Reactive oxidized nitrogen speciation and partitioning in urban and rural New York State

ABSTRACT

This study examined reactive oxidized nitrogen (NO_y) speciation and partitioning at one urban site, Queens College (QC) in New York City, and one rural site, Pinnacle State Park (PSP) in Addison, New York (NY) from September 2016 to August 2018 and June 2016 to September 2018, respectively. Oxides of nitrogen (NO_x), nitric acid (HNO₃), particle nitrate (pNO₃), peroxy nitrates (PNs), alkyl nitrates (ANs), and NO_y measurements were made at both sites. Across all seasons at QC, the median NO_x, HNO₃, pNO₃, PNs, ANs, and NO_y concentrations were 10.99, 0.49, 0.24, 0.62, 0.94, and 13.95 parts per billion (ppb), respectively. All-season median percent contributions of NO_x, HNO₃, pNO₃, PNs, and ANs to the total NO_y at QC were 77, 4, 2, 5, and 7%, respectively. Therefore, the sum of the individual NO_y species (NO_yi ≈ NO_x + HNO₃ + pNO₃ + PNs + ANs) accounted for 95% of the total NO_y at QC, which was well within measurement uncertainties. At PSP, the median NO_x, HNO₃, pNO₃, PNs, ANs, and NO_y concentrations were 0.65, 0.16, 0.12, 0.13, 0.18, and 1.56 ppb, respectively, over all seasons. The median percent contributions of NO_x, HNO₃, pNO₃, PNs, and ANs to NO_y over all seasons at PSP were 42, 10, 8, 9, and 12%, respectively. NO_yi comprised 81% of NO_y across all seasons at PSP, and deviations from 100% closure were generally within measurement uncertainties. Since both datasets yielded NO_y budget closure results that were either fully or largely explained by the measurement uncertainties, the observed NO_yi is likely representative of ambient NO_y in urban and rural New York. The results have implications for understanding the fate of NO_x emissions and their impact on local and regional air quality in urban and rural New York State.

Implications: Continuous speciated and total reactive oxidized nitrogen (NO_y) measurements were made in urban and rural New York from 2016 to 2018. Different NO_y species have contrasting effects on the chemistry that impacts ozone (O₃) and fine particulate matter (PM_2.5) formation and concentrations. Since O₃ and PM_2.5 are regulated pollutants that have proven difficult to control, the results have implications for current and future air quality policy.

Introduction

Oxides of nitrogen (NO_x = nitric oxide (NO) + nitrogen dioxide (NO₂)) emissions contribute to ozone (O₃) and fine particulate matter (PM_2.5) formation. O₃ and PM_2.5 are pollutants that impact human health and are subject to regulatory action (U.S. EPA, 2010, 2014). In 2016, exposure to fine particles led to approximately 4 million premature deaths worldwide (World Health Organization 2018). NO_x oxidation products (e.g., nitric acid (HNO₃), particle nitrate (pNO₃)) contribute to acid deposition and nutrient loading, which affect ecosystems (U.S. EPA 2019). Different reactive oxidized nitrogen (NO_y) species exert differing effects on the atmospheric chemical reactions that influence O₃ and PM_2.5 concentrations, atmospheric acidity, and atmospheric oxidation potential (e.g., Crutzen 1979; Singh, Brune, and Crawford 2003; Singh et al. 2007). Measurements of NO_y composition are therefore needed to understand and predict the effects of NO_x emissions on air pollutant concentrations and atmospheric deposition. The principal components of NO_y are NO_x, HNO₃, pNO₃, peroxy nitrates (PNs), and alkyl nitrates (ANs). Although other NO_y species, such as dinitrogen pentoxide (N₂O₅), nitryl chloride (ClNO₂), peroxynitric acid (HO₂NO₂), gas-phase nitrate (NO₃), and nitrous acid (HONO) significantly influence nighttime chemistry, they are (1) less abundant, (2) photolyzed rapidly during daytime (e.g., Brown et al. 2006; Singh et al. 2007), and (3) difficult to measure continuously. A general schematic summarizing how many of the above NO_y species interact is shown in Figure 1.

Figure 1. General schematic of oxidized and reduced nitrogen chemistry in the troposphere. Compounds that are most important to daytime and nighttime chemistry are in red and black, respectively

Previous studies have assessed the impact of NO_y compounds on tropospheric chemistry by examining the NO_y budget. Early investigations of NO_y partitioning in Europe (e.g., Harrison et al. 1999; Zellweger et al. 2000), the United States (e.g., Aneja et al. 1996; Parrish et al. 1993), and Canada (e.g., Hayden et al. 2003; Parrish et al. 1993) found that NO_x, HNO₃, and peroxyacetyl nitrate (PAN) contributed approximately 70%–90% to NO_y concentrations. Due to (1) ongoing NO_x reductions throughout the United States (e.g., Hidy and Blanchard 2015) and (2) HNO₃, PAN, and ANs being closely related to the NO_x level (e.g., Browne and Cohen 2012; Singh 1987; Singh and Hanst 1981), other NO_y species–-such as ANs–-may currently constitute a larger fraction of NO_y (Li et al. 2018). Revisiting the NO_y budget to study the present distribution of NO_y species is warranted for two reasons.

First, even though recent studies in the United States (e.g., Benedict et al. 2018; Li et al. 2018; Wild et al. 2016) and China (e.g., Chen, Wang, and Lu 2018; Li et al. 2016; Xue et al. 2011) have conducted NO_y budget analyses, a key limitation associated with these studies was that the authors focused on the NO_y budget during a particular season rather than during all seasons. For example, Benedict et al. (2018) studied the summertime NO_y budget at Rocky Mountain National Park (ROMO) in Colorado during the Front Range Air Pollution and Photochemistry Experiment (FRAPPЀ). From July 16 to August 31, 2014, the authors found that NO, total nitrate (= HNO₃ + pNO₃), ANs, and PAN comprised 7%, 11%, 4%, and 24%, respectively, of NO_y at ROMO (Benedict et al. 2018). Since NO₂ was not measured, the authors suggested that NO₂ and other NO_x oxidation products (NO_z = NO_y – NO_x) accounted for 54% of summertime NO_y at ROMO (Benedict et al. 2018).

Li et al. (2018) used the Geophysical Fluid Dynamics Laboratory (GFDL) AM3 model to explore decadal changes in summertime (July-August) NO_y from 2004 to 2013 over the southeast United States. In summer 2004 (2013), the authors found that the model-predicted NO_y consisted of 45% (42%) HNO₃, 31% (32%) NO_x, 14% (14%) PNs, and 9% (12%) ANs (Li et al. 2018). Furthermore, Li et al. (2018) found that HNO₃, NO_x, PNs, ANs, and NO_y concentrations decreased by 38%, 32%, 34%, 19%, and 34%, respectively, from 2004 to 2013. The authors attributed these decreases to anthropogenic NO_x emissions reductions, and they also concluded that lower hydroxyl radical (OH) concentrations resulting from lower NO_x levels likely increased the lifetime of NO_x and ANs, thereby contributing to the slower ANs reduction rate (Li et al. 2018).

In a wintertime study, Wild et al. (2016) examined NO_y partitioning at the Horsepool measurement site in the Uintah Basin, Utah, from January 15 to February 27, 2012 and January 23 to February 21, 2013. The winter 2012 and winter 2013 field campaigns had no snow cover and snow cover, respectively. The authors found that NO_x was about 80% and 30% of the NO_y concentration during winter 2012 and winter 2013, respectively, even though NO_x concentrations were similar for the two measurement periods. Wild et al. (2016) concluded that the snow cover present during the winter 2013 measurement period led to enhanced photochemistry and more NO_x being oxidized to NO_z, yielding a lower NO_x-to-NO_y ratio. In addition, they found a dramatic difference in nighttime NO_z partitioning during the two campaigns. Namely, N₂O₅ and ClNO₂ were approximately 50% of nighttime NO_z in winter 2012, while they were a small fraction of nighttime NO_z in winter 2013. The authors asserted that the nighttime N₂O₅ lifetime in winter 2013 was a factor of 2.6 shorter than in winter 2012, which likely contributed in part to the differences in NO_z partitioning during the two campaigns (Wild et al. 2016).

Second, although previous studies in Europe (e.g., Harrison et al. 1999), the United States (e.g., Day et al. 2003; Horii et al. 2005; Murphy et al. 2006; Parrish et al. 1993), and Canada (e.g., Hayden et al. 2003; Parrish et al. 1993; Zhang et al. 2008) have made speciated NO_y measurements for multiple seasons, the measurements were (1) not made recently (i.e., within the last 10 years), and (2) primarily made in rural areas. This research gap is noteworthy because it is important to compare recent NO_y speciation in urban versus rural locations using a comprehensive suite of measurements, especially since NO_x concentrations are continuing to decline in the continental United States (e.g., Hidy and Blanchard 2015). Parrish et al. (1993) studied the NO_y partitioning at six rural sites in the eastern United States and Canada from late summer--early fall 1988 by making NO_x, PAN, HNO₃, pNO₃, and NO_y measurements. During daytime hours at multiple lower-elevation sites, Parrish et al. (1993) found that the median contributions of NO_x, PAN, and HNO₃ to the NO_y concentration ranged from approximately 25%–40%, 12%–25%, and 20%–30%, respectively. The authors also reported that the speciated NO_y measurements made at all sites consistently accounted for approximately 90% of NO_y (Parrish et al. 1993).

Hayden et al. (2003) explored the seasonal NO_y budget from February 1998 to October 1999 by making NO_x, HNO₃, pNO₃, PAN, and NO_y measurements at the rural Sutton research site, which is located in southern Quebec at an elevation of 845 m a.s.l. During winter (summer), the authors determined that NO_x, HNO₃, and PAN contributed approximately 70% (35%), 18% (20%–25%), and 9% (13%), respectively, to the NO_y concentration (Hayden et al. 2003). Hayden et al. (2003) also found that an approximately 25%–30% shortfall in the NO_y budget occurred during spring and summer. The authors hypothesized that unmeasured, photochemically produced NO_y species originating from aged air masses may have been an important contributor to this NO_y budget deficit (Hayden et al. 2003).

Day et al. (2003) studied NO_y speciation using data collected from October 2000 to December 2001 near the University of California-Blodgett Forest Research Station (UC-BFRS). At UC-BFRS, the authors found that ANs were typically 10%–20% of the NO_y concentration (Day et al. 2003). Day et al. (2003) suggested that ANs may have been an important contributor to the NO_y budget shortfall found by previous studies at other rural locations (e.g., Aneja et al. 1996; Buhr et al. 1990; Hayden et al. 2003; Williams et al. 1997). In addition, Horii et al. (2005) assessed NO_y partitioning at Harvard Forest, Massachusetts (MA) by making NO_x, HNO₃, PAN, and NO_y measurements from June to November 2000. When the prevailing wind direction was northwesterly, Horii et al. (2005) found that the sum of NO_x, HNO₃, and PAN was comparable to NO_y. Namely, HNO₃ comprised approximately 33% (20%) of NO_y during the summer (fall), with NO_x accounting for most of the remaining NO_y (Horii et al. 2005). In contrast, the authors found an NO_y budget deficit similar in magnitude to that reported by Hayden et al. (2003) when the prevailing wind direction was southwesterly. Specifically, NO_x and HNO₃ accounted for less than 70% of the observed NO_y, while PAN contributed 7%–14% to NO_y during summer (Horii et al. 2005). Ultimately, the authors concluded that unmeasured alkyl and hydroxyalkyl nitrates may have been an important reason for the observed NO_y budget shortfall during southwesterly flow conditions (Horii et al. 2005).

Zhang et al. (2008) studied the NO_y budget in eastern Canada by measuring NO_x, PAN, peroxypropionyl nitrate (PPN), HNO₃, pNO₃, and NO_y during 14 field intensives at eight rural sites from 2001–2005. Collectively, these field campaigns made speciated NO_y measurements during each season (Zhang et al. 2008). The authors distinguished between cold-season, warm-season, and hot-season field campaigns by using the average daytime (9:00–17:00 local time [LT]) air temperature during each campaign (Zhang et al. 2008). Zhang et al. (2008) determined that, on average, NO_x accounted for 50%–80% and 30%–60% of NO_y during the cold seasons and warm/hot seasons, respectively; PAN, HNO₃, and pNO₃ each contributed up to 19% of NO_y; and PPN concentrations were 15%–25% and 10%–15% of PAN during the cold seasons and warm/hot seasons, respectively.

This study addresses the two above research gaps by analyzing the observed NO_y (≈ NO_x + HNO₃ + pNO₃ + PNs + ANs) and NO_z (≈ HNO₃ + pNO₃ + PNs + ANs) speciation and NO_y partitioning in one urban location, Queens College (QC) in Flushing, New York, and one rural location, Pinnacle State Park (PSP) in Addison, NY, from September 2016 to August 2018 and June 2016 to September 2018, respectively. The analysis was conducted at both sites using continuous, hour-averaged measurements of NO_x, HNO₃, total pNO₃, PNs, ANs, and NO_y. Several steps were taken to explore the observed NO_y and NO_z speciation and NO_y partitioning at QC and PSP in more detail. The corresponding analysis approach–-in addition to the instrument and site descriptions–-is presented below.

Instrumentation and site descriptions

The instrumentation used in this study is summarized in Table 1. NO_x measurements were made using Thermo Model 42 C-Trace Level (42 C-TL) NO_x analyzers equipped with photolytic converters that selectively converted NO₂ to NO prior to chemiluminescent detection. NO_y and HNO₃ data from September 2016 to June 2018 and June 2016 to June 2018 at QC and PSP, respectively, were obtained from NO_y analyzers developed by Atmospheric Research & Analysis, Inc. (ARA). These analyzers used a continuous two-channel denuder difference method to determine NO_y and HNO₃ concentrations (Arnold et al. 2007). Due to missing data for these analyzers from July--August 2018 and July--September 2018 at QC and PSP, respectively, NO_y and/or HNO₃ data from collocated analyzers operated by the New York State Department of Environmental Conservation (DEC) and the Atmospheric Sciences Research Center (ASRC) were used for those timeframes. Particle nitrate (pNO₃) data were collected using a continuous three-channel denuder-filter difference approach (Edgerton et al. 2006). PNs and ANs measurements were made via a continuous three-channel thermal-dissociation photolytic difference technique that used two photolytic converters kept at different temperatures–-160 °C and 380 °C for the PNs and ANs channels, respectively-–to distinguish between PNs and ANs (Ninneman et al. 2017). Sampling lines from the converter boxes located on the roof to the instruments located in the climate-controlled shelter were approximately 10–15 m long.

Table 1. Instrument information for the QC and PSP speciated NO_y measurements

Species	Organization(s)^1	Converters	Sampling details	Reference(s)
NO	ASRC	None	Manifold	Ninneman et al. (2017, 2019)
NO2		Photolytic (BLC)^2
NOy	ARA	Molybdenum	None	Arnold et al. (2007)
	ASRC (PSP only)			Ninneman et al. (2017, 2019); Schwab, Spicer, and Demerjian (2009)
	DEC (QC only)			NYS DEC (2015)
HNO3	ARA	Molybdenum	KCl denuder	Arnold et al. (2007)
	ASRC (PSP only)		NaCl denuder	Schwab, Spicer, and Demerjian (2009)
pNO3	ARA	Molybdenum	PM2.5 cyclone plus two denuders (one Na2CO3 and one C6H8O7)^3	Edgerton et al. (2006)
PNs	ARA	Thermal denuder plus BLC	None	Ninneman et al. (2017)
ANs

¹ASRC = Atmospheric Sciences Research Center; ARA = Atmospheric Research & Analysis, Inc.; DEC = New York State Department of Environmental Conservation.

²BLC = Blue Light Converter.

³Na₂CO₃ = sodium carbonate, C₆H₈O₇ = citric acid.

Span and converter efficiency checks for all instrumentation were performed against a standard NO source and either a standard NO₂ or isopropyl nitrate source, respectively. Converter efficiencies for NO_y conversion were tested weekly, and these converters were replaced if the converter efficiency fell below 95%. Each analyzer developed by ARA was subjected to automated zero and span checks at least once a week. The zeros and spans were extracted from the data stream and used to adjust concentrations in the ARA data management system. Zero and span adjustments were linearly interpolated on a 1-minute time base to account for analyzer drift between automated calibrations. In addition, manual calibrations were performed once a month via a two-step procedure. First, 13-minute zero calibrations were performed using ultra-zero grade air from a cylinder. Second, if the analyzer readings during the 13-minute zero calibrations were stable (± 0.1 ppb), 13-minute span checks were completed. If the analyzer readings were more than ± 2% from the expected value during the span checks, the photomultiplier tube high voltage (PMTHV) was adjusted upward or downward accordingly. In contrast, the instruments maintained by DEC and ASRC were calibrated differently. Namely, weekly calibration checks and semimonthly multipoint calibrations were performed on these analyzers at QC. At PSP, weekly calibration checks and quarterly multipoint calibrations were performed on the analyzers operated by ASRC (Schwab, Spicer, and Demerjian 2009). Multipoint calibrations for these instruments at QC and PSP were done more frequently when warranted.

QC (40.74° N, 73.82° W) is approximately 20 m a.s.l., and it is located in the densely-populated neighborhood of Flushing in the Queens borough of New York City. Ninneman et al. (2017) describes the QC site in further detail. Moreover, Schwab, Spicer, and Demerjian (2009), Ninneman et al. (2017), and Ninneman, Demerjian, and Schwab (2019) provide a detailed description of the PSP site, so the site is briefly described here. PSP (42.09° N, 77.21° W) is approximately 504 m a.s.l., and it is located in the village of Addison, which is (1) located in southwestern New York, and (2) has a population of about 1,800 (Ninneman, Demerjian, and Schwab 2019; Ninneman et al. 2017; Schwab, Spicer, and Demerjian 2009). A map showing the locations of QC and PSP within New York State is given by Figure 2.

Figure 2. Map of the northeast United States with the Queens College (QC, black star) and Pinnacle State Park (PSP, blue star) sites indicated (adapted from D-Maps (2018))

Methodology

The observed NO_y and NO_z speciation and NO_y partitioning at QC and PSP were assessed from September 2016 to August 2018 and June 2016 to September 2018, respectively, using hour-averaged data. Speciated NO_y (NO_yi) was calculated as the sum of NO, NO₂, HNO₃, pNO₃, PNs, and ANs. Speciated NO_z (NO_zi) was calculated as the sum of HNO₃, pNO₃, PNs, and ANs. NO_y was also independently measured at QC and PSP. Monthly data completeness for the above parameters was generally 80% or greater at both sites (Tables S1--S2). However, due to instrument malfunctions, there is (1) no PSP NO_x data for August 2017, (2) no PNs and ANs data for QC from December 2017 to January 2018 and April 2018, (3) no continuous pNO₃ data for QC and PSP from May--August 2018 and May--September 2018, respectively, and (4) no NO_y and HNO₃ data from July--August 2018 and July--September 2018 at QC and PSP, respectively, for the analyzers developed by ARA. In addition, the QC NO_x data is missing for January 2017 due to a data logger issue.

The continuous pNO₃ data gap at both sites was accounted for by using daily-average, filter-based pNO₃ measurements made once every three days by the Environmental Protection Agency’s (EPA) Chemical Speciation Network (CSN). Additionally, all available continuous pNO₃ data at QC and PSP were adjusted to match the filter-based pNO₃ measurements since similar data adjustments have been made elsewhere (Edgerton et al. 2006). The procedure for adjusting the continuous pNO₃ data at both sites consisted of five steps. First, the continuous pNO₃ data were compared to the filter-based pNO₃ data on all days from September 2016 to April 2018 at QC and June 2016 to April 2018 at PSP where there were (1) at least 18 hours of continuous pNO₃ data, and (2) available filter-based pNO₃ data. Second, the continuous, hour-averaged pNO₃ data were converted to daily-averaged data to match the filter-based pNO₃ measurements (Edgerton et al. 2006). Third, the data were filtered for times where the ratio of the continuous pNO₃ data to the filter-based pNO₃ data (pNO_3,c/pNO_3,f) was less than 3. This was done to focus on instances where the continuous and filter-based pNO₃ data were in reasonably good agreement. Fourth, the remaining continuous and filter-based pNO₃ data were used to generate scatter plots and linear regression statistics resulting from forcing the y-intercept to zero (Figures S1--S2). In this study, all linear regressions were performed using robust linear regression with bisquare weights in Matrix Laboratory (MATLAB) version R2017a, where extreme outliers in the data were neglected and mild outliers in the data were down-weighted through the use of a tuning constant (MathWorks 2018). Finally, the continuous pNO₃ data were adjusted to match the filter-based pNO₃ data based on the slopes of the linear regressions shown in Figures S1--S2. All continuous pNO₃ data at QC and PSP were adjusted downward by factors of 1.3 and 1.5, respectively (Figures S1--S2).

As discussed previously, the NO_y data gap was addressed by using July--August (July--September) 2018 NO_y data from the NO_y analyzer operated by DEC (ASRC) at QC (PSP). These NO_y measurements were adjusted to match the NO_y measurements from the analyzers developed by ARA due to small systematic differences well within the uncertainties of the combined measurements. These data adjustments were made for two reasons. First, since most of the analysis used measurements made by the ARA-developed analyzers, adjusting the other NO_y measurements increased the consistency of the observed NO_y speciation and partitioning results. Second, scatterplots of the different NO_y measurements at QC (Figure S3) and PSP (Figure S4) over extended timeframes revealed that the NO_y concentrations from the ARA-developed analyzers were consistently 8%–9% greater. Accordingly, the NO_y measurements made by the DEC-operated (ASRC-operated) analyzer at QC (PSP) from July--August (July--September) 2018 were adjusted upward by 8.5% (9%).

Then, the observed NO_y and NO_z speciation and NO_y partitioning at QC and PSP were analyzed by constructing monthly and seasonal stacked bar charts depicting (1) the total median concentration of speciated NO_y and NO_z, and (2) the median percent each individual NO_y species contributed to the NO_y concentration. An analysis of variance (ANOVA) statistical method was used in MATLAB to determine p-values for the monthly and seasonal median NO_x, HNO₃, pNO₃, PNs, and ANs concentrations at both sites. These p-values tested the hypothesis that the median concentrations of the above species varied as a function of month or season. All available continuous data were used to determine NO_y and NO_z speciation at QC and PSP. NO_y partitioning from September 2016 to November 2017 and March 2018 (June 2016 to April 2018) at QC (PSP) was assessed using coincident, continuous data. Due to multiple data gaps at QC from December 2017 to February 2018, April 2018, and June--August 2018, all available continuous data were used to calculate NO_y partitioning during those periods. For May 2018 and May--September 2018 at QC and PSP, respectively, daily-averaged, filter-based pNO₃ measurements made once every three days by the EPA’s CSN and coincident, continuous, and hour-averaged NO_x, HNO₃, PNs, ANs, and NO_y data were used to calculate NO_y partitioning. Therefore, data used to study NO_y partitioning during those timeframes were specified to be “mostly coincident.”

It needs to be noted that (1) NO_y speciation and/or partitioning at QC was not studied for February and November 2017 due to suspect NO_y and NO_x data, respectively, and (2) the QC NO_x data for February 2018 were adjusted by a factor of 0.8 to reflect a change in the NO_x analyzer’s NO coefficient. Additionally, NO_y speciation and partitioning at PSP were not studied for August 2017 due to suspect NO_x data. To explore the differences in the NO_y speciation and partitioning in urban versus rural New York State, the monthly and seasonal median NO_x to NO_y ratio (NO_x/NO_y) was examined from September 2016-August 2018 and June 2016-September 2018 at QC and PSP, respectively. For both sites, discussion will mostly focus on the seasonal median NO_y and NO_z speciation, NO_y partitioning, and NO_x/NO_y ratio results.

The diurnal variability in the seasonal median NO_y and NO_z speciation and NO_y partitioning at QC and PSP were examined to further elucidate the fate of NO_y. These analyses were done for seasons when (1) there was good data completeness and (2) all continuous measurements were made. The seasons that met the two above requirements at QC (PSP) were spring 2017, summer 2017, and fall 2016–2017 (spring 2017, summer 2016–2017, fall 2016–2017, and winter 2016/7–2017/8). For all seasonal analyses in this study, the months corresponding to the spring, summer, fall, and winter seasons were March--May, June--August, September--November, and December--February, respectively.

Due to the number and complexity of the speciated NO_y measurements made at both sites, the measurement uncertainties were estimated to support their validity. Two steps were taken to estimate the measurement uncertainties for QC and PSP. First, the Allan variance–-a method of quantifying the noise within a measurement signal over time (e.g., Stein 1985)-–was used to help estimate the uncertainty of the NO, NO₂, HNO₃, pNO₃, PNs, ANs, and NO_y measurements during stable periods at PSP. Stable periods corresponded to instances when ambient NO_yi and NO_y concentrations changed by less than ± 0.2 ppb for 4 hours or longer within the June 2016 to July 2017 timeframe, and they were only found for PSP since the site is relatively unaffected by fresh, local emissions. Nonetheless, since (1) the analyzers used to measure the individual NO_y species and the total NO_y were identical at QC and PSP, and (2) the maintenance performed on the instrumentation was identical at both sites, the Allan variances found for PSP data were assumed to also apply to QC data. To estimate the Allan variance for each NO_y species at PSP, 5-minute averaged HNO₃, pNO₃, PNs, ANs, and NO_y data were converted to 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, and 60-minute averaged data for the 8–10 stable periods found for each species. The same was done for the 1-minute averaged NO and NO₂ data, except that it was first converted to 5-minute averaged data. Next, the Allan variance was calculated and plotted as a function of the averaging time for all stable periods by using the following equation (Allan 1999):

(1) $\mathrm{A}\mathrm{V}=\left(\frac{1}{2}\right)\ast \overline{\mathrm{\Delta }{\mathrm{y}}^2}$(1)

where AV is the Allan variance, y is the chemical species of interest, ∆y is the change in concentration, represented here by y over the averaging period (i.e., ∆y = y_t+1 – y_t), and $\overline{\mathrm{\Delta }{\mathrm{y}}^2}$ is the average of the square of each change of y over the averaging time period. Based on visual inspection of the initial Allan variance plots, a second Allan variance plot was generated for NO, HNO₃, PNs, ANs, and NO_y that showed only the stable periods that (1) exhibited a consistent relationship between the Allan variance and the averaging time, and (2) consisted of measurements that were likely above the method detection limit (MDL). For clarity, these stable periods were denoted as “selected stable periods.” Then, the percent error was computed using the following equations:

(2) $\mathrm{M}\mathrm{D}\mathrm{L}=\mathrm{k}\ast \mathrm{A}\mathrm{V}$(2)

(3) $\mathrm{\%}\mathrm{E}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r}=\left(\frac{\mathrm{M}\mathrm{D}\mathrm{L}}{\bar{\mathrm{M}}}\right)\ast 100\mathrm{\%}$(3)

where k is the confidence factor, which was defined to be 3 (covering 99% of the variance for normally distributed data), and $\stackrel{-}{M}$ is the average concentration of the NO_y species of interest during the selected stable periods. Ultimately, the percent errors calculated using Equation (3)(3) $\mathrm{\%}\mathrm{E}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r}=\left(\frac{\mathrm{M}\mathrm{D}\mathrm{L}}{\bar{\mathrm{M}}}\right)\ast 100\mathrm{\%}$(3) were averaged for each species to estimate part of each measurement’s uncertainty. An example of how the Allan variance, the MDL, and the percent error were calculated for HNO₃ is given by Figure S5 and Table S3 to further show how the above approach was carried out, and the estimated MDLs for hour-averaged NO_yi and NO_y are given in Table 2.

Table 2. Estimated method detection limit (MDL) and overall uncertainty for each measured NO_y species. The estimated MDLs and overall uncertainties apply to hour-averaged data

Species (abbreviation)	Estimated method detection limit (MDL, ppt)	Estimated overall uncertainty (%)
Nitric oxide (NO)	26	19
Nitrogen dioxide (NO2)	121	20
Nitric acid (HNO3)	31	22
Particle nitrate (pNO3)	32 (QC), 37 (PSP)	28 (QC), 33 (PSP)
Peroxy nitrates (PNs)	32	24
Alkyl nitrates (ANs)	41	18
Reactive oxidized nitrogen (NOy)	74	10

Second, three other factors were considered in the estimation of the speciated and total NO_y measurement uncertainties. These included (1) analyzer-to-analyzer variability, (2) the aforementioned scaling of continuous pNO₃ to adjusted values, and (3) the contribution from the calibration standard and dynamic dilution. Analyzer-to analyzer variability was assumed to be ± 5% for all measurements based on the findings of Sun, Wang, and Zhang (2010), and the uncertainty due to the calibration standard and dynamic dilution was assumed to be ± 2% and ± 4%, respectively, for all measurements. Meanwhile, the average percent error and the average MDL for the continuous pNO₃ data at QC and PSP were adjusted upward by a factor of 1.3 and 1.5, respectively, consistent with the slopes found after taking the linear regression of continuous pNO₃ versus filter-based pNO₃ and forcing the y-intercept to zero (Figures S1-S2). Table 2 shows the estimated overall uncertainties found for the hour-averaged speciated and total NO_y measurements made at QC and PSP after adding the above factors in quadrature.

Results

Queens College

The seasonal and monthly median speciated NO_y (NO_z) concentrations at QC are shown in Figures 3a and S6a (3b and S6b), respectively. The numerical values and the data completeness of the monthly speciated NO_y concentrations are given in Table S1. Median NO_x concentrations at QC showed statistically significant seasonal and monthly variability (p < .01; Tables 3 and S4), and they were about 17–18 ppb during the winter seasons, 8–12 ppb during the spring, summer, and fall seasons, and 10.99 ppb across all seasons (Figure 3a and Table 3). The larger median wintertime NO_x concentration at QC was expected due to the combined effect of lower temperatures, lower planetary boundary layer heights, and anthropogenic NO_x emissions.

Table 3. All-season median NO_y speciation, p-values for the seasonal median NO_x, HNO₃, pNO₃, PNs, and ANs concentrations that were determined using an analysis of variance (ANOVA) statistical method, and all-season median NO_y partitioning at QC and PSP

Parameter	Site	NO	NO2	NOx	HNO3	pNO3	PNs	ANs	NOyi	NOy
Concentration (ppb)	QC	0.97	10.02	10.99	0.49	0.24	0.62	0.94	13.28	13.95
	PSP	0.03	0.62	0.65	0.16	0.12	0.13	0.18	1.24	1.56
P-value	QC	NA	NA	< 0.01	0.37	0.17	0.17	0.01	NA	NA
	PSP	NA	NA	< 0.01	0.91	0.06	0.61	0.14	NA	NA
% Contribution to NOy	QC	6	71	77	4	2	5	7	95	NA
	PSP	2	40	42	10	8	9	12	81	NA

Figure 3. Seasonal median speciated (a) NO_y concentrations and (b) NO_z concentrations at QC from September 2016 to August 2018. Data that are 50%–75% complete (DJF 2017/8 pNO₃; MAM 2018 PNs and ANs) and less than 50% complete (DJF 2017/8 PNs and ANs; MAM 2018 pNO₃; JJA 2018 HNO₃) are denoted by the red and purple hatching, respectively. (c) Seasonal median NO_y partitioning at QC from September 2016 to August 2018. The dashed red line denotes 100% closure of the NO_y budget, and the x-axis is color-coded according to data usage. Reasons for missing data in (a) and (c) are outlined in the Methodology. See Table S1 for monthly numerical values and data completeness, respectively. In all figures, the combined uncertainty of the measurements is given by the black error bars

Seasonal and monthly variability in the NO_z species were also evaluated to gain insight into the underlying chemistry. Figure 3b reveals that the median concentration of NO_zi (≈ HNO₃ + pNO₃ + PNs + ANs) showed little seasonality, with values ranging from approximately 2–3 ppb during all seasons. Seasonal and monthly variability in the HNO₃ concentration at QC were not statistically significant, with p-values of 0.37 and 0.63, respectively (Tables 3 and S4). Still, Figure 3b shows that HNO₃ concentrations were highest during summer, consistent with active photochemical processing of NO_x in warm, sunlit conditions (e.g., Blanchard and Hidy 2018). Although the seasonal and monthly variability in pNO₃ were not statistically significant at the 95% confidence level (Tables 3 and S4), pNO₃ concentrations were noticeably greater during winter (Figure 3b). This was expected since higher pNO₃ concentrations typically occur at lower temperatures (e.g., Seinfeld and Pandis 2016). Monthly median concentrations of PNs exhibited a relatively small range (approximately 0.88 ppb; Table S1), with maximum and minimum values occurring in September 2016 and November 2016, respectively. Nonetheless, the seasonality and the monthly variability of the PNs concentrations shown in Figure 3b and S6b were modestly significant (Tables 3 and S4), though not at the 95% confidence level. ANs concentrations were greatest during fall (Figures 3b and S6b), and the seasonality and month-to-month variability in ANs were statistically significant (p = .01) and modestly significant (p = .13), respectively, at the 95% confidence level (Tables 3 and S4).

Seasonal and monthly median NO_y partitioning at QC are given by Figures 3c and S6c, respectively. The numerical values of the monthly percent contributions of each measured NO_y species to the NO_y concentration are shown in Table S1. Figure 3c and Table 3 demonstrate that NO_yi (≈ NO_x + HNO₃ + pNO₃ + PNs + ANs) constituted most of the ambient NO_y at QC. Over all seasons, the median NO_y concentration was composed of 6% NO, 71% NO₂, 4% HNO₃, 2% pNO₃, 5 % PNs, and 7% ANs (Table 3). Therefore, NO_yi made up 95% of NO_y (Table 3). Any deviations from 100% closure are well within the estimated measurement uncertainties (Figure 3c), which further indicates excellent closure of the NO_y budget at QC. Moreover, Figures 3c, 4, S6c, S7, and Table 3 show that NO_x was the dominant component of NO_y at QC. Specifically, the seasonal median NO_x/NO_y ratio at QC was about 0.8 across all seasons, and it showed little seasonal variability (Figure 4). Overall, the seasonal median NO_x/NO_y ratios at QC shown in Figure 4 are representative of a site that is impacted by fresh emissions.

Figure 4. Seasonal median NO_x/NO_y ratio at QC from September 2016-August 2018. The dashed purple line denotes the median NO_x/NO_y ratio for the full measurement period. Reasons for missing data are outlined in the Methodology, and see Table S1 for monthly numerical values and data completeness, respectively. The combined uncertainty of the NO, NO₂, and NO_y measurements is given by the black error bars

The diurnal variability in the seasonal median NO_y and NO_z speciation at QC for spring 2017, summer 2017, and fall 2016–2017 is shown in Figure 5. Peak NO_x concentrations occurred in the morning and early evening, which was likely due to increased emissions from vehicular traffic (e.g., Roberts-Semple, Song, and Gao 2012). In addition, the diurnal pattern of NO varied seasonally. Figure 5a, c, and e show that daytime NO concentrations declined more slowly during spring compared to summer and fall. This suggests that the decay rate of fresh emissions differed as a function of season. The diurnal variation of NO_zi was generally consistent across the spring, summer, and fall seasons, further indicating that NO_y speciation at QC is dominated by NO_x. However, it is important to note that nighttime concentrations of ANs were approximately 1 ppb greater during fall compared to spring and summer (Figure 5b, d, and f). Since ANs are often permanent NO_x sinks (e.g., Farmer et al. 2011), the higher nighttime ANs concentrations during fall may have implications for next-day O₃ production at QC. Studying the impact of NO_z speciation on O₃ production is an important area of future work that will be discussed in more detail later.

Figure 5. (a) The diurnal variation of the hourly median NO_y speciation at QC during spring 2017. (b) The diurnal variation of the hourly median NO_z speciation at QC during spring 2017. (c) Same as (a), except for summer 2017. (d) Same as (b), except for summer 2017. (e) Same as (a) and (c), except for fall 2016–2017. (f) Same as (b) and (d), except for fall 2016–2017. In all figures, the combined uncertainty of the measurements is given by the black error bars

Pinnacle State Park

Figures 6a (6b) and S8a (S8b) show the seasonal and monthly median speciated NO_y (NO_z) concentrations, respectively, at PSP. The numerical values and the data completeness of the monthly speciated NO_y concentrations are provided in Table S2, and the all-season median speciated NO_y concentrations are shown in Table 3. According to Table 3, median NO_x, HNO₃, pNO₃, PNs, and ANs concentrations at PSP were 0.65, 0.16, 0.12, 0.13, and 0.18 ppb, respectively, across all seasons. In addition, Figure 6b reveals that the median NO_zi concentration at PSP was about two times greater in winter compared to summer. This difference was influenced by relatively significant seasonality (p = .06) and strongly significant monthly variability (p < .01) in pNO₃ (Tables 3 and S4), which had wintertime (summertime) concentrations of approximately 0.35 (0.03) ppb (Figure 6b). It is also noteworthy that the pNO₃ concentration during spring 2018 was comparable to those measured during the winter seasons (Figure 6b). Since higher pNO₃ concentrations are expected at lower temperatures (e.g., Seinfeld and Pandis 2016), this may be partially due to the similar median temperatures of approximately 4, −1, and −3 °C at PSP during spring 2018, winter 2016/7, and winter 2017/8, respectively (not shown). In addition, the seasonal variability in NO_zi indicates that the fraction of NO_x oxidized to NO_z in airmasses arriving at PSP varied across seasons. This is explicitly shown in Figure 7 and S9, which depict the seasonal and monthly median NO_x/NO_y ratio, respectively, at PSP. Figure 7 shows that the median NO_x/NO_y ratio was (1) about 0.3–0.4 in spring and summer, (2) about 0.5–0.6 in fall and winter, and (3) about 0.4 across all seasons. Overall, the lower NO_x/NO_y ratios at PSP are consistent for a rural location that is mainly affected by photochemically aged air masses, rather than fresh emissions.

Figure 6. Seasonal median speciated (a) NO_y and (b) NO_z concentrations at PSP from June 2016 to August 2018. Data that are 50%–75% complete (MAM 2018 pNO₃) are denoted by the red hatching. (c) Seasonal median NO_y partitioning at PSP from June 2016 to August 2018. The dashed red line denotes 100% closure of the NO_y budget, and the x-axis is color-coded according to data usage. Reasons for missing data in (a) and (c) are outlined in the Methodology. See Table S2 for numerical values and data completeness, respectively. In all figures, the combined uncertainty of the measurements is given by the black error bars

Figure 7. Seasonal median NO_x/NO_y ratio at PSP from June 2016 to August 2018. The dashed purple line denotes the median NO_x/NO_y ratio for the full measurement period. Reasons for missing data are outlined in the Methodology, and see Table S2 for numerical values and data completeness, respectively. The combined uncertainty of the NO, NO₂, and NO_y measurements is given by the black error bars

Even though the median NO_x/NO_y ratio varied seasonally and monthly (Figures 7 and S9), the statistical significance of the seasonal and monthly variation in the individual NO_y species differed (Tables 3 and S4). Similar to QC, NO_x showed statistically significant seasonal and monthly variability (p < .01; Tables 3 and S4) at PSP, with higher concentrations during winter (Figures 6a and S8a). Seasonality and monthly variability in HNO₃ were statistically insignificant (Tables 3 and S4), and concentrations deviated little from the all-season median value of 0.16 ppb (Figures 6b and S8b and Table 3). As described above, seasonal variability in pNO₃ was relatively significant (p = .06; Table 3); month-to-month variability was strongly significant (p < .01; Table S4); and concentrations were highest during colder months and seasons (Figures S8b and 6b). PNs showed no clear seasonal or monthly trend (Figures 6b and S8b and Tables 3 and S4). Finally, seasonal and monthly variability in ANs were modestly significant (Tables 3 and S4), with maximum concentrations during winter (Figures 6b and S8b).

Figures 6c and S8c show the seasonal and monthly median NO_y partitioning, respectively, at PSP. The numerical values of the monthly percent contributions of each NO_y species to the NO_y concentration are provided in Table S2. Figure 6c and Table 3 illustrate that NO_yi generally accounted for most of the ambient NO_y, with NO_y consisting of 2% NO, 40% NO₂, 10% HNO₃, 8% pNO₃, 9% PNs, and 12% ANs across all seasons. As a result, NO_yi represented 81% of the total NO_y at PSP (Table 3). However, it is noteworthy that the NO_y budget closure at PSP appeared to vary seasonally (Figure 6c). Namely, Figure 6c indicates that NO_yi made up approximately 80%–95% of NO_y during fall and winter, and approximately 75%–85% of NO_y during spring and summer. Although the observed NO_y budget shortfall at PSP may simply be due to measurement uncertainties (Figures 6c and S8c), potential reasons for the shortfall during spring and summer will be discussed in the next section.

The diurnal pattern in the seasonal median NO_y and NO_z speciation at PSP for spring 2017, summer 2016–2017, fall 2016–2017, and winter 2016/7–2017/8 is given by Figure 8. In spring, summer, and fall, NO_x levels were higher at night due to a combination of lower nighttime mixing heights and daytime photochemistry (Figure 8a, c, and e). Meanwhile, wintertime NO_x levels were higher than those during spring, summer, and fall, with concentrations ranging from 1.1–1.6 ppb for all hours of the day (Figure 8g). The higher wintertime NO_x concentrations were likely due to the combined effect of lower temperatures, lower mixing heights, and reduced photochemical activity (e.g., Schwab, Spicer, and Demerjian 2009). Additionally, the diurnal variation of HNO₃ during spring, summer, and fall indicates that greater photochemical processing and HNO₃ production occurs in airmasses arriving at PSP during those seasons (Figure 8b, d, and f). This is because HNO₃ concentrations are maximized in the early afternoon, particularly during summer (Figure 8d). Figure 8 also shows that nighttime HNO₃ concentrations are non-negligible (i.e., 0.09 ppb or greater) during all seasons. This suggests that nighttime HNO₃ sources, such as heterogeneous hydrolysis of N₂O₅ or the reaction of NO₃ with hydrocarbons, are likely present at PSP. Finally, the higher nighttime pNO₃ concentrations during winter demonstrate how nitrate partitions into the particle phase at low temperatures (Figure 8h; e.g., Seinfeld and Pandis 2016).

Figure 8. (a) The diurnal variation of the hourly median NO_y speciation at PSP during spring 2017. (b) The diurnal variation of the hourly median NO_z speciation at PSP during spring 2017. (c) Same as (a), except for summer 2016–2017. (d) Same as (b), except for summer 2016–2017. (e) Same as (a) and (c), except for fall 2016–2017. (f) Same as (b) and (d), except for fall 2016–2017. (g) Same as (a), (c), and (e), except for winter 2016/7–2017/8. (h) Same as (b), (d), and (f), except for winter 2016/7–2017/8. In all figures, the combined uncertainty of the measurements is given by the black error bars

Discussion and potential implications

The results presented above show that NO_y speciation at QC is (1) NO_x-dominant during all seasons, and (2) comprised of NO_zi that exhibits little seasonal variability (Figure 3). Across all seasons, median NO_x concentrations were 10.99 ppb and median NO_zi concentrations were 2.29 ppb, yielding a NO_yi concentration of 13.28 ppb (Table 3). Overall, the NO_x-dominant nature of NO_y speciation at QC was expected since the site is in an urban location surrounded by several major roadways and large emissions sources (e.g., Ninneman et al. 2017). In addition, the speciated NO_y measurements account for most of the ambient NO_y concentration at QC during all seasons (Figure 3c), with NO_yi constituting 95% of the NO_y concentration (Table 3). As mentioned previously, departures from 100% closure are well-explained by the measurement uncertainties (Figure 3c). This indicates that NO_yi is a good representation of NO_y speciation at QC.

Figure 7 demonstrates that the seasonal median NO_x/NO_y ratio at PSP ranged from approximately 0.3–0.4 (0.5–0.6) during spring and summer (fall and winter). This implies that the air is more oxidized during the warmer seasons. However, Figure 6a and b show that both NO_x and NO_zi concentrations are higher during the winter. Higher wintertime NO_zi concentrations may be due in part to (1) greater availability of NO_x, (2) the formation channels of pNO₃, PNs, and ANs occurring through HNO₃, and/or (3) reduced thermal decomposition of PNs. Additionally, median NO_x and NO_zi concentrations were approximately 0.65 and 0.59 ppb, respectively, at PSP across all seasons (Table 3). This resulted in a median NO_yi concentration of 1.24 ppb that was approximately an order of magnitude less than what was measured at QC (Table 3). The lower NO_yi concentrations may partially explain why NO_yi did not account for as much of NO_y at PSP compared to QC, particularly during spring and summer (Figure 6c). This is because the speciated NO_y concentrations at PSP are closer to the MDL than those at QC (Tables 2 and 3). Still, NO_yi accounted for 81% of NO_y across all seasons at PSP (Table 3), and deviations from 100% closure can mostly be explained by the measurement uncertainties (Figures 6c and S8c).

Nonetheless, Figure 6c reveals that more noticeable NO_y budget shortfalls occasionally occurred at PSP during spring and summer. Consequently, it was hypothesized that these larger NO_y budget shortfalls may have been due in part to unmeasured NO_y species. To test this hypothesis, the hourly median NO_y partitioning was plotted as a function of time of day on a seasonal basis for the seasons when (1) data completeness was high and (2) all the continuous measurements were available (Figure 9). For completeness, the same was done for QC (Figure 10). Figure 9b shows that the NO_y budget shortfall at PSP was maximized during summer from mid-morning to mid-afternoon, which suggests that photochemical processes led to the formation of the unidentified NO_y (e.g., Hayden et al. 2003). For example, it is possible that isoprene oxidation by OH in the presence of NO_x formed various isoprene nitrates (e.g., hydroxyalkyl nitrates; Grossenbacher et al. 2001) that were not fully detected by the instrument that measured PNs and ANs. Figure 9b also indicates that HONO, gas-phase NO₃, N₂O₅, and ClNO₂ were unimportant contributors to the unmeasured NO_y since they are (1) photolyzed rapidly during daytime, and (2) typically maximized at night (e.g., Brown et al. 2006; Singh et al. 2007). If the above species constituted a significant fraction of the unmeasured NO_y at PSP, then a maximum (minimum) in the NO_y budget shortfall would be expected during nighttime (daytime), which is the opposite of what is shown in Figure 9b. Furthermore, since pNO₃ concentrations were less than 0.05 ppb during the warm seasons (Figure 6b and Table S2), using the unadjusted continuous pNO₃ data would have accounted for little of the approximately 20%–30% shortfall observed in spring and summer during daytime (Figure 9a and b). Future studies are needed to determine what led to the NO_y budget shortfalls that sometimes occurred at PSP. Meanwhile, Figure 10 shows that NO_yi is representative of the diurnal pattern of ambient NO_y at QC.

Figure 9. The diurnal variation of the hourly median NO_y partitioning at PSP during selected seasons. The dashed red line denotes 100% closure of the NO_y budget, and the combined uncertainty of the measurements is given by the black error bars

Figure 10. The diurnal variation of the hourly median NO_y partitioning at QC during selected seasons. The dashed red line denotes 100% closure of the NO_y budget, and the combined uncertainty of the measurements is given by the black error bars

Previous work has shown that the seasonality of light ANs follows the seasonality of their dominant parent hydrocarbons (HCs; e.g., Russo et al. 2010). Figures 3b, 6b, S6b, and S8b indicate that concentrations of ANs at QC and PSP were maximized during the fall and winter, respectively. Overall, maximum ANs concentrations during the colder seasons were expected for QC, but not for PSP. This is because a fall or winter maximum in ANs suggests that they are mainly derived from anthropogenic HC emissions rather than biogenic HC emissions. One factor that may have affected the seasonal variability in ANs at PSP is that the site is located in the Marcellus Shale region, where extensive oil and natural gas (O&NG) operations occur. O&NG operations are known to be a source of anthropogenic HCs (e.g., ethane, propane, etc.; Ren et al. 2019; Vinciguerra et al. 2015). As a result, anthropogenic HC emissions from O&NG operations may have partially contributed to the seasonality in ANs at the PSP site, and future studies are needed to either support or refute this hypothesis.

An additional limitation regarding this study also needs to be noted. Namely, the analyzers that measured PNs, ANs, and pNO₃ were not species-specific (Edgerton et al. 2006; Ninneman et al. 2017). By only measuring total PNs, ANs, and pNO₃ concentrations, the concentrations of speciated PNs, ANs, and pNO₃ and their contributions to the NO_y concentration could not be assessed. Hence, future NO_y speciation and partitioning studies at QC and PSP should consider using different measurement techniques to (1) detect individual PN, AN, and pNO₃ species, (2) further elucidate the fate of NO_y, and (3) explore what may have contributed to the occasional NO_y budget shortfalls found at PSP during spring and summer. Examples of such techniques include gas chromatography/electron capture detection (GC/ECD; Flocke et al. 2005; Roberts 2007) or chemical ionization mass spectrometry (CIMS; Beaver et al. 2012; St. Clair et al. 2010).

Despite the limitations and shortcomings described above, the estimated MDLs and uncertainties (Table 2) show that the continuous, speciated NO_y data collected at QC and PSP were sufficiently robust to analyze NO_y and NO_z speciation and NO_y partitioning over multiple years, thereby addressing the research gaps identified in the Introduction. The data can also be compared to selected observational NO_y speciation and partitioning studies conducted previously at different locations. This is shown in Table S5, which summarizes the observed NO_y budget closure found from the early 1990s to the present. However, the combination of (1) few of the previous studies being done recently, (2) differences in measurement methods, measurement suites, and how the results were reported, and (3) the lack of previous work done in urban areas, make direct and detailed comparisons between the NO_y budget closure found by this study versus previous studies difficult (Table S5). Still, a comparison of the PSP NO_y concentration to those reported by Parrish et al. (1993), Aneja et al. (1996), and Hayden et al. (2003) reveals that the NO_y concentration has declined in rural eastern North America over the past 20–30 years (Table S5). This is consistent with reductions in NO_x concentrations and NO_x emissions found over a similar timeframe in the continental United States (e.g., Civerolo et al. 2017; Hidy and Blanchard 2015). Moreover, the extensive speciated NO_y datasets for QC and PSP will be useful for future O₃ assessments, and both datasets can be used in a variety of future studies. For example, future studies can examine the impact of NO_z speciation on the O₃ production regime at QC and PSP using (1) the speciated NO_z data, (2) O₃ data that are collected at both sites, and (3) an air quality model (e.g., the Community Multiscale Air Quality (CMAQ) model). Such future work will be important since previous studies have indicated that rural New York is in a NO_x-sensitive O₃ production regime, while New York City is likely trending toward NO_x sensitivity (e.g., Jin et al. 2017). Thus, it is becoming increasingly important to understand how the different NO_z species affect O₃ production and loss, and these speciated NO_y datasets for QC and PSP will be useful in gaining such an understanding.

Conclusion

This study assessed observed NO_y and NO_z speciation and NO_y partitioning at one urban site, QC in New York City, and one rural site, PSP in Addison, NY, from September 2016-August 2018 and June 2016-September 2018, respectively. All-season median NO_x, HNO₃, pNO₃, PNs, ANs, and NO_y concentrations at QC (PSP) were 10.99 (0.65), 0.49 (0.16), 0.24 (0.12), 0.62 (0.13), 0.94 (0.18), and 13.95 (1.56) ppb, respectively (Table 3). The predominance of NO_x relative to NO_yi at QC was characteristic of a site that is frequently impacted by fresh emissions, while the higher fraction of NO_z at PSP was characteristic of a rural site that is typically affected by photochemically aged air masses. In addition, all-season median percent contributions of the NO_x, HNO₃, pNO₃, PNs, and ANs concentrations to the NO_y concentration at QC (PSP) were 77 (42), 4 (10), 2 (8), 5 (9), and 7 (12) %, respectively (Table 3). Consequently, NO_yi comprised 95% and 81% of the NO_y at QC and PSP, respectively, across all seasons (Table 3). At both sites, departures from 100% NO_y budget closure are either fully or largely explained by the estimated measurement uncertainties. This implies that both datasets were sufficiently robust to evaluate seasonal and monthly NO_y and NO_z speciation and NO_y partitioning. Overall, the results presented here are illustrative of the possibilities still available to us with both of these unique datasets. For example, future work can use the speciated NO_y datasets at QC and PSP to (1) study the effect of NO_z speciation on the O₃ production regime and/or (2) evaluate air quality model performance. Such future research can further improve our understanding of the fate of NO_y and its impact on local and regional air quality in urban and rural New York State.

Data availability

The data used in this paper are available from a web‐accessible data archive that is cited in the References (Schwab et al. 2018). The link to the data archive is http://atmoschem.asrc.cestm.albany.edu.

Acknowledgment

The authors acknowledge the contributions of Mike Christophersen and Santosh Mahat at QC, John Spicer at PSP, and ARA engineers Brad Gingrey and Chad Haker. Brian Crandall produced the finalized ASRC NO_x, NO_y, and HNO₃ data for PSP.

Disclosure statement

The authors declare that they have no conflict of interest.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.

Additional information

Funding

This work was supported by the New York State Energy Research and Development Authority [grant numbers 48971, 58907].

Notes on contributors

Matthew Ninneman

Matthew Ninneman earned his bachelor’s degree from North Carolina State University in 2015. He is a PhD student advised by Drs. James Schwab and Sarah Lu at the State University of New York at Albany. His research is focused on examining ozone and reactive oxidized nitrogen chemistry in the northeast United States.

Joseph Marto

Joseph Marto earned his bachelor’s degree from Skidmore College in 2014. He is a Ph.D. student advised by Dr. James Schwab at the State University of New York at Albany. His research is focused on examining new particle formation and growth events in the northeast United States.

Stephanie Shaw

Stephanie Shaw, Ph.D., is a Technical Executive at the Electric Power Research Institute where she assesses the environmental impacts of fossil and renewable energy generation and storage. Dr. Shaw has more than 20 years of experience in the areas of greenhouse gas and criteria pollutant emissions and air quality impacts from energy, industrial, and natural sources determined through laboratory and field measurements.

Eric Edgerton

Eric Edgerton helped found Atmospheric Research & Analysis, Inc. (ARA) in 1997. ARA is an air quality consulting firm that is focused on high-quality, research-grade air quality monitoring and data analysis. Mr. Edgerton currently serves as the company’s president.

Charles Blanchard

Charles Blanchard is an independent consultant specializing in receptor modeling and the statistical analysis of air quality measurements. He received a Ph.D. and an M.S. in Energy and Resources and an M.A. in Statistics from the University of California, Berkeley.

James Schwab

James Schwab is a Research Professor and Senior Research Associate with the Atmospheric Sciences Research Center at the University at Albany. His research group studies atmospheric chemistry and air pollution, with an emphasis on policy-relevant scientific work. His research team makes continuous and ongoing measurements of air pollution at locations across New York State. His stations in the Adirondacks and Southern Tier have been continuously monitoring for thirty-plus and almost twenty-five years, respectively. He has also been involved in and organized field campaigns and projects to measure chemically important trace gases and aerosols, mainly across New York State.