shown in Fig. 8, higher-than-expected levels of formate and acetate are observed in the particle phase. This has also been reported in previous studies (Liu et al., 2012). Laboratory tests showed that the disagreement cannot be explained by positive biases in the particle-phase formate and acetate PILS-HPIC measurements resulting from less than 100 % gas removal by the carbon denuder. The measured denuder efficiency for formic acid was ≥ 99.97 % (Supplement Sect. S4). The possibility that formic and acetic acid dimers in the aqueous phase (Schrier et al., 1964;Gilson et al., 1997;Chen et al., 2008) may result in higher-than-predicted molar fractions of formate and acetate in the particle phase was explored but also could not explain the observed gas-particle partitioning of these acids (Supplement Sect. S5). The disagreement could be due to incorrect Henry's law constants for formic and acetic acids. However, the Henry's law constants for formic and acetic acid would have to be ∼ 10 4 times and ∼ 3 × 10 5 times larger than their literature values, respectively, in order for their S curves to match our measured molar fractions of formic and acetic acid in the particle phase. In addition, formic and acetic acids may not be inter- Figure 8. Analytically calculated S curves of ε(HCOO − ) and ε(CH 3 CO − 2 ) (solid black lines) and ambient data from 13 September to 6 October 2016 plotted against ISORROPIA-predicted particle pH (shown in a and b, respectively). For the ambient data, a narrow range in W i (0.5 to 4 µg m −3 ) and RH (20 % to 90 %) is chosen to be close to the analytically calculated outputs. For the analytically calculated S curves, we used γ HCOOH = 0.334 and γ CH 3 COOH = 2.150 (AIOMFAC predicted). We also assumed that γ H + γ HCOO − = γ H + γ CH 3 COO − = γ H + γ NO − 3 and used the ISORROPIA-predicted γ H + −NO − 3 = γ H + γ NO − 3 = 0.265. The black lines are S curves calculated using the selected time period's average temperature (23.4 ± 4.0 • C) and W i (1.6 ± 1.7 µg m −3 ). The grey lines are S curves calculated using 1 standard deviation from the average temperature and W i (i.e., temperature = 27.4 • C and W i = 0.5 µg m −3 for dotted grey line; temperature = 19.4 • C and W i = 3.3 µg m −3 for solid grey line).
nally mixed with most of the other PM 1 aerosol components (e.g., SO 2− 4 , NO − 3 , NH + 4 , C 2 O 2− 4 ) and thus are not associated with acidic aerosols, as assumed above. They may instead be associated with aerosols largely composed of non-volatile cations and have a pH closer to neutral. More research is needed to explain this disagreement.

Summary
Gas-and particle-phase measurements were conducted in Yorkville, Georgia (a rural field site), during fall 2016. The goal of the field study was to understand how NH 3 affects particle acidity and consequently SOA formation through the gas-particle partitioning of semi-volatile inorganic and organic compounds. Since it is a rural site surrounded by forest, agricultural land, and CAFOs, this study provided an opportunity for ambient observations in an area impacted by high local emissions of BVOCs and NH 3 . NH 3 concentrations measured by the NH 3 -CIMS ranged from 0.7 to 39.0 ppb (average 8.1 ± 5.2 ppb), which were substantially higher than typical levels in the southeastern US. PM 1 inorganic chemical composition, gas-phase HNO 3 and NH 3 concentrations, temperature, and RH were used as model inputs in the ISORROPIA II thermodynamic model to calculate PM 1 pH and W i . PM 1 pH ranged from 0.9 to 3.8, with an average pH of 2.2 ± 0.6. The measured and predicted HNO 3 -NO − 3 and NH 3 -NH + 4 gas-particle partitioning ratios were in good agreement. The measured gasphase organic acids were estimated to contribute 30 % of the overall WSOC g on a carbon mass basis, whereas measured particle-phase organic acids comprised 6 % of the total organic aerosol mass concentration and 4 % of the overall WSOC p on a carbon mass basis. Formic and acetic acids were the most abundant gas-phase organic acids, with averages of 2.2 ± 1.6 and 1.9 ± 1.3 µg m −3 , respectively. Oxalate was the most abundant particle-phase water-soluble organic acid anion, with an average of 0.07 ± 0.05 µg m −3 . Measured oxalic acid gas-particle partitioning ratios generally agreed with analytical predictions, which were based on oxalic acid's physicochemical properties (specifically, its Henry's law constants, acid dissociation constants, and activity coefficients), temperature, W i , and particle pH. The partitioning of oxalic acid to the particle phase is highly sensitive to temperature and W i . In contrast, the partitioning of formic and acetic acids to the particle phase was higher than predicted for reasons currently unknown.
Although past air regulations have resulted in decreased sulfate, nitrate, and ammonium aerosol mass concentrations across the US, our study suggests that the current limited regulation of NH 3 emissions may result in some increase in the organic aerosol mass concentration due to increased gas-toparticle partitioning of some organic acids. However, in this study, the effect was small since the organic acids comprised a small fraction of the overall organic aerosol mass.