We examine the implications of new estimates of the anthropogenic and lightning nitrogen oxide (NOx) source for the budget of oxidized nitrogen (NOy) over the United States in summer using a 3-D global chemical transport model (Model of Ozone and Related Tracers-4). As a result of the Environmental Protection Agency (EPA) State Implementation call, power plant NOx emissions over the eastern United States decreased significantly, as reflected by a 23% decrease in summer surface emissions from the 1999 U.S. EPA National Emissions Inventory to our 2004 inventory. We increase the model lightning NOx source over northern midlatitude continents (by a factor of 10) and the fraction emitted into the free troposphere (FT, from 80% to 98%) to better match the recent observation-based estimates. While these NOx source updates improve the simulation of NOx and O3 compared to the Intercontinental Chemical Transport Experiment-North America aircraft observations, a bias in the partitioning between nitric acid (HNO3) and peroxyacetylnitrate (PAN) remains especially above 8 km, suggesting gaps in the current understanding of upper tropospheric processes. We estimate a model NOy export efficiency of 4%−14% to the North Atlantic in the FT, within the range of previous plume-based estimates (3%−20%) and lower than the 30% exported directly from the continental boundary layer. Lightning NOx contributes 24%−43% of the FT NOy export from the U.S. to the North Atlantic and 28%−34% to the NOy wet deposition over the United States, with the ranges reflecting different assumptions. Increasing lightning NOx decreases the fractional contribution of PAN to total NOy export, increases the O3 production in the northern extratropical FT by 33%, and decreases the regional mean ozone production efficiency per unit NOx (OPE) by 30%. If models underestimate the lightning NOx source, they would overestimate the background OPE in the FT and the fractional contribution of PAN to NOy export. Therefore, a model underestimate of lightning NOx would likely lead to an overestimate of the downwind O3 production due to anthropogenic NOx export. Better constraints on the lightning NOx source are required to more confidently assess the impacts of anthropogenic emissions and their changes on air quality over downwind regions.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Nitrogen oxides (NOx = NO + NO2) play a key role in atmospheric chemistry [e.g., Seinfeld and Pandis, 2006]. In the troposphere, NOx contributes to the formation of ozone (O3) and affects the oxidizing capacity of the atmosphere. NOx is mainly released as NO from combustion, lightning, and soil emissions and then is quickly converted to NO2. It can further be oxidized to peroxyacetylnitrate (PAN), nitric acid (HNO3), and other minor oxidation products. NOx and its oxidation products are collectively known as oxidized nitrogen (NOy). Longer-lived NOy, especially PAN and to a lesser extent HNO3, have the potential to act as reservoirs for NOx and contribute to O3 production on a global scale following export from the NOx source region [e.g., Moxim et al., 1996; Liang et al., 1998; Horowitz and Jacob, 1999; Li et al., 2002, 2004; Parrish et al., 2004; Hudman et al., 2004; Penkett et al., 2004; Auvray and Bey, 2005]. Deposition of NOy compounds, mostly in the form of highly soluble HNO3, provides an important source of nutrients for marine, freshwater and terrestrial ecosystems, influencing their productivity and thereby affecting the global carbon cycle [Galloway et al., 2004; Prentice et al., 2001; Vitousek et al., 1997]. Understanding the budget of NOy is thus necessary to assess the environmental impacts of NOx sources.
 The United States is a major source of anthropogenic NOx, accounting for 22% of global anthropogenic emissions in 2000 [Olivier and Berdowski, 2001] and it has been actively controlling its summertime NOx emissions for more than a decade. For example, the Environmental Protection Agency (EPA) NOx State Implementation (SIP) call in 1998 has led to a reduction of NOx emissions from power plants by 50% in the eastern United States between 1999 and 2003, as confirmed with Continuous Emission Monitoring System measurements on power plant stacks [Frost et al., 2006]. Satellite based observations of tropospheric NO2 provide additional evidence of this decline in NOx, showing a consistent trend of 7% yr−1 decrease for 1996–2006 over the eastern United States [van der A et al., 2008]. Stavrakou et al.  further inferred a 4.3% yr−1 decrease of NOx emissions over the Ohio River Valley from 1997 to 2006 from the tropospheric NO2 column.
 The NASA Intercontinental Chemical Transport Experiment-North America (INTEX-NA) [Singh et al., 2006] was part of the International Consortium for Atmospheric Research on Transport and Transformation (ICARTT) field campaign over North America during July–August 2004. Within the North American troposphere, significant differences in NOy partitioning between observations and models as well as among models suggest a need for additional studies with models incorporating new constraints on NOx sources and chemistry provided by the INTEX-NA campaign [Singh et al., 2007]. For example, Hudman et al.  found that the INTEX-NA observations are consistent with a 22% decrease in fossil fuel NOx from 1999 in the U.S. National Emission Inventory and that increasing lightning NOx in the GEOS-Chem model by a factor of 4 over Northern Hemisphere midlatitude continents improved the match between simulated NOy species and observations. Hudman et al.  derive a higher observed dO3/dCO ratio of 0.47 from the North American outflow during summer 2004 than that reported by studies in early 1990s, reflecting a decrease in the NOx/CO emission ratio as well as an increase in the ozone production efficiency per unit NOx. They also found North American NOx emissions during summer 2004 enhanced the hemispheric tropospheric ozone burden by 12.4%, with comparable contributions from fossil fuel and lightning (5%–6%). Kaynak et al.  reported a small impact from lightning NOx on maximum 8 h O3 in surface air over the United States (less than 2 ppb in 71% of cases) in summer 2004.
 Here we extend these studies by comparing the impacts of changing anthropogenic and lightning sources on the U.S. NOy budget (with a focus on export and deposition), as well as the implications for O3 production and O3 air quality. We use a global 3-D model, Model of Ozone and Related Tracers (MOZART-4) [Horowitz et al., 2003, 2007; Emmons et al., 2010]. The description and evaluation of our simulations are provided in sections 2 and 3. In section 4, we present (1) the tropospheric NOy budget of the United States during the summer of 2004, (2) the sensitivity of the NOy budget to the reported anthropogenic NOx emission changes and to the new constraints on lightning NOx, and (3) an estimate of the export efficiency of U.S. surface NOx emissions to the North Atlantic. The implications of changing NOx emissions on O3 production and O3 air quality are discussed in section 5. Conclusions are given in section 6.
2. Model Description
2.1. The MOZART4 Model
 The Model of Ozone and Related Tracers (MOZART) version 4 is updated from MOZART-2 [Horowitz et al., 2003], as described by Emmons et al. . The particular configuration used in our study (model resolution, meteorology, isoprene chemistry) has been modified from that described by Emmons et al.  and is the same as that used by Horowitz et al. . Briefly, the model resolution for this simulation is 1.9° latitude by 1.9° longitude, with 64 vertical levels. The driving meteorological fields are from the Global Forecast System and updated every 3 h [Horowitz et al., 2007]. The influx of O3 from the stratosphere is prescribed by the SYNOZ (SYNthetic OZone) technique (500 Tg yr−1) [McLinden et al., 2000]. Convective mass fluxes are diagnosed by the model, using the shallow and midlevel convective transport formulation of Hack  and deep convection scheme of Zhang and McFarlane . Vertical diffusion within the boundary layer is based on the parameterization of Holtslag and Boville . Wet deposition is taken from the formulation described in the study by Brasseur et al. . The monthly mean dry deposition velocities are from Horowitz et al.  except for O3 (taken from Bey et al. ) and PAN (taken from a separate MOZART-4 simulation in which it was calculated interactively to reflect the updates described by Emmons et al. ). Global anthropogenic, biomass burning, and natural emissions are based on the POET (a project studying Precursors (CO, NOx, CH4 and NMVOC) of Ozone and their Effect on the Troposphere) emission inventory for 1997 [Olivier et al., 2003; http://www.aero.jussieu.fr/projet/ACCENT/POET.php], except for biogenic emissions of isoprene and monoterpene, which are calculated interactively using Model of Emissions of Gases and Aerosols from Nature (v.0) [Guenther et al., 2006]. Over North America, we use the daily biomass burning emissions from Turquety et al.  (including CO, NO, SO2, NH3, carbonaceous aerosols, and nonmethane volatile organic compounds) for our simulation of the INTEX-NA field campaign during summer 2004. Anthropogenic emissions over North America are modified from their POET values as described in section 2.2.
2.2. Anthropogenic Emissions Over the United States
 The U.S. surface anthropogenic emissions are from the EPA National Emission Inventory (NEI99, version 3, http://www.epa.gov/ttn/chief/net/1999inventory.html, represented as NOx99). To simulate the reported anthropogenic NOx emissions reduction, we replace the NEI99 anthropogenic NOx emissions over the United States (including part of southern Canada and Mexico) with a new inventory for summer 2004 (applied for June, July, and August in 2004, denoted as NOx04). The 2004 anthropogenic NOx emission inventory was prepared as part of Visibility Improvement State and Tribal Association of the Southeast (VISTAS) [MACTEC, 2005]. It is projected from the VISTAS 2002 U.S. emissions inventory using growth factors from the Economic Growth Analysis System Version 4.0 and is integrated with the actual power plant NOx emissions obtained from the continuous emissions monitoring data set.
Figure 1 shows the spatial distribution of the differences in surface NOx emissions from the 1999–2004 inventories (due to anthropogenic emission changes). Besides the expected NOx reductions in the northeastern United States and the Ohio River Valley [Environmental Protection Agency (EPA), 2005] due to regulations on NOx emissions under the NOx SIP call, we also see NOx reductions in southern California, which may reflect regulations on automobile NOx emissions there [Frost et al., 2008] or an overestimate of NOx emissions in the late 1990s (mostly from motor vehicles) [Kim et al., 2009]. After updating the NOx emission inventory from 1999 to 2004, the total July surface NOx emissions (including 0.1 Tg N from soil emissions) in the continental United States decreases from 0.68 Tg N in 1999 to 0.52 Tg N in 2004, consistent with the sum of soil and fuel emissions constrained by the INTEX-NA observations in the study by Hudman et al. .
2.3. Lightning NOx Parameterization
 In the MOZART-4 model, lightning flash frequency and the resulting NOx source are parameterized using the scheme of Price et al. , with the flash frequency determined by the maximum cloud top height, and the cloud-to-ground (CG) to intracloud (IC) flash ratio determined as in the study by Price and Rind . The NOx production estimated by this parameterization is around 110 and 1100 moles N per IC and CG flash, respectively. Using the original parameterization, the resulting model lightning NOx source is 0.027 Tg N during July 1 to August 15 over the United States, much lower than the 0.27 Tg N estimate constrained by the INTEX-NA observations in the study by Hudman et al. . Our model-simulated CG lightning frequency is also biased low compared to that observed by the National Lightning Detection Network (NLDN) in 2004, especially over the regions extending southwest from the Midwest to Texas (Figures 2a and 2b). However, the lightning flash frequency over the United States is close to the annual mean distribution from Schumann and Huntrieser  in which they review 3 decades of global lightning NOx studies: strong lightning around the Gulf of Mexico (up to 6–9 flashes km−2 month−1 in our model versus up to 64 flashes km−2 yr−1 and over the Midwest (around 1–3 flashes km−2 month−1 in our model versus 8–32 flashes km−2 yr−1) [Boccippio et al., 2001].
 In the model, the total lightning NOx source can be adjusted by modifying the flash frequency, the energy per flash, or the NOx emitted per unit energy. We choose to increase the NOx emitted per unit energy over the midlatitude continents in the Northern Hemisphere (north of 25°N) by a factor of 10 to achieve a similar regional production (0.27 Tg N from July 1 to August 15) as Hudman et al. . Figure 2c shows the lightning NOx source after the adjustment. Clearly, lightning NOx production follows the flash distribution (Figure 2a), maximizing near the east coast and Florida.
 We also modify the original vertical profile of lightning NOx [Pickering et al., 1998] to account for recent updates from Pickering et al. . The new profile reduces the fraction of lightning NOx in the surface layer from 20% to 2% and redistributes the remainder proportional to the original profiles over the range 2–16 km (Figure 3). Our adjusted profile in the northern hemispheric midlatitudes has a two-peak shape with peaks at 5 and 12 km as in the midlatitude lightning profiles suggested by Ott et al. ; however, the altitude and the magnitude of the second peak are both overestimated compared to their recommendation (15% at 12 km in this paper versus 10% at 9 km in the study by Ott et al. ).
2.4. Model Simulations
 We conducted four experiments with various combinations of anthropogenic NOx emissions: NEI99 NOx emissions (“NOx99”) and 2004 NOx emissions (“NOx04”), and lightning parameterizations: original lightning NOx (“LowLght”) and updated lightning NOx (“HighLght”) (Table 1). The experiment with the 2004 anthropogenic emissions and scaled lightning NOx sources (hereafter, referred to as NOx04HighLght) is used to analyze the budget and mean export of NOy (section 4.1). All simulations were conducted from May 2004 through the INTEX-NA period (July–August 2004) to allow for spin-up to capture changes in summertime continental boundary layer chemistry. Note that our NOx99LowLght simulation is identical to the “base” simulation described in the study by Horowitz et al. .
Table 1. U.S. NOx Emissions for July 2004 in the MOZART-4 Simulationsa
 MOZART-4 has previously been used to help interpret observed distributions of trace gases during field campaigns and has been shown to capture the major feature of observed CO and O3 distributions. Both Pfister et al. [2008a] and Fang et al.  show that the model CO bias compared to observations during the ICARTT campaign is generally within 10%. Fang et al.  also show that MOZART-4 fails to reproduce extreme biomass burning CO signals (600 ppbv) sampled by the NASA DC-8 flight over northeast North America and attribute the problem to the excessive mixing in the model and a possible underestimate of the biomass burning emission injection height. Pfister et al. [2008a] find that MOZART represents the overall shape of the O3 profiles monitored by the IONS-04 (INTEX Ozonesonde Network Study) throughout most of the troposphere but overestimates concentrations near the surface at most east coast IONS sites. The overestimate of surface O3 and relatively low bias in the free troposphere may be attributed to an overestimate in surface emissions, coarse resolution in global models, and insufficient boundary layer ventilation in the model. When we compare our MOZART-4 simulated monthly mean daily maximum 8 h surface (MDA8) O3 during summer time in 2004 with that observed at Clean Air Status and Trends Network sites (not shown), we find a similar overestimate of surface ozone over the eastern United States of 10–15 ppbv and an underestimate over the western United States by about 8 ppbv. The bias in the eastern United States also exists in other simulations with previous versions of MOZART [Lin et al., 2008; Murazaki and Hess, 2006] and occurs systematically in the multimodel estimates of surface O3 for the year 2001 of the Task Force on Hemispheric Transport of Air Pollution [Fiore et al., 2009; Reidmiller et al., 2009]. Horowitz et al.  compare MOZART-4 simulations and observations from the ICARTT field campaign over the eastern United States during summer 2004 and find simulated concentrations of trace species (including O3) generally match observations to within 30% in the U.S. boundary layer (below 2 km), except for NOx (overestimated by ∼30%) and PAN (overestimated by a factor of ∼2).
3.2. MOZART-4 Simulations of the INTEX-NA Campaign in Summer 2004
 In this section we compare our suite of MOZART-4 sensitivity experiments (Table 1) with observations made on board the NASA DC-8 aircraft during the INTEX-NA campaign [Singh et al., 2006]. The wide spatial coverage of this aircraft is suitable for comparison with our global model. Our evaluation focuses on NOy (NOx, PAN, and HNO3) and other related species (O3 and OH). We also compare the simulated nitrate wet deposition during summer 2004 with that monitored by the National Atmospheric Deposition Program [NADP, 2010].
3.2.1. Boundary Layer Distributions
 Here we focus on the NOx 99HighLght and NOx 04HighLght simulations (Table 1). Comparisons of selected species below 2 km in the eastern United States are presented in Figure 4. In general, we find that updating the anthropogenic NOx emissions from 1999 emission inventory to 2004 has little influence on the spatial correlation with observed NOy, O3, and OH (r > 0.5, except for HNO3, r = 0.3), but it improves the model biases (represented as the absolute/relative differences between the model value averaged across all sampled grid cells and the observed mean value) except for HNO3 (Figure 4): The mean NOx bias decreases from 153 ppt (30%) in NOx 99HighLght to 15 ppt (3%) in the NOx04HighLght simulation and the O3 bias decreases from 6 ppb (12%) to 2 ppb (4%). The apparent degradation of the spatial correlation for NOx when switching from the 1999 to 2004 emission inventory is due to one model grid cell (covering Ohio and Pennsylvania), where the NOx04HighLght simulation underestimates observations by more than NOx99HighLght simulation (excluding that point yields equivalent correlations for the simulations versus observations, r = 0.58). PAN is overestimated (by ∼350 ppt in NOx99HighLght and ∼250 ppt in NOx04HighLght). This bias persists when we update all the VOC emissions in addition to NOx from the 1999 to the 2004 inventory. The model underestimates HNO3 (mean bias of −37 pptv in NOx99HighLght and −379 pptv in NOx04HighLght). The underestimate of HNO3 and overestimate of PAN will likely cause an overestimate of NOy export from the model boundary layer (BL).
3.2.2. Vertical Distributions
Figure 5 shows the mean profiles of different species within the eastern United States for all the model simulations. In general, the factor of 10 increase in lightning NOx improves simulated NOx, O3, and HNO3 in the free troposphere: the upper tropospheric NOx underestimate decreases from 750 to 540 ppt, comparable to the results from Hudman et al. ; the O3 discrepancy decreases to within 10 ppb; the HNO3 discrepancy decreases to within 40 ppt above 6 km. The modeled OH and PAN worsen as compared to the observations. OH in the upper troposphere is underestimated by the LowLght simulations by around 50%; the lightning adjustment, while improving the OH prediction above 10 km, leads to an overestimate of OH in the midtroposphere with a maximum bias of 60%. PAN is overestimated by around 70 ppt in the midtroposphere and less than 30 ppt in the upper troposphere in the NOx04HighLght and NOx99HighLght simulations (mean biases are both less than 20%). The different sign and location of maximum biases for NOx, PAN, and HNO3 suggest that a further modification of the lightning NOx profile will not help improve the prediction of all these species; therefore, there must be other causes of these biases. As a check on the total NOy simulation, we add HNO3, PAN, and NOx (the major NOy species) from both model and observations and find that they agree better than the individual species do. For example, between 4 and 8 km, the sum of the major NOy species agrees with observations to within 10% after the lightning adjustment.
Hudman et al.  also found problems in HNO3 and PAN, especially in the mid-upper troposphere, but their PAN was underestimated by ∼30% in the upper troposphere, while their HNO3 was twice the observed value; simply adjusting lightning will not fix that HNO3 and PAN partitioning bias either. Henderson et al.  noted an NOy partitioning problem as compared to observations when they applied several chemical mechanisms (CB05, RACM2, SAPRC07, and GEOS-Chem) in a box model. The NOy partitioning bias reflects that some processes important for NOy partitioning in the upper troposphere are missing in the model, e.g., overestimates of radical sources and subsequent chemical cycling [Henderson et al., 2009]. A recommendation from their study (already adopted in MOZART-4 [Emmons et al., 2010]) is to use the Blitz et al.  estimate for the acetone quantum yield, which reduces the acetone photolysis and decreases PAN and HOx (= OH + HO2) in the upper troposphere throughout the Northern Hemisphere [Arnold et al., 2005]. Uncertainties in isoprene chemistry are also known to affect PAN formation [e.g., Emmerson and Evans, 2009; Paulot et al., 2009; Pfister et al., 2008b]. We use a 4% yield of isoprene nitrate production from the reaction of isoprene hydroxyperoxy radicals with NO, which best captures the boundary layer concentrations of organic nitrates and their correlation with ozone observed in the INTEX-NA field campaign; however, many uncertainties remain [Horowitz et al., 2007]. The excess PAN relative to HNO3 in the midtroposphere in our model is likely to lead to an overestimate of the potential influence of U.S. NOx emissions on O3 production and NOy deposition in downwind regions.
3.2.3. Wet Deposition of Inorganic Nitrate
 We compare the total summer time inorganic nitrate (including nitric acid and aerosol nitrate) wet deposition from the NOx99LowLght, NOx04LowLght, and NOx04HighLght simulations with observations from NADP  (Figure 6) during June–August 2004. The NOx99LowLght simulation is well correlated with observations (r = 0.72). There is a significant overestimate of the nitrate wet deposition over northeastern California, the Midwest and the northeast corridor, associated with the overestimate of NOx emissions over these regions. When we update from the 1999–2004 anthropogenic NOx emission, these overestimates are reduced in general (+15% to −5%), especially over the northeast corridor. The modeled correlation also increases to 0.75. Further scaling of the lightning NOx by a factor of 10 does not affect the correlation but leads to a better match with observations over the midwest. However, the uniform scaling causes excessive wet deposition along the east coast. The overestimate of the mean nitrate wet deposition (+30%) compared to the NADP observations in our NOx04HighLght simulation indicates that the lightning NOx source (180 Gg in July) is probably too high. Rather, the discrepancies likely reflect a problem with the flash frequency distribution as seen in Figure 2.
4. Sensitivity of the NOy Budget on Anthropogenic and Lightning NOx Emissions
4.1. Budget of NOy Over the United States
 In this section we discuss the chemical processing, deposition, and export of the U.S. NOx emissions in July, focusing on the NOx04HighLght simulation. We define the contiguous U.S. boundary layer (BL) as the region extending horizontally from 24°N–48°N and 67.5°W–127.5°W, from the surface to about 800 hPa (around 2 km). We also define the contiguous U.S. total column (TC) to refer to the same horizontal region but vertically extending to about 200 hPa. Hereafter, we refer to these lateral boundaries (up to 200 hPa) as “walls”; for example, the lateral boundary at 67.5°W, 24°N–48°N, from the surface to 200 hPa is referred to as the “east wall.”
 Most of the NOy exported from the BL is transported vertically to the free troposphere (about 90%). NO2 and HNO3 are the most abundant NOy species in the BL (HNO3 accounts for more than 30%, whereas NO2 accounts for more than 25% of the BL NOy burden) since PAN is thermally unstable in the BL. NO2 and HNO3 are also the most abundant components of NOy exported from the U.S. BL (each accounts for more than 30%, with PANs contributing over 10%).
 Lightning NOx accounts for almost 30% of the total U.S. July NOx source in the model, similar to that estimated from NLDN [Kaynak et al., 2008]. Around 80% of all emitted NOx in the U.S. TC over the United States is deposited and about 20% is exported (Table 2), consistent with the estimate by Sanderson et al. . Eastward export through the east wall of the United States dominates the export from the TC (116 versus 19 and 17 Gg N exported through the north and south walls, respectively; there is a 29 Gg N inflow through the west wall in July). PANs (PAN + MPAN) contribute 50% and NOx contributes 17% to the NOy species composing the eastward flux.
Figure 7 shows the NOy flux through the east wall of the United States. Outflow dominates the transport here except south of 30°N, where weak inflow into the United States exist. This pattern reflects the dominance of the Bermuda High during July. Because of the location of North American jet stream, a maximum eastward outflow is located between 200 and 400 hPa, centered at around 45°N. Vertical profiles of the relative contribution of major components of NOy to this eastward flux in the NOx04HighLght simulation are shown in Figure 8. Near the surface, HNO3 is by far the largest component of exported NOy with a maximum contribution of more than 40%. Wet deposition decreases the HNO3 contribution with altitude from the surface to 600 hPa, and then its contribution remains almost constant at 30% up to 300 hPa. PANs contribute most to the exported NOy above the boundary layer with a maximum contribution of more than 60% located at around 500 hPa, reflecting their longer lifetime at the colder temperatures in the free troposphere (FT). Above 500 hPa, the relative contribution of PANs decreases as that of NOx increases.
4.2. Effects of Recent Anthropogenic NOx Emission Reductions on the U.S. NOy Budget and Export to the North Atlantic
 We determine the effects of the NOx emission reductions in response to the SIP call on the budget of NOy by comparing the NOx99HighLght and the NOx04HighLght simulations (Table 2). From NOx99HighLght to NOx04HighLght, BL NOx emissions decrease by 23%. As a result, the NOy burden within the BL decreases by 19% (Table 2, from 19 to 15 Gg N). Meanwhile, the net export of reactive nitrogen from the BL decreases by 20%, while the total deposition from the BL decreases by 24% (Table 2). This nonlinearity between NOx emission changes and the response of the NOy budget terms is associated with changes in NOy partitioning [Liang et al., 1998]. As surface NOx emissions decrease, PAN increases from 22% to 24% while HNO3 decreases from 36% to 32% of the NOy burden. A smaller contribution from HNO3 (which is efficiently removed by dry and wet deposition) implies a longer NOy lifetime (+5% in the BL) and hence a relative increase in the NOy burden and export, resulting in a less-than-linear reduction of these two terms. For the same reason, the deposition of NOy decreases more than linearly. Similar results occur for the TC over the United States (Table 2). Decreasing the U.S. anthropogenic NOx emissions thus lowers the NOy burden, NOy deposition, and NOy export from the United States in summer but decreases NOy deposition more efficiently than the NOy burden and export.
 Decreasing anthropogenic NOx emissions also reduces the eastward export through the east wall of the United States (Figure 7b), with the largest reduction near the surface around 42°N and at 300 hPa at around 45°N. The corresponding NOy concentration changes along the east wall of the United States (not shown) have a similar pattern. Surface emissions thus have a strong potential to affect not only surface export but also the free tropospheric eastward export north of 40°N, consistent with the dominant export mechanism that is associated with frontal passages at this latitude [e.g., Fang et al., 2009; Owen et al., 2006; Li et al., 2005; Cooper et al., 2001, 2002; Merrill and Moody, 1996].
4.3. Impact of Lightning NOx on the U.S. Export to the North Atlantic and Deposition
 By comparing the NOx04LowLght and NOx04HightLght simulations, we found that due to the lightning adjustment, the NOx source to the FT over the United States (includes inflow through the west wall, BL ventilation, and emissions within the FT) increases by 77%, while the total deposition (only the wet deposition since dry deposition only occurs in the model lowest level) increases by 99% and the export (includes northward, southward, and eastward export) increases by 53%. The less-than-linear variation of NOy export and more-than-linear response of NOy deposition to the source change reflect the partitioning change (the contribution of HNO3 increases and that of PANs decreases), consistent with our findings from the anthropogenic emission sensitivity experiment in section 5 (note the opposite sign of the emission perturbations).
 The lightning adjustment results in a different pattern of the eastward NOy flux change (Figure 7c) from that associated with anthropogenic emissions (Figure 7b), with one maximum located around 400 hPa, centered at 42°N. The PAN contribution changes up to 10% in the FT in the lightning perturbation case, whereas it is only about −2% in the anthropogenic emission perturbation case (Figure 8) even though the NOx emission perturbations within the total column are similar in both experiments. If the adjusted lightning were correct, this partitioning change implies the LowLght simulations would likely overestimate the relative contribution of PAN to the total exported NOy while underestimating the absolute export of NOy produced by lightning. Using the NOy partitioning simulated by the LowLght models would, in this case, overestimate the impact of the U.S. anthropogenic NOx emission reductions on global and downwind O3 air quality even with good constraints on U.S. anthropogenic NOx emissions.
 Here we estimate the contributions from lightning NOx sources to the total FT eastward export, the largest export pathway for NOy. For this estimate, we use the NOx04HighLght and NOx04LowLght simulations in which lightning NOx changes by an order of magnitude. The export change is not linear to the emission change, as the export through the east wall increases from 74 to 106 Gg N (Table 3, +43%,) in July after the lightning adjustment (which increases the total FT source by 77%, Table 3). We assume that this change is only due to the lightning NOx increase over the U.S. plus the increase in the inflow through the west wall (from 24 to 30 Gg N, reflecting the lightning NOx increase imposed over all the northern hemisphere midlatitude continents). We bracket our estimate of the lightning contribution to the U.S. FT eastward fluxes using the following two limits: (A) the NOy entering through the west wall blows directly across the region and is exported through the east wall and (B) the NOy from the west wall is all deposited within the United States. As shown in Table 4, the increase in lightning NOx in the FT thus contributes from 26 (scenario A) to 32 GgN (scenario B) to the total July NOy eastward export in the NOx04HighLght simulation. The original lightning NOx source over the United States provides 14 Gg N during July. Applying the previous upper and lower bound assumptions of 0%–100% deposition of lightning NOx in this simulation yields a range of 0–14 Gg N contribution from the original lightning to the July eastward export. Combining this range with the estimated range of the enhanced export from the increased lightning yields an estimate for the total lightning contribution of 26–46 Gg N or 24%–43% of the total U.S. NOy export to the North Atlantic in the FT.
Table 3. July Budget of NOy Species in the U.S. (24°N–48°N, 127.5°W–67.5°W) Free Troposphere (800–200 hPa)a
Assumptions: (A) NOy entering through the west wall blows directly across the region and is exported eastward through the east wall; (B) NOy from the west wall is all deposited within the United States.
Defined as the eastward NOy export flux through the east boundary of the United States (24°N–48°N, 67.5°W) from the NOx04HighLght simulation.
Calculated as the difference between the NOx04HighLght and NOx04LowLght simulations, see section 4.3.
Original lightning in the NOx04LowLght simulation, see section 4.3.
This term is derived by assuming a 0%–100% contribution from these emission sectors.
This term is derived as the FT NOy eastward export, other components (inflow from the west wall, local lightning increase, original lightning, biomass burning, and aircraft source).
 With similar assumptions, we estimate the contribution from lightning to wet deposition over the United States. Combining the lightning NOx in the NOx04LowLght simulation (0–14 GgN) with the increase in FT wet deposition from the NOx04LowLght to the NOx04HighLght simulation (96–102 Gg N), we estimate that lightning NOx accounts for 96–116 GgN (47%–57%) of total wet deposition from the FT. Since the lightning source within the BL is only around 2% of the total BL NOx emissions, we assume that its contribution to wet deposition in the BL (134 GgN) is negligible and estimate the lightning NOx contribution to the total U.S. NOy wet deposition to be 28%–34%. Over remote regions, the lightning NOx contribution to total wet deposition may be much higher. Changes in lightning activity may thus be detectable from measurements of nitrate deposition, particularly if subjected to nitrogen isotope analysis that can distinguish among anthropogenic and lightning sources [Hastings et al., 2003].
4.4. Export Efficiency of U.S. Surface NOx Emissions to the North Atlantic
 Since the export of NOy to the North Atlantic in the FT is the component affecting O3 production over downwind regions, we modify the traditional Eulerian budget analysis (section 4.1) to isolate this component. Specifically, we calculate the NOy export efficiency (f): f = NOyFT eastward export/NOx emissions, where NOx emissions are surface emissions and the NOyFT eastward export is the NOy flux through the east wall that comes from surface ventilation. Table 4 shows the contributions of different components to the total July FT NOy eastward export in the NOx04HighLght simulation. The eastward export in the FT originating from the surface NOx emissions is estimated to be 22–74 GgN (out of the total surface emissions of 520 Gg N in July), yielding a 4%–14% range for the FT export efficiency to the North Atlantic, with the range resulting from various assumptions (see section 4.3 and Table 4).
 Another method to derive the export efficiency of NOx emissions (hereafter, referred to as the “plume-based method”) uses CO as an inert tracer and estimates the export efficiency using the NOy-CO relationship sampled in the outflow plumes downwind. Parrish et al.  analyzed the difference between the traditional Eulerian budget estimate and the plume-based estimate and attributed it to losses in NOy occurring between export from the continental BL and arrival at the point sampled. Applying this plume-based method, Hudman et al.  reported an NOy export efficiency of 14% ± 8% from GEOS-Chem model and of 16% ± 10% from the ICARTT observations between 2.5 and 6.6 km. Our modified Eulerian budget estimate (4%–14%) falls into the range of this and other plume-based estimates [3%–20%, e.g., Stohl et al., 2002; Parrish et al., 2004; Li et al., 2004; Hudman et al., 2007].
 The accuracy of the plume-based export efficiency estimate depends on many factors, such as the treatment of the background CO [Li et al., 2004] and the altitude of the plumes sampled. Our approach provides an average view of the NOy export efficiency in the FT to the North Atlantic.
 Results from the anthropogenic sensitivity experiment (section 4.2), in which 157 Gg increases of anthropogenic emissions causes a 5 Gg N increases in the FT eastward export (Table 2), suggest that the anthropogenic NOy export efficiency from the increased anthropogenic emission is 3%. However, this estimate is unlikely to represent the “real” anthropogenic export efficiency due to the nonlinearity discussed in section 4.2. Indeed, our analysis above indicates that the export efficiency of the total anthropogenic emissions is greater than 3%. To avoid the impact of nonlinearity induced by emission perturbations from sensitivity studies on the export efficiency calculation, a possible solution is to tag anthropogenic NOx within the United States [Horowitz and Jacob, 1999] and thereby quantify its contribution to the eastward NOy export.
5. Implications for O3 Production
5.1. O3 Production in the U.S. BL and In the Northern Hemispheric FT
 We examine the implications for O3 production from the changes in anthropogenic and lightning NOx sources. Our simulations indicate that the 23% decrease in surface NOx emission from the 1999–2004 inventory decreases the gross O3 production in the United States BL by 10%, from 17.0 to 15.3 G mol d−1. The less-than-linear decrease reflects an increase in ozone production efficiency (OPE, from 19.1 to 23.9, calculated as regional mean gross O3 production divided by regional mean NOx loss rate), similar to findings from previous studies [Liu et al., 1987; Lin et al., 1988; Jacob et al., 1993; Thompson et al., 1994; Horowitz et al., 1998]. The OPE over the northeastern United States increases by around 20%, higher than the 9% estimate from Hudman et al. , due to a larger NOx emission change in our experiment within this region (24%) relative to their work (15%). The direct O3 export out of the U.S. BL decreases by 0.55 Gmol d−1 (by 28% of the O3 direct export), around 32% of the decrease in the gross O3 production (1.7 Gmol d−1) within the U.S. BL. This ratio is comparable to that found by Li et al.  when the North American NOx anthropogenic emissions were turned off. The decrease in FT O3 production over the United States, due to decreased NOy exported from the BL, is 0.4 G mol d−1, 73% of the direct O3 export reduction in the U.S. BL from the change in anthropogenic NOx emissions.
 The enhanced lightning NOx increases gross O3 production in the FT over the United States by 49% (from 10.6 to 15.5 G mol d−1). Meanwhile, the OPE is reduced by 46% (from 56.3 to 30.2). A previous study showed a mean FT OPE of 61 over the United States [Liang et al., 1998], similar to our NOx04LowLght simulation. The gross O3 production in the FT north of 30°N is 55.4 G mol d−1 in the NOx04LowLght simulation, also consistent with the estimate of Wang et al.  (51 Gmol d−1 annually). After increasing the lightning NOx source in the NOx04HighLght simulation, the gross O3 production within this region is enhanced to 73.7 G mol d−1 (+33%) with the mean OPE reduced from 56 to 39 (−30%). A model with a lightning NOx source biased low will overestimate the background OPE along with the relative contribution of PAN to NOy export, both of which may lead to an overestimate of the contribution from anthropogenic NOx emissions to O3 in downwind regions even if the anthropogenic emissions are well represented. As an illustration, we examine the 500 hPa O3 concentration response over southern Europe to the same anthropogenic emission change over the United States under the original lightning case (O3 (NOx99LowLght) - O3 (NOx04LowLght)) and under the increased lightning case (O3 (NOx99HighLght) - O3 (NOx04HighLght)). We find that the O3 response in the original lightning (LowLght) case is about 30% higher than that in the increased lightning (HighLght) case. Further efforts are needed to better quantify the lightning NOx source to obtain a more reliable evaluation of the impact of anthropogenic NOx emissions on global-scale O3 production.
5.2. Surface O3 Over the Contiguous United States
 The anthropogenic NOx emission reductions from 1999 to 2004 improve O3 air quality, especially in the eastern United States and California (Figure 9). July MDA8 O3 decreases by 9–12 ppbv in these two regions where total O3 is above 70 ppbv. The maximum decrease in MDA8 O3 over the eastern United States occurs to the south of the maximum NOx emission change (over Kentucky, Tennessee, and Missouri, Figure 1), consistent with the results of Hudman et al.  and Kim et al. . Within the northeastern and mid-west United States, the monthly mean MDA8 O3 decreases by around 5–7 ppbv (comparable to 4–6 ppbv from Hudman et al.  and 7 ppbv from Kim et al. ). Over Los Angeles, MDA8 O3 increases despite the decrease of NOx emissions, likely reflecting the NOx-saturated regime for O3 production there. This trend from 1999 to 2004 is consistent with the trend reported by EPA (www.epa.gov/airtrends/weather/region09.pdf), where O3 is shown to increase over Los Angeles, despite the decrease over most California. Decreases of the monthly mean 24 h average surface O3 concentrations show a similar pattern to MDA8 O3, except the maximum change is lower (6 ppbv). Stronger sensitivity of the MDA8 O3 reflects the rapid photochemical production fueled by local precursor emissions as compared to the complex nighttime processes that affects the 24 h average surface O3 [Russell et al., 1986].
 Surface O3 concentrations are weakly affected by the adjustment to the northern hemispheric lightning NOx source. The monthly mean MDA8 O3 concentration increases by up to 3.5 ppbv, with the largest increase occurring to the northeast of Florida and to the east of California. We scaled the lightning NOx source by a factor of 10 in the NH midlatitudinal continents, and we also reduced the fraction of lightning NOx in the surface layer from 20% to 2% (see section 2). These adjustments compensate in the boundary layer. Hence, the surface O3 change is mostly due to the lightning NOx change in the FT. The maximum O3 concentration change near Florida is associated with the maximum lightning NOx emission change over Florida (Figure 2b). However, the lightning NOx change corresponding to the maximum O3 concentration change to the east of California is weak. In this region, subsidence of air dominates, bringing the lightning-influenced air parcels from the free troposphere to the surface over the western United States. This downward O3 flux into the planetary boundary layer over western North America is also shown by Parrington et al. . The monthly mean 24 h average surface O3 shows both a similar pattern and similar magnitude (up to 3 ppbv) as that of MDA8 O3, confirming that lightning NOx affects surface O3 through downward transport of free tropospheric air rather than through local photochemical production. In general, the lightning-induced surface O3 concentration changes are small and not located in the most polluted regions, e.g., 8 h O3 nonattainment regions at northeast corridor and California (http://www.epa.gov/air/oaqps/greenbk/map8hrnm.html), and therefore, the lightning adjustment has a relatively small impact on the typical maximum O3 over those regions where it is of greatest concern [Kaynak et al., 2008]. However, it does affect the background O3 (i.e., O3 that is not produced by U.S. anthropogenic emissions). The contribution from lightning NOx to surface O3, estimated as the difference between the NOx04HighLght and NOx04LowLght simulations, is as high as 3 ppbv over the western United States and near Florida and around 1 ppbv elsewhere. These value suggest lightning NOx contributes one tenth (or less) of the surface O3 background, assuming estimates of 15–45 ppb (highest in the western United States; this range incorporates several methods for estimating background O3 level) [Altshuller and Lefohn, 1996; Lin et al., 2000; Jaffe et al., 2003; Fiore et al., 2002, 2003; Vingarzan, 2004].
 Nitrogen oxides (NOx) and their oxidation products affect the tropospheric O3 budget and the oxidizing power of the atmosphere, as well as ecosystem productivity and thereby the global carbon cycle. U.S. anthropogenic NOx emissions have decreased since the mid-1990s, especially from 1999 to 2004 in response to the U.S. Environmental Protection Agency (EPA) State Implementation (SIP) call [Kim et al., 2006; Stavrakou et al., 2008; van der A et al., 2008; Frost et al., 2006]. In addition, recent work suggests that lightning NOx is higher than previously thought over the midlatitudes of the Northern Hemisphere and that the empirical vertical profile of lightning-produced nitrogen applied in the current generation of CTMs (chemical transport model) needs revision [Ott et al., 2007, 2010; Pickering et al., 2006]. We use the MOZART-4 chemical transport model to examine the budget of NOy species in the United States during summer 2004 and investigate its sensitivity to these changes in NOx sources.
 Our evaluation of the model with observations from INTEX-NA shows that using the 2004 NOx emission inventory versus the 1999 inventory (NEI99) reduces the mean bias of the boundary layer (BL) simulation for most species (including NOx, OH, PAN, and O3) while maintaining the spatial correlations (r > 0.5 for most NOy species). The simulated bias is also smaller than the mean bias in Singh et al.  (bias decreases from 40% to 6% and >100% to 67% for NOx and PAN, respectively). Scaling up the FT NOx lightning emission by a factor of 10 improves the profiles of NOx and O3 relative to observations, but there is still a considerable underestimate (60%) in the upper troposphere for NOx. HNO3 and PAN are not notably improved in our model after the emission adjustments, indicating a bias in chemical partitioning, as found in other models [Singh et al., 2007; Henderson et al., 2009; Yu et al., 2010], possibly reflecting uncertainties in isoprene chemistry [e.g., Emmerson and Evans, 2009; Ito et al., 2009; Pfister et al., 2008b] or the convection related processes that affect the upper troposphere [Bertram et al., 2007]. The simulation using 2004 NOx emissions yields a spatial correlation of 0.75 between modeled nitrate wet deposition and the National Atmospheric Deposition Program (NADP) observations. Increasing the lightning NOx improves nitrate wet deposition prediction over the midwest and inland of southeast United States but worsens over the east coast (Figure 6), likely reflecting a problem in simulated lightning frequency (Figure 2).
 We assessed the impact of the anthropogenic and lightning NOx changes on U.S. O3 air quality. Anthropogenic NOx emission reductions lower the monthly mean daily maximum 8 h surface (MDA8) O3 by up to 12 ppb over Kentucky, Tennessee, Missouri, and California and by 5–7 ppb over Northeast Corridor and midwest, consistent with prior studies [Hudman et al., 2009; Kim et al., 2006]. The lightning impact on surface O3 is weaker, especially over the highly polluted regions [Kaynak et al., 2008], although it does raise background O3 levels by 3 ppbv over the western United States and near Florida and 1 ppbv in other regions.
 In our model, most NOx emitted in July is removed by wet or dry deposition within the United States (∼70% in the boundary layer, BL, and 80% in the total column, TC), with the rest exported. The imposed NOx emission changes yield a more-than-linear response of NOy deposition and a less-than-linear response of NOy export, resulting from the NOy partitioning and lifetime changes [e.g., Liang et al., 1998]. We estimated the range of lightning NOx contribution to the total NOy wet deposition over the United States and to the U.S. FT export to the North Atlantic to be 28%–34% and 24%–43% in the NOx04HighLght simulation, respectively, with the range reflecting different assumptions (section 4.3 and Table 4).
 We further calculate the range of the anthropogenic contribution to the U.S. FT export to the North Atlantic to be 22–74 GgN in July, yielding an estimate for NOy export efficiency (the ratio of the anthropogenic export to the North Atlantic between 800 and 200 hPa over total U.S. surface NOx emissions) of 4%–14%. The high end of this range is consistent with the plume-based estimates of NOy export efficiency during the INTEX-NA campaign (14% ± 8% in the GEOS-Chem model and 16% ± 10% from the INTEX-NA observations between 2.5 km and 6.6 km) [Hudman et al., 2007] and the range of previous plume-based estimates [e.g., Parrish et al., 2004; Li et al., 2004; Stohl et al., 2002].
 The adjustment to the lightning NOx source increases the fractional contribution of HNO3 to NOy while decreasing that of PAN (Figure 8), suggesting that an underestimate of lightning NOx would cause the relative PAN contribution to the total NOy export to be overestimated. The Ozone Production Efficiency (OPE) in the FT over the United States and the extratropical Northern Hemisphere decrease from 55 to 30 and from 56 to 39, respectively, when lightning NOx is increased. Even if anthropogenic emissions are well represented, an excessive fractional PAN contribution to NOy export and northern hemispheric OPE would bias high the estimated impact of the U.S. anthropogenic NOx emission reduction on global and downwind O3 air quality (e.g., at 500 hPa over southern Europe, the response to the same U.S. anthropogenic NOx changes are 30% higher in the low versus high lightning cases). Better constraints on the lightning NOx source are required to more confidently assess the impacts of anthropogenic emissions and their changes on air quality over downwind regions.
 The authors would like to thank Rynda Hudman, Robert Pinder, Barron Henderson, Emily Fischer, Kenneth E. Pickering, and Anand Ganadesikan for informative conversations; Rynda Hudman and Songmiao Fan for insightful reviews, the INTEX-NA campaign team for providing the NASA DC-8 aircraft observational data; and the MOZART-4 development team.