The changing role of organic nitrates in the removal and transport of NOx

A better understanding of the chemistry of nitrogen oxides (NOx) is crucial to effectively reducing air pollution and predicting future air quality. The response of NOx lifetime to perturbations in emissions or in the climate system is set in large part by whether NOx loss occurs primarily by the direct formation of HNO3 or through the formation of alkyl and multifunctional nitrates (RONO2). Using 15 years of detailed in situ observations, we show that in the summer daytime continental boundary layer the relative importance of these two pathways can be well approximated by the relative likelihood that OH will react with NO2 or instead with a volatile organic compound (VOC). Over the past decades, changes in anthropogenic emissions of both NOx and VOCs have led to a significant increase in the overall importance of RONO2 chemistry to NOx loss. We find that this shift is associated with a decreased effectiveness of NOx emissions reductions on ozone production in polluted areas and increased transport of NOx from source to receptor regions. This change in chemistry, combined with changes in the spatial pattern of NOx emissions, is observed to be leading to a flatter distribution of NO2 across the United States, potentially transforming ozone air pollution from a local issue into a regional one.


Introduction
Nitrogen oxides (NO x ≡ NO+NO 2 ) play a central role in the formation of toxic air pollutants including O 3 and secondary aerosols.More broadly, NO x chemistry controls the rates and pathways of atmospheric oxidation by determining the concentration of the three most important tropospheric oxidants: OH, O 3 , and NO 3 .NO x emissions also directly contribute to nitrogen deposition in sensitive ecosystems (Fowler et al., 2013).Due to its harmful effects to the environment and human health, NO x has been the target of emissions control strategies since the 1970s, causing anthropogenic NO x emissions in the United States to have decreased by a factor of 2 or more over the past 30 years (United States Environmental Protection Agency, 2018).Understanding the consequences of these past changes and predicting the results of future emissions reductions on the atmosphere requires a quantitative description of feedbacks between NO x concentrations and NO x chemistry.
After emission to the atmosphere, removal of NO x occurs through two primary pathways: conversion to HNO 3 and conversion to alkyl and multifunctional nitrates (RONO 2 ).Once formed, HNO 3 is nearly chemically inert in the troposphere, with a lifetime to reaction or photolysis of over 50 h.HNO 3 is therefore removed almost entirely by wet and dry deposition.RONO 2 represents a class of diverse molecules, with atmospheric lifetimes ranging from hours to days depending on the properties of the organic backbone (R group).The loss of RONO 2 is divided among reactions that release NO x from the R group and recycle it back to the atmosphere, reactions that result in heterogeneous hydrolysis to form HNO 3 , and direct deposition.The later two pathways permanently remove NO x from the atmosphere (Nguyen et al., 2015;Romer et al., 2016;Fisher et al., 2016).Other NO x oxidation products, such as peroxy acetyl nitrate (PAN) or HONO, can play an important role in the transport and redistribution of NO x but do not generally lead to permanent NO x removal.
Historically, direct HNO 3 production was thought to be the only important NO x loss pathway, with RONO 2 chemistry playing at most a minor role.However, several studies have shown that the formation rate of RONO 2 in cities or forested regions can be competitive with or greater than the direct production rate of nitric acid (Rosen et al., 2004;Farmer et al., 2011;Browne et al., 2013;Romer et al., 2016;Sobanski et al., 2017).
The relative importance of HNO 3 and RONO 2 production is an important factor in setting the lifetime of NO x (Romer et al., 2016), and it affects the response of NO x loss to temperature (Romer et al., 2018).Due to their different production pathways, the relative importance of HNO 3 and RONO 2 production also controls how NO x loss and ozone production are affected by changes to emissions of NO x or volatile organic compounds (VOCs).By terminating the radical chain reactions, the formation of RONO 2 serves to suppress ozone formation in polluted areas (Perring et al., 2010;Farmer et al., 2011;Edwards et al., 2013;Lee et al., 2014).Several studies have also shown that RONO 2 can efficiently partition into aerosols, potentially explaining a large portion of secondary organic aerosol in a wide range of environments (Rollins et al., 2012;Pye et al., 2015;Xu et al., 2015b;Lee et al., 2016).
Multiple previous studies have used chemical transport models to investigate how the relative production of RONO 2 and HNO 3 varies in different environments.Browne and Cohen (2012) modeled NO x loss over the Canadian boreal forest using WRF-Chem and Fisher et al. (2016) and Zare et al. (2018) studied NO x loss in the southeast United States using GEOS-Chem and WRF-Chem, respectively.These studies agree that in rural and forested areas with lower NO x emissions and higher biogenic VOC emissions, RONO 2 chemistry is often the largest sink of NO x .
However, these studies diverge in their conclusions about the overall importance of RONO 2 chemistry as a NO x sink and how it is likely to change in the future.In a WRF-Chem simulation identical to those described in Zare et al. (2018), RONO 2 chemistry is found to be 60 % or more of the total NO x loss across broad swathes of the southeast United States (Fig. 1), while Fisher et al. (2016) found RONO 2 production to be concentrated in rather small sections of the southeast.Furthermore, Fisher et al. (2016) suggested that the contribution of RONO 2 chemistry to NO x loss across the region is unlikely to change significantly in the future due to the spatial segregation of NO x and VOC emissions.On the other hand, Zare et al. (2018) and Browne and Cohen (2012) suggested that the contribution of RONO 2 chemistry to NO x loss was likely to grow significantly if anthropogenic NO x emissions decreased across the United States.
Here we use in situ observations from a collection of 13 different field deployments to investigate how the relative daytime production of RONO 2 and HNO 3 varies across the United States and how this fraction may change in the future.We show that the relative production of RONO 2 and HNO 3 Figure 1.Average (24 h) fraction of total NO x loss via RONO 2 chemistry over the southeast United States in summer 2013 simulated using the RACM2_Berkeley2 mechanism in WRF-Chem (Zare et al., 2018).can be well described by the relative OH reactivity of NO 2 and of the combined VOC mixture.As both anthropogenic NO x and anthropogenic VOC emissions have decreased substantially in the United States over the past 20 years, the relative role of these two pathways has shifted as well.While the shift has generally been towards an increasing role for RONO 2 chemistry, the shift has been smallest in large cities and largest in the transitional regime around them.Combined with changing emission patterns of NO x , the shift in NO x chemistry is leading to a flatter distribution of NO x across the continental United States.
2 NO x chemistry and production of RONO 2 and HNO 3 NO x is emitted to the atmosphere as NO from a range of anthropogenic and biogenic sources, including motor vehicles, power plants, lightning, fires, and soil bacteria.In the daytime, NO interconverts with NO 2 on a timescale of minutes through Reactions (R1-R2), forming the chemical family NO x .When NO x is combined with VOCs and hydrogen oxides (HO x ), a set of linked radical chain reactions is formed (Reactions R3-R6).As part of these reactions, two molecules of NO are oxidized to NO 2 , leading to the net production of O 3 through Reaction (R2).
The reactions that propagate the catalytic cycle occur at the same time as reactions that remove NO x from the atmosphere, terminating the cycle.Direct HNO 3 production occurs through the association of OH with NO 2 (Reaction R7).RONO 2 compounds are produced as a minor channel of the RO 2 + NO reaction.While the RO 2 + NO reaction typically produces NO 2 and leads to the production of ozone (Reaction R4b), for a small fraction of the time these two radicals will instead associate to form an organic nitrate (Reaction R4a).
The branching ratio k R4b /(k R4a + k R4b ) is designated α and is determined by the nature of the R group as well as the temperature and pressure.Longer carbon backbones and lower temperatures increase α, while lower pressures and oxygenated functional groups decrease it (Wennberg et al., 2018).Typical values of α in the summertime continental boundary layer range from near 0 for small hydrocarbons and highly oxygenated compounds to over 0.20 for large alkanes and alkenes (Perring et al., 2013).
The total rate of RONO 2 production can be calculated from the properties of individual VOCs measured in the atmosphere via Eq.(1).In Eq. (1), Y RO 2i represents the yield of RO 2 radicals from VOC oxidation and f NO i represents the fraction of those RO 2 radicals that react with NO instead of reacting with HO 2 or undergoing unimolecular isomerization (e.g., Teng et al., 2017).f NO i is close to 1 under polluted or moderately polluted conditions but decreases as the concentration of NO x decreases.

P (RONO
If the contributions from individual VOCs are summed and averaged, the total production of RONO 2 can also be calculated from the effective behavior of the VOC mixture via Eq.( 2), where VOCR is the sum of all measured VOC concentrations weighted by their reaction rate with OH.
In a similar fashion, the production of HNO 3 can be calculated via Eq.( 3), where NO2R is the NO 2 reactivity, or the concentration of NO 2 multiplied by k OH+NO 2 .At 298 K and 1 atm, 10 ppb of NO 2 is equivalent to an NO2R of 2.3 s −1 .

P (HNO
Total NO x loss is the sum of the conversion to HNO 3 and conversion to RONO 2 .The fraction of NO x loss via RONO 2 production can be expressed analytically as Eq. ( 4).
The relative production of RONO 2 and HNO 3 is seen to be controlled by two factors: the first describing the chemistry of RO 2 radicals (α eff , f NO eff , Y RO 2eff ) and the second the ratio of NO2R to VOCR, which describes whether OH is more likely to react with a VOC or with NO 2 .Because Eq. ( 4) concerns fractional loss of NO x , the concentration of OH, which affects RONO 2 and HNO 3 production equally, does not appear in the result.
We show below that in the summertime continental boundary layer, the terms describing RO 2 radical chemistry vary significantly less than the NO2R/VOCR ratio, allowing the relative importance of RONO 2 and HNO 3 chemistry to be roughly estimated from only a single ratio.
3 Observed contributions of HNO 3 and RONO 2 chemistry to NO x loss

Daytime chemistry
Relative RONO 2 and HNO 3 production rates were calculated for 13 separate campaign deployments in the Northern Hemisphere over the past 20 years.Campaigns were selected that included measurements of NO x , HNO 3 , O 3 , HCHO, a wide range of VOCs, and total organic nitrates ( RONO 2 ).Although they do not include measurements of RONO 2 , ITCT2k2 and CALNEX-P3 were also included to provide a pair of measurements of VOCs and NO x in the same geographic location separated in time.A list of all campaigns used in this study is given in Table 1.Where available, measurements of OH and HO 2 were used to directly calculate RO 2 formation and loss.When these radicals were not available, OH and HO 2 radical concentrations were also calculated iteratively based on the total rate of HO x radical production by O 3 photolysis, HCHO photolysis, and alkene ozonolysis.When HONO was measured, HONO photolysis was also included as an OH source.In a small fraction of cases (3 % of all data points), NO measurements were not available and NO concentrations were calculated based on the concentrations of O 3 and NO 2 .Details of the radical modeling, including the equations used to calculate the production and loss of these radicals, are given in Appendix A.
Although these field campaigns do not constitute a random sample of the atmosphere, the combined dataset provides an excellent survey of atmospheric chemistry over a wide range of conditions.The combined dataset includes nearly 8000 data points for which fractional NO x loss can be calculated, spanning nearly 3 orders of magnitude in the ratio of NO2R to VOCR with no significant gaps (Fig. 2).(red line in Fig. 3a).This functional form corresponds to a linear relationship between P (RONO 2 )/P (HNO 3 ) and NO2R/VOCR on a loglog scale.If m is fixed to 1, then this form also corresponds to the expected behavior if the VOC mixture did not change between environments, and so all parameters other than NO2R/VOCR remained constant (gray line in Fig. 3a).
The calculated increase in fractional NO x loss via RONO 2 chemistry as NO2R/VOCR decreases is matched by an increase in the observed ratio of RONO 2 to the sum of RONO 2 and HNO 3 (Fig. 3b).However, the increase in fractional concentrations as NO2R/VOCR decreases is much less than the increase in fractional production.At low NO2R/VOCR ratios, the dominant RONO 2 species are typically short lived and can undergo heterogeneous hydrolysis to produce HNO 3 (e.g., Browne et al., 2013).This indirect source of HNO 3 can be the greatest source of HNO 3 in forested environments, and it leads to the relatively weak dependence of fractional concentration on NO2R/VOCR.
While the fraction of NO x loss occurring via RONO 2 chemistry can be well predicted from just the NO2R/VOCR ratio, the observations exhibit a sharper transition from HNO 3 -dominated to RONO 2 -dominated NO x loss than would be expected if the VOC mixture remained constant.This effect can be explained by variation in Y RO 2eff , α eff , and f NO eff as NO2R/VOCR changes.The behavior of these three parameters is shown in Fig. 4. As NO2R/VOCR decreases, f NO eff consistently decreases from 0.8 to 0.2, due almost entirely to the decrease in NO x concentrations.In contrast, both Y RO 2eff and α eff are larger in areas with low NO2R/VOCR ratios, due to changes in the VOC mixture between environments.In areas where NO2R/VOCR is high, many of the predominant VOCs, including CO, HCHO, and aromatics, either do not produce RO 2 radicals when oxidized by OH or produce RO 2 radicals that do not efficiently produce organic nitrates, leading to the relatively low values of Y RO 2eff and α eff .In areas with low NO2R/VOCR ratios, the VOC mixture is often dominated by biogenic alkenes such as isoprene and monoterpenes that efficiently produce organic nitrates, leading to higher values of both Y RO 2eff and α eff .However, although variation in these parameters can help explain some of the observed behavior of fractional NO x loss, the overall variation is much smaller than the variation of the NO2R/VOCR ratio.Each of the three parameters varies by a factor of 4 or less, while the NO2R/VOCR ratio varies by a factor of 1000.The conclusion that variation in VOC parameters is small compared to the variation in the NO2R/VOCR ratio does not hold outside of the summertime continental boundary layer.In the remote marine boundary layer or in the upper troposphere, α eff is extremely low, as the dominant VOCs produce alkyl nitrates at yields of 0.01 or less (Mao et al., 2009;Perring et al., 2013).Under these conditions, HNO 3 dominates NO x loss even when NO2R/VOCR is less than 3 × 10 −2 .
The trend calculated from the in situ observations matches that found in model simulations: in areas with high ratios of NO2R to VOCR, HNO 3 is the dominant NO x sink, but as concentrations of NO x decrease and concentrations of VOCs increase, the opposite is true.The combined in situ observations show that the importance of RONO 2 chemistry to NO x loss is a nonlinear function of the NO2R/VOCR ratio, leading to a sharp transition between the HNO 3 -dominated and RONO 2 -dominated regimes.The sharp transition suggests there is a strong gradient in chemical NO x loss between urban and rural areas, especially in areas with significant biogenic VOC emissions.Furthermore, the sharp transition indicates that some regions may quickly shift from HNO 3dominated to RONO 2 -dominated regimes if NO2R/VOCR decreases.

Nighttime chemistry
While the primary focus of this analysis is on daytime chemistry, a conceptually similar transition may also occur at night.At night, OH concentrations are near zero, and the first step in NO x oxidation is the reaction of NO 2 with O 3 to produce NO 3 .This radical can in turn react either with NO 2 to form N 2 O 5 or with an alkene to form an organic nitrate (Reactions R8-R9).
Finally, N 2 O 5 can either thermally decompose to re-form NO 3 and NO 2 or it can hydrolyze on aerosol surfaces to produce HNO 3 (Reactions R9-R10).
Although the details are different, the nighttime chemical system shares some fundamental similarities with the daytime system: NO x can be lost through the production of RONO 2 or of HNO 3 , and a key step controlling the relative importance of these two sinks is whether an oxidant reacts with NO 2 or with a VOC.These similarities suggest that the relative importance of RONO 2 and HNO 3 as NO x sinks at night may be controlled by the relative reactivities of NO 2 and VOCs towards NO 3 .In areas where NO 3 is more likely to react with NO 2 , HNO 3 production is likely to dominate NO x loss, while the opposite is likely to be true in areas where NO 3 is more likely to react with a VOC.
However, quantitatively estimating the relative fraction of NO x lost through these different pathways is not practical with the combined dataset presented here.There have been relatively few measurements of the nocturnal atmosphere (only 4 of the 13 campaigns in Table 1 include nighttime measurements) and there remain significant uncertainties in the kinetics of nighttime NO x loss.In particular, the overall rate of N 2 O 5 hydrolysis is controlled by the reactive uptake parameter γ and the aerosol surface area, both of which can vary by multiple orders of magnitude (Brown et al., 2009;McDuffie et al., 2018).Variation in the rate of N 2 O 5 hydrolysis may therefore also play a major role in controlling the relative importance of RONO 2 and HNO 3 chemistry to NO x loss at night.While developing a more quantitative understanding of the trends in the chemical mechanisms of nocturnal NO x loss is an important area for future research, the conceptual similarity between the daytime and nighttime regimes suggests that conclusions based on daytime NO x chemistry may also be relevant to the nighttime.

Predicted trends over time
Using the trends in Fig. 3a to understand trends in NO x chemistry over time is only possible if the response to variation across space is equivalent to the response to variation across time.Two direct comparisons of fractional NO x loss in the same environment but at different times are found to fall along the same curve as the variation between campaigns in different locations (Fig. 3), indicating that such a substitution is valid in this analysis.The first case, INTEX-NA and SEAC4RS, sampled the southeast United States (SEUS) in Together, these cases indicate that the trend from Fig. 3a can be used to predict changes in fractional loss if the trend in NO2R/VOCR is known.Over the past decade, satellite measurements of NO 2 show a significant decrease in national NO 2 concentrations, reporting an average decrease of 4.5 %-7 % per year between 2005 and 2011 (Russell et al., 2012).No comparable satellite observations of VOCs exist, but studies in multiple locations have reported a decrease in primary anthropogenic VOC concentrations of 5.5 %-7.5 % per year over 2000-2010(Geddes et al., 2009;;Warneke et al., 2012;Pollack et al., 2013;Pusede et al., 2014).In contrast, biogenic VOC concentrations have been either constant or increasing over that same time period (Geddes et al., 2009;Hidy et al., 2014).Oxygenated VOCs show no major trend with time, although there are few long-term measurements of these species (Geddes et al., 2009;Pusede et al., 2014).
These varied trends in NO x , anthropogenic VOCs, and biogenic VOCs mean that NO2R/VOCR has not changed uniformly over the past decade.Past NO2R/VOCR ratios were calculated by assuming a 6.5 % yr −1 decrease to anthropogenic VOC concentrations, a 5.5 % yr −1 decrease to NO x concentrations, and a 1.5 % yr −1 increase in biogenic VOC concentrations over the past 15 years.We also extrapolate these same trends to estimate NO2R/VOCR 15 years into the future.The calculated NO2R/VOCR ratios are combined with the relationship from Fig. 3 to estimate fractional NO x loss at different times (Fig. 5).Based on these trends, RONO 2 chemistry is seen to have become a larger portion of total NO x loss over the past 15 years, although the change is not evenly distributed.The similar trends in NO x and anthropogenic VOCs cause there to have been little to no change in the regions with the highest NO2R/VOCR ratios (typically large cities).The largest changes are projected to occur in regions with moderate NO2R/VOCR ratios.In these regions, biogenic VOCs often account for a greater fraction of the VOCR, leading to significant decreases in NO2R/VOCR over the past 15 years.In addition, the response of fractional NO x loss to changes in the NO2R/VOCR ratio is magnified in areas where both RONO 2 and HNO 3 chemistry contribute to NO x loss.In this transitional regime, if recent trends continue, the fraction of NO x loss occurring via RONO 2 chemistry could double in the next 15 years.Given the large number of data points sampled in this transition regime (Fig. 2), many regions of the United States are therefore likely to transition from a regime where HNO 3 dominates NO x loss to a mixed or RONO 2 -dominated regime.
5 Impacts of the transition from the HNO 3 to the RONO 2 regime The growing importance of RONO 2 chemistry to NO x loss has several implications for air quality.Most directly, it means that understanding NO x chemistry in all but the most polluted megacities requires including the effects of RONO 2 chemistry.More theoretically, the transition from HNO 3 -to RONO 2 -dominated NO x loss affects how atmospheric chemistry will respond to changes in emissions of NO x and VOCs.
Because RONO 2 species are produced in the same set of reactions that produce O 3 , the fractional loss of NO x via RONO 2 chemistry is directly proportional to the ozone production efficiency (OPE), the ratio of ozone production to NO x loss (Eq.5).
Fundamentally, OPE represents the total amount of ozone produced for each molecule of NO x emitted.When considering ozone pollution on regional scales, OPE is a more appropriate metric than instantaneous ozone production, because it accounts for ozone production both locally and further afield.ratio decreases, OPE increases, reaching an inflection point exactly at the crossover point between the HNO 3 -dominated and RONO 2 -dominated regimes (Fig. 6a-b).For the polluted areas of the country, where HNO 3 is currently the dominant NO x loss pathway, this means that, for example, interventions to improve air quality by reducing NO x emissions will be fighting uphill because every incremental fractional decrease in NO x emissions will be associated with a growing incremental increase in OPE (Fig. 6c).
In addition, as RONO 2 chemistry becomes a more important part of the NO x budget, changes to α eff have an increasing effect on OPE (Fig. 6c).Policy interventions that reduce VOCR but preferentially target high-α compounds (e.g., long-chain alkanes) could inadvertently increase ozone production or OPE (Farmer et al., 2011;Perring et al., 2013).
In addition to the large effects on aerosol yield that changes to NO x and VOC emissions have directly (e.g., Xu et al., 2015a;Pusede et al., 2016), they also affect aerosols by changing the fate of NO x .While both HNO 3 and RONO 2 can form aerosols (Stelson and Seinfeld, 1982;Pye et al., 2015), the properties of the resulting aerosols are likely to differ.Because HNO 3 is a strong acid, a shift towards RONO 2 chemistry is likely to increase aerosol pH.An increase in the role of RONO 2 chemistry will also cause more of the nitrate aerosol to be organic rather than inorganic, potentially affecting the viscosity and morphology of aerosols.Further effects of changing NO x chemistry arise from the distinct fates of RONO 2 and HNO 3 .Many RONO 2 compounds, especially those derived from isoprene, are remarkably reactive in the troposphere, with lifetimes of a few hours or less.A fraction of this RONO 2 loss returns NO x to the atmosphere, allowing RONO 2 production to effectively transport NO x downwind (Romer et al., 2016;Xiong et al., 2016).In contrast, HNO 3 is effectively chemically inert in the troposphere, with a chemical lifetime of 50 h or more.
As a result of the differing chemical fates and lifetimes, transitioning from a HNO 3 -dominated regime to a mixed or RONO 2 -dominated regime has implications for the distribu- tion of NO x on regional to continental scales.If a greater fraction of NO x in polluted or moderately polluted regions is converted into RONO 2 compounds rather than into HNO 3 , then more of the NO x may be re-released downwind, where it can participate in radical chemistry and ozone production.Simulations of RONO 2 chemistry using WRF-Chem and the RACM2_Berkeley2 mechanism (Zare et al., 2018) were used to investigate the RONO 2 lifetime and NO x recycling efficiency of RONO 2 across the southeast United States in summer 2013 (Fig. 7).Across much of the region, RONO 2 is calculated to have a lifetime of roughly 4 h, and the release of NO x from RONO 2 oxidation was between 40 % and 75 % of the instantaneous RONO 2 production rate.Combined, these findings demonstrate a significant role for RONO 2 chemistry in the transport of NO x between regions in the southeast United States.The effects of organic nitrate chemistry on the distribution of NO x is likely to vary greatly across different regions of the United States and should be studied in further detail.
Enhanced NO x transport between source and receptor regions is one aspect of a combined trend that is transforming the spatial distribution of NO x .Over the past decade, NO x emission reductions have been concentrated in the most polluted environments.In these areas, motor vehicles and power plants, targets of emission control strategies, account for almost all of the NO x emissions.In less polluted regions, other sources of NO x , including soil microbes (both in agricultural and nonagricultural regions), off-road vehicles, fires, and lightning, play a greater role in the NO x budget, reducing the effectiveness of typical combustion-related NO x emission controls.In addition, hemispheric background concentrations of NO x and O 3 have risen slightly over the past 2 decades (Cooper et al., 2012).The combination of all three of these trends suggests that the distribution of NO x across the United States is getting flatter over time.This trend matches satellite observations of NO 2 over the continental United States.Figure 8 shows the cumulative frequency distribution of summertime tropospheric NO 2 columns from 2005-2007 and 2015-2017 using the BErkeley High-Resolution (BEHR) v3.0A retrieval (Laughner et al., 2018a) of slantcolumn measurements from ozone monitoring instrument (OMI).Over this time, the highest percentiles of NO 2 concentrations have decreased and the lowest percentiles increased, leading to a significantly narrower distribution of NO 2 concentrations.
In summary, over the past 15 years, decreases in anthropogenic NO x and VOC emissions have led to a significant shift in the mechanisms of daytime NO x loss.Many places where HNO 3 production dominated NO x loss are now mixed or have switched to a situation where the majority of NO x loss occurs through RONO 2 chemistry.If past trends continue, RONO 2 chemistry will grow to become an even more important fraction of NO x chemistry in coming decades.As a result of this combination of changing NO x chemistry, decreasing NO x emissions, and increasing background concentrations, air pollution in the United States may transform from a highly local issue to a more extended regional one.Efforts to control air pollution focused only on local sources are less likely to be effective; future improvements in air quality and attaining the most recent National Ambient Air Quality Standards are likely to require coordinated efforts on regional scales to broadly reduce NO x emissions.indicating that the use of modeled radicals does not significantly affect our results.Furthermore, Fig. A1b-d show that the use of modeled OH or HO 2 concentrations alone does not lead to noticeable changes in P (RONO 2 )/P (HNO 3 ).Use of modeled NO concentrations can cause small but noticeable changes in P (RONO 2 )/P (HNO 3 ), but modeled NO concentrations are used in less than 3 % of all data points used in this analysis (238 out of 7988 data points).

A2 Determination of α
Accurately calculating the RONO 2 production rate requires accurate knowledge of α i for all VOCs.If values of α had been reported for a specific compound from laboratory measurements, the most recent value was applied (Perring et al., 2013;Teng et al., 2015;Rindelaub et al., 2015;Praske et al., 2015;Wennberg et al., 2018).In cases where no reliable laboratory measurements are available, the parameterization for α from Wennberg et al. (2018) was used.In all cases, the temperature and pressure dependencies described in Wennberg et al. (2018) were used to scale the laboratory measurements of α to the conditions of the atmosphere.

Figure 2 .
Figure 2. Number of points in each bin for which the fraction of NO x loss occurring via RONO 2 chemistry could be calculated.

Figure 3 .
Figure 3.Comparison of the relative production rates of RONO 2 and HNO 3 as a function of NO2R/VOCR.Used data points are restricted to the continental summer daytime boundary layer (i.e., over land, less than 1.5 km above ground level, and average temperature > 10 • C).Panel (a) shows the fraction of NO x loss attributable to RONO 2 chemistry, as well as a least-squares fit to the data and the expected behavior if α eff , f NO eff , and Y RO 2eff were constant.Panel (b) shows the ratio of RONO 2 to the sum of HNO 3 and RONO 2 .In each panel, the blue diamonds show the median in each bin, and the vertical lines show the interquartile range.

Figure 4 .
Figure 4. VOC oxidation parameters (α eff , f NO eff , Y RO 2eff ) as a function of NO2R/VOCR.Used data points are restricted to the continental summer daytime boundary layer (i.e., over land, less than 1.5 km above ground level, and average temperature > 10 • C).The line and solid shapes show the median in each bin, and the vertical lines show an example of the interquartile range for each binned parameter.

Figure 5 .
Figure5.Predicted trends in fractional NO x loss over time, calculated from the estimated NO2R/VOCR ratio assuming a constant 6.5 % yr −1 decrease in anthropogenic VOC concentrations, a 5.5 % yr −1 decrease in NO x concentrations, and a 1.5 % yr −1 increase in biogenic VOC concentrations.

Figure 6
uses the theoretic framework described inRomer et al. (2018) to investigate how ozone and NO x chemistry change as a function of NO2R/VOCR.As the NO2R/VOCR

Figure 6 .
Figure 6.Theoretical picture of NO x and O 3 chemistry, calculated using variable NO x concentrations and fixed VOCR, P (HO x ), and α eff .Panel (a) shows how P (O 3 ) and OPE change as NO x changes; panel (b) shows how the fractional NO x loss changes as NO2R/VOCR decreases; panel (c) shows that changes to NO x and VOCR have their greatest effect on OPE not when P (O 3 ) is at a maximum, but at the crossover point between the RONO 2dominated and HNO 3 -dominated regimes.

Figure 7 .
Figure 7. WRF-Chem simulation of RONO 2 chemistry over the southeast United States for summer 2013 as described in Zare et al. (2018).Panel (a) shows the overall lifetime of RONO 2 , defined as the concentration of RONO 2 divided by their chemical loss rate for the daytime boundary layer.Panel (b) shows the average NO x recycling efficiency, defined as the local rate of NO x production from RONO 2 oxidation divided by the rate of RONO 2 production.

Figure 8 .
Figure 8. Cumulative frequency distribution of ozone monitoring instrument tropospheric NO 2 columns over the continental United States using the BErkeley High-Resolution (BEHR) v3.0A retrieval for summer (April-September) in 2005-2007 and 2015-2017.

Figure A1 .
Figure A1.Comparison of P (RONO 2 )/P (HNO 3 ) when measured concentrations of all possible radicals are used (x axis) versus when measured concentrations are replaced by modeled concentrations (y axis).Panel (a) shows the result when modeled concentrations of OH, HO 2 , and NO are all used simultaneously; panels (b-d) show the effect of replacing measured with modeled values one species at a time.