We examine the sources and chemistry affecting nitrogen oxides (NOx = NO + NO2) over the tropical Pacific (30°S–20°N) using observations from the Pacific Exploratory Mission to the Tropics B (PEM-Tropics B) aircraft mission conducted in March–April 1999. A global model of tropospheric chemistry driven by assimilated meteorological data is used to interpret the observations. Median concentrations observed over the South Pacific during PEM-Tropics B were 7 pptv NO, 16 pptv peroxyacetyl nitrate (PAN), and 34 pptv nitric acid (HNO3); the model generally reproduces these observations but overestimates those over the North Pacific. Lightning was the largest source of these species in the equatorial and South Pacific tropospheric column and in the tropical North Pacific upper troposphere. The oceanic source of acetone implied by high observations of acetone concentrations (mean 431 pptv) allows an improved simulation of PAN/NOx chemistry. However, the high acetaldehyde concentrations (mean 78 pptv) measured throughout the troposphere are inconsistent with our understanding of acetaldehyde and PAN chemistry. Simulated concentrations of HNO3 and HNO3/NOx are highly sensitive to the model representation of deep convection and associated HNO3 scavenging. Chemical losses of NOx during PEM-Tropics B exceed chemical sources by a factor of 2 in the South Pacific upper troposphere. The chemical imbalance, also apparent in the low observed HNO3/NOx ratio, is explained by NOx injection from lightning and by frequent convective overturning which depletes HNO3. The observed imbalance was less during the PEM-Tropics A campaign in September 1996, when aged biomass burning effluents over the South Pacific pushed the NOx budget toward chemical steady state.
 The capacity of the atmosphere to remove pollutants is largely dependent on the supply of oxidants in the troposphere [Thompson, 1992]. Nitrogen oxides (NOx = NO + NO2) are critical for generating ozone (O3) and the hydroxyl radical (OH), the two most important tropospheric oxidants. Primary sources of NOx to the troposphere include combustion, lightning, and microbial processes in soils [Lamarque et al., 1996; Lee et al., 1997]. These sources are almost exclusively from land. Because of its short lifetime against atmospheric oxidation (1–10 days), NOx tends to be low over oceans. This depletion of NOx over much of the troposphere limits the rate of O3, and hence, OH production. The dominant sources of NOx over oceans are thought to include long-range transport from continental source regions as well as chemical recycling of nitric acid (HNO3) and peroxyacetyl nitrate (PAN) [Moxim et al., 1996; Penner et al., 1998]. Both HNO3 and PAN can have lifetimes of weeks in the troposphere and therefore provide a secondary source of NOx far from the point of original emission.
 We examine here the factors affecting NOx concentrations over the remote tropical Pacific using observations obtained during the Pacific Exploratory Mission to the Tropics phase B (PEM-Tropics B), a NASA Global Tropospheric Experiment (GTE) aircraft campaign conducted in March–April 1999 [Raper et al., 2001]. The campaign employed two aircraft, a DC-8 and a P-3B, outfitted with instruments to measure O3, hydrogen oxide radicals (HOx = OH + HO2), NOx, HNO3, PAN, CO, hydrocarbons, and a large number of related species. Operating out of Hawaii, Christmas Island, Fiji, Tahiti, and Easter Island, flights during PEM-Tropics B extensively surveyed the Pacific from 40°S to 40°N with numerous vertical profiles (flight tracks are shown in Figure 1). The PEM-Tropics B data complement observations obtained during previous GTE campaigns over the Pacific, in particular PEM-Tropics A [Hoell et al., 1999], which surveyed a similar domain during August–September 1996. We interpret the PEM-Tropics B observations with a global 3-D chemical transport model driven by assimilated meteorological observations to assess the importance of primary sources and chemical recycling in controlling the abundance of NOx in the region.
 Previous analyses of the PEM-Tropics A observations identified some of the factors determining the NOx budget over the South Pacific in austral spring, when biomass burning in the Southern Hemisphere is at its seasonal maximum. Schultz et al.  used a photochemical box model constrained with PEM-Tropics A observations to show that decomposition of PAN originating from biomass burning accounted for most of the NOx in the lower troposphere (below 6 km). The NOx budget in the upper troposphere during PEM-Tropics A was found by Schultz et al. [1999, 2000] to deviate from chemical steady state by a factor of 2–5 as diagnosed by the ratio LNOx/PNOx of chemical loss of NOx (LNOx) to chemical production (PNOx), computed with their photochemical model. With a global 3-D model, Staudt et al.  found that transport and subsidence of NOx produced by lightning over the equatorial Pacific warm pool was responsible for perturbing the NOx budget in the upper troposphere during PEM-Tropics A away from chemical steady state.
 Concentrations of NOx, HNO3, and PAN observed in the upper troposphere over the South Pacific during PEM-Tropics B were a factor of 2–3 less than those observed during PEM-Tropics A. PEM-Tropics B was conducted in the wet season of the southern tropics, when biomass burning is minimum. Using a photochemical box model to analyze the PEM-Tropics B observations, Olson et al.  calculated a LNOx/PNOx of 2.0–3.3, similar to Schultz et al. [1999, 2000] for PEM-Tropics A. Likewise, Wang et al.  found in a 1-D model for the PEM-Tropics B conditions that a large primary source of NOx was needed in the middle and upper troposphere to explain the surplus of PAN and HNO3 relative to NOx. We will examine this issue further here using the capability of our global 3-D model to explicitly resolve the primary sources of NOx during PEM-Tropics B.
 A number of global 3-D models have been applied previously to the investigation of NOx in the remote troposphere, and have had some success in reproducing observed PAN and NOx concentrations [Singh et al., 1998; Levy et al., 1999; Emmons et al., 1997; Thakur et al., 1999; Bey et al., 2001a, 2001b; Staudt et al., 2002]. A general result from these models is that decomposition of PAN is the principal source of NOx in much of the remote troposphere. This result is sensitive to the abundances of acetaldehyde (CH3CHO) and acetone, the prinicipal precursors responsible for conversion of NOx to PAN. PEM-Tropics B included the first extensive measurements of CH3CHO and acetone in the Pacific troposphere, and these indicated concentrations higher than would be expected from current models by a factor of 3 or more [Singh et al., 2001]. Such a model underestimate could challenge the conventional view of the importance of PAN as a source of NOx in the remote troposphere. We will examine this issue in this paper.
 PEM-Tropics B included a number of flights over the northern tropical Pacific out of Hawaii, in contrast to PEM-Tropics A, which focused on the South Pacific. Observations of CO, hydrocarbons, and halocarbons in that region indicated considerable combustion influence, with fossil fuel sources from Eurasian sources dominating in the boundary layer and biomass burning sources from Southeast Asia dominating in the free troposphere [Blake et al., 2001; Staudt et al., 2001]. Biomass burning in the northern tropics (including Southeast Asia and the Indian subcontinent) was at its seasonal maximum during PEM-Tropics B. We will examine here the importance of these sources in determining the concentrations of NOx over the northern tropical Pacific, particularly in relation to lightning, which was found in previous global model studies to be the principal source of NOx in the region [Levy et al., 1999; Tie et al., 2001].
2. PEM-Tropics B Meteorological and Chemical Environment
 As shown in Figure 1, the PEM-Tropics B aircraft sampled extensively the troposphere above the tropical Pacific. Transport to this region was governed by semipermanent subtropical high pressure systems located above the northeast and southeast Pacific, which bring air from midlatitudes toward the equator, feeding into the easterly flow of the trade winds [Merrill, 1989]. Major barriers to horizontal air flow were the Intertropical Convergence Zone (ITCZ) and the South Pacific Convergence Zone (SPCZ) [Fuelberg et al., 2001]. The ITCZ was less defined than on average over the Pacific during PEM-Tropics B because of normal seasonal variability amplified by the strong La Nina conditions at the time [Fuelberg et al., 2001]. La Nina conditions also shifted the SPCZ westward of its typical position [Fuelberg et al., 2001]. This shift created a window in the vicinity of the dateline where northern hemispheric air could travel into the Southern Hemisphere near the surface. Transport of northern hemispheric midlatitude anthropogenic effluents around the North Pacific High into the trade winds and through this window has been termed the “river of pollution” [Staudt et al., 2001].
Figure 2 (top) shows the observed distributions of NO, HNO3, and PAN as a function of latitude and altitude during PEM-Tropics B. Longitudinal variations were relatively small. The observations have been averaged across all longitudes and into 0.5° latitude by 0.5 km altitude boxes. The nitrogen oxide radicals (NO and NO2) were measured on the DC-8 using the photofragmentation two-photon laser-induced fluorescence technique [Bradshaw et al., 1999]. The NO2 concentrations measured by this method have been shown to match closely the expected photochemical equilibrium with NO [Bradshaw et al., 1999; Schultz et al., 1999]. On the P-3B, NO and NO2 were measured by chemiluminescence using a method similar to that described by Ridley et al. . We focus here on the measurements of NO, which were made with greater precision. Nitric acid was measured with a mist chamber technique [Talbot et al., 1999], with temporal resolution ranging from 4 min near the surface to 8 min in the upper troposphere. Peroxyacetyl nitrate was measured on the DC-8 aircraft only, using gas chromatography with electron capture detection from a cryogenically enriched sample of ambient air [Singh et al., 2001]. Eighty percent of the PAN observations below 1 km altitude were below the 1 pptv limit of detection [Olson et al., 2001].
 The observations of NO in the Southern Hemisphere show a strong vertical gradient with higher concentrations in the upper troposphere (Figure 2). From the surface up to 4–6 km altitude, NO concentrations were generally less than 5 pptv. Enhanced NO concentrations at altitudes greater than 9 km above the southern and equatorial Pacific were likely the result of lightning, which was active in the warm pool area of the southwest equatorial Pacific, but also more generally over the South Pacific [Fuelberg et al., 2001]. Somewhat lower NO concentrations were observed in the upper troposphere above the North Pacific (Figure 1), which was more distant from areas of active convection and lightning.
 Low concentrations of HNO3 were observed in the marine boundary layer above the southern and equatorial Pacific, reflecting deposition and uptake by sea salt (J. E. Dibb et al., Soluble N and S relationships in the equatorial Pacific, submitted to Journal of Geophysical Research, 2001, hereinafter referred to as Dibb et al., submitted manuscript, 2001). Nitric acid was also depleted in the upper troposphere at 15°S–10°N, due to the efficient scavenging in convective updrafts [Mari et al., 2003]. Nitric acid concentrations in the upper troposphere increased south of 20°S due to less active convection. The highest concentrations of HNO3 were sampled in the lower troposphere of the northern subtropics, where other tracers indicated a strong Eurasian pollution influence [Staudt et al., 2001; Blake et al., 2001]. Bulk filter measurements of aerosol nitrate (NO3−) during PEM-Tropics B (Dibb et al., submitted manuscript, 2001) had relatively sparse spatial coverage and were frequently below the limit of detection (typically 7–17 pptv). As shown in Figure 3, the ratio of total nitrate (HNO3(gas) + NO3−(aerosol)) to HNO3(gas) was typically 1–2 for the ensemble of data where NO3− concentrations were above the detection limit.
 Observed PAN concentrations generally increased with altitude (Figure 2), reflecting the strong temperature dependence of PAN loss. Concentrations were less than 10 pptv throughout the lower troposphere. In equatorial regions, rapid convective mixing of air between the boundary layer and the free troposphere depressed PAN concentrations throughout the column. At 10–20°S, PAN concentrations in the upper troposphere were generally less than 20 pptv while midtropospheric (4–8 km altitude) concentrations were higher, suggesting a strong coupling between the lower and upper troposphere by deep convection. PAN was elevated in the middle and upper troposphere of the northern subtropics, reflecting biomass burning outflow from the Indian subcontinent and Southeast Asia as seen for other tracers [Staudt et al., 2001; Blake et al., 2001]. Indeed, PAN concentrations observed at 4–8 km altitude over the North Pacific were well correlated with CH3Cl and CO, tracers of biomass burning (Figure 4).
 The concentrations of NOx, HNO3, and PAN observed over the tropical South Pacific (10–30°S) during PEM-Tropics B were a factor of 2 or more lower than those observed during PEM-Tropics A when a pervasive biomass burning influence was present, as illustrated in Table 1 [also see Talbot et al., 2000]. Over the equatorial Pacific, the difference is less because biomass burning influence during PEM-Tropics A was largely confined south of 10°S [Gregory et al., 1999]. Concentrations of NO and PAN in the lower troposphere over the equatorial Pacific were actually higher during PEM-Tropics B, when the river of Eurasian pollution [Staudt et al., 2001] carried northern midlatitude effluents as far south as the SPCZ. During PEM-Tropics B, HNO3/NOx concentration ratios in the upper troposphere above the South Pacific were a factor of 3 less than those observed during PEM-Tropics A, indicating a larger chemical imbalance in the NOx budget for this region.
Table 1. Reactive Nitrogen Species Concentrations in the Tropical Pacific Tropospherea
Upper Troposphere (6–12 km)
Lower Troposphere (0–6 km)
Median and interquartiles (values in parentheses) observed for 28 August to 6 October 1996 (PEM-Tropics A) and 1 March to 19 April 1999 (PEM-Tropics B). The observations have been averaged over the HNO3 measurement interval.
The sum of observed NO and modeled NO2 for PEM-Tropics A and of observed NO and NO2 for PEM-Tropics B.
 We use the Harvard/Florida State University (FSU) global chemical transport model (CTM) to investigate the factors affecting tropospheric NOx over the remote Pacific during March–April 1999. This model has been previously applied to simulate tropospheric O3-NOx-hydrocarbon chemistry during PEM-Tropics A [Staudt et al., 2002], to simulate sources and transport of CO during PEM-Tropics B [Staudt et al., 2001], and to evaluate our understanding of the sources of oxygenated organics during PEM-Tropics B [Singh et al., 2001]. It is driven by assimilated meteorological fields produced by the FSU Global Spectral Model [Krishnamurti et al., 1993, 1996] for 22 January to 19 April 1999. A detailed treatment of tropical dynamics makes the FSU analysis especially useful for studies of the tropical Pacific. In particular, the precipitation rates used in the physical initialization process are derived from satellite observations of outgoing long-wave radiation and precipitation [Krishnamurti et al., 1983; Gairola and Krishnamurti, 1992]. The CTM has 14 vertical layers along sigma coordinates (centered at 0.99, 0.95, 0.9, 0.85, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1, 0.07, 0.03) and a horizontal resolution of 4° latitude by 5° longitude.
 Transport processes in the model include grid-scale advection using a second-order moments scheme [Prather, 1986], dry convection which redistributes tracer uniformly up to the mixed layer height calculated by the FSU model, and deep moist convection using convective mass fluxes diagnosed in the FSU model as described by Staudt et al. . Because the deep convective adjustment process in the FSU model is not easily adapted to tracer transport, the convective mass fluxes are not used in the FSU model but are parameterized in a postprocessing step. To account for possible errors in this parameterization step, we reduce the convective mass fluxes used in this analysis by 25% compared to those used by Staudt et al.  and Singh et al. . This correction improves the simulation of HNO3 as discussed in section 5. Dry deposition is as described by Bey et al. [2001a]. Nitric acid and H2O2 are scavenged by liquid and solid precipitation in convective updrafts and in large-scale systems as described by Liu et al. . This includes a scavenging efficiency of 40% km−1 in wet convective updrafts, which we reduce here to 20% km−1 for temperatures less than −15°C. There is substantial uncertainty regarding the uptake of HNO3 by ice, with some studies indicating quantitative uptake [Zondlo et al., 1997; Abbatt, 1997] and others suggesting column scavenging efficiencies of 10% or less [Tabazadeh et al., 1999; Meilinger et al., 1999; Clegg and Abbatt, 2001]. We discuss the sensitivity of the model results to the assumed scavenging efficiencies in section 5.
 Chemical perturbation to the NOx budget from dust aerosols is also an important consideration for our simulation. Maximum export of desert dust from Asia occurs during spring [Ginoux et al., 2001] and possibly impacts tropospheric chemistry over the North Pacific [Dentener et al., 1996; Tabazadeh et al., 1998; He and Carmichael, 1999; Zhang and Carmichael, 1999; Martin et al., 2002]. We have implemented radiative and chemical effects of dust in the model as described by Martin et al.  using monthly mean fields of mineral dust from Ginoux et al. . Partitioning of HNO3 between gas and aerosol phases is not resolved in the model, and it is assumed for chemical reactivity purposes that HNO3 is in the gas phase. Sequestration of HNO3 in the aerosols could affect the NOx budget if the gas-phase conversion rate of HNO3 to NOx differs largely from the parallel process involving aerosol nitrate, but the related aerosol chemistry is highly uncertain [Jacob, 2000].
 The model simulates 120 species to describe tropospheric O3-NOx-hydrocarbon chemistry, including 24 chemical tracers transported in the model. The chemical mechanism is that of Horowitz et al.  with minor updates as discussed by Bey et al. [2001a], and is integrated with a fast Gear solver [Jacobson and Turco, 1994]. The rate of the CH3O2 + HO2 reaction has been increased here by a factor of 3 at low temperature (220 K) relative to the standard recommendation [DeMore et al., 1997], following the analysis of the HOx budget in PEM-Tropics B by Ravetta et al. . Heterogeneous reactions involving aerosols use reaction probabilities from Jacob  with sulfate aerosol fields from Chin et al.  and dust aerosol fields as described above. Photolysis rates are computed with the fast J radiative transfer code of Wild et al. , including Mie scattering by clouds with vertically resolved cloud optical depths from the FSU meteorological fields.
 Emissions of CO, NOx, and hydrocarbons from fossil fuel combustion and industry are those described by Wang et al.  for 1985 and are scaled to 1995 levels, the most recent year for which national emission estimates could be obtained, as described by Bey et al. [2001a]. Emissions of NOx from lightning are parameterized from the FSU cloud top data following Price and Rind  and Pickering et al.  and are adjusted to yield a global source of 6 Tg N yr−1, consistent with the Martin et al.  model analysis of tropical tropospheric ozone. The sensitivity of results to the magnitude of the lightning source will be examined. Isoprene is emitted from terrestrial vegetation using the Guenther et al.  scheme. Sources of acetone and CH3CHO are as described by Wang et al.  and include oxidation of hydrocarbons and emissions from biomass burning and biofuels. Acetone is also emitted by vegetation and industry. A minimum acetone concentration of 400 pptv is assumed, based on the PEM-Tropics B observations of Singh et al.  and presumably reflecting an oceanic source [Jacob et al., 2002]. Other biogenic emissions of organic species are assumed to play little photochemical role but are included as sources of CO [Duncan et al., 2003].
 A particularly important issue for our simulation is the representation of biofuel use and biomass burning in Asia. We have implemented here a new inventory for emissions of CO from biofuel use with 1° × 1° spatial resolution from Yevich and Logan . We have also included biofuel emissions of NOx and hydrocarbons using emission factors from Andreae and Merlet . Biomass burning emissions for the PEM-Tropics B period are specified using fire count observations from the Along Track Scanning Radiometer-2 (ATSR-2) satellite [Arino and Rosaz, 1999] to scale a climatological inventory of biomass burned from R. M. Yevich and J. A. Logan (personal communication, 2001) to 1999 levels [Duncan et al., 2003]. We apply emission factors (Table 2) for different vegetation types mostly following the recommendations of Andreae and Merlet  to obtain emissions for CO, NOx, and hydrocarbons.
 The distributions of NOx emissions in the model from lightning, fossil fuel, biofuel, and biomass burning are shown in Figure 5, and regional emission totals for these different sources are summarized in Table 3. The largest NOx source is from fossil fuel burning with comparable contributions from North America, Europe, and Asia. The lightning source is concentrated over continents in the Southern Hemisphere and over Indonesia, but there is also some activity over the South Pacific. Maximum biomass and biofuel burning emissions are from Southeast Asia and the Indian subcontinent.
Table 3. Global Emissions of NOx for 1 March to 19 April 1999a
North and Central America
Europe and Northern Africa
Asia and Indonesia
Units of Tg N.
 The fluxes from the stratosphere of O3, NOx, and HNO3 contribute to the budgets of these species in the troposphere. The FSU model does not have the vertical resolution needed for an accurate simulation of stratospheric dynamics. We therefore use a cross-tropopause flux boundary condition for O3 at 100 hPa altitude as described by Wang et al.  with an annual global flux of 400 Tg yr−1 distributed by latitude and month of the year. We also use a flux of 0.48 Tg N yr−1 of NOy computed by Wang et al.  which is transported across the tropopause as NOx and HNO3 with a molar ratio of 1:4.
 Simulations are conducted for a 3-month period from 22 January to 19 April 1999. The model is initialized on 21 January with mean January concentrations from the full-year CTM simulation of Wang et al.  using general circulation model meteorology. Simulation through 1 March provides initialization of the transport patterns in the model, and we focus our attention on the 1 March to 19 April period.
4. Evaluation of Standard Simulation
 The CO concentrations produced by the model were discussed by Staudt et al.  and found to reproduce the major vertical, latitudinal, and zonal gradients in the observations with no significant biases. In particular, the model reproduces the decreasing concentrations of CO with altitude in the Northern Hemisphere, the increasing concentrations with altitude in the Southern Hemisphere, and the longitudinal gradients above the equatorial Pacific driven by interhemispheric transport. Enhanced concentrations were observed and simulated in the lower troposphere above the western equatorial Pacific, where the river of pollution brought effluents from northern midlatitudes southward to about 10°S. In the upper troposphere, relatively high concentrations were found above the eastern equatorial Pacific, where northern hemispheric air was carried southward through a region of equatorial northwesterlies and into the general circulation of the Southern Hemisphere.
 The model simulation of oxygenated organic species, including formaldehyde (HCHO), PAN, acetone, and CH3CHO was presented by Singh et al. . The model reproduces the vertical profile of HCHO well, with maximum underestimates of 40% above 9 km. The simulation of PAN, which will be discussed in more detail below, overestimates observations above the northern tropical Pacific and underestimates observations above the southern tropical Pacific. Simulated acetone concentrations as presented by Singh et al.  were much lower than observed, and the imposition here of a minimum concentration of 400 pptv provides an ad hoc correction. A more thorough investigation of acetone sources and sinks using the PEM-Tropics B observations is presented by Jacob et al. .
 Acetaldehyde concentrations over the South Pacific are underestimated by the model; simulated concentrations are 1–30 pptv, while the observations are 50–120 pptv and have little vertical gradient [Singh et al., 2001]. No such underestimate is found when comparing model results with observations at continental sites (0.1–1 ppbv) [Granby et al., 1997; Riemer et al., 1998; Christensen et al., 2000]. With a lifetime of about 1 day due to reaction with OH and photolysis, long-range transport of CH3CHO from continental source regions to the remote Pacific troposphere is minimal. Singh et al.  suggest that oxidation of nonmethane hydrocarbons not currently included in the model or photochemical degradation of organic material present in surface oceans may provide additional sources for CH3CHO. However, imposing CH3CHO concentrations in the model at the levels observed by Singh et al.  would lead to severe discrepancies in the simulation of PAN and NOx, as discussed in section 6.
Figure 6 shows comparisons of model results (sampled along the flight tracks and for the flight days) with observed concentrations of O3, OH, HO2, NO, HNO3, and PAN, as well as HNO3/NOx and PAN/NOx concentration ratios, as a function of altitude and latitude. The model is largely successful at reproducing the observations over the tropical South Pacific. The low observed concentrations of O3, NO, HNO3, and PAN are well simulated south of 10°N. Rates of O3 production and loss are similar to those obtained by Olson et al.  with a photochemical point model constrained by local PEM-Tropics B observations.
 The model is less successful at reproducing the observations north of 10°N, where O3 is overestimated by about 50%, NO by as much as a factor of 2, and HNO3 and PAN by a factor of 2–5. We think that the model overestimates over the North Pacific are caused by excessive vertical mixing at high latitudes in the model which brings air from the upper troposphere into the lower tropospheric flow around the North Pacific High. This dynamical problem in the model will affect species that have large vertical gradients in the far northern latitudes, including O3, HNO3, and PAN. The dynamically induced errors in these species in turn cause overestimates when they react above the tropical North Pacific. For example, high O3 is responsible for OH overestimates near the surface north of the equator and high HNO3 and PAN cause NO to be too high at all altitudes north of about 10°N. The overestimate of O3 above the North Pacific is largely insensitive to emissions from biomass burning, fossil fuel, and lightning. Likewise, we found that neither realistic variations in the initial conditions nor a zero flux boundary condition from the stratosphere provided much improvement in the simulation.
 The HOx species (OH and HO2) are well simulated in the lower troposphere, but are about 50% too high in the upper troposphere (Figure 6) with little dependence on latitude. Point photochemical model results of Ravetta et al.  similarly overestimate OH and HO2 observed above 8 km altitude during PEM-Tropics B, while other models overestimate HO2 and underestimate OH [Davis et al., 2001; Tan et al., 2001; Olson et al., 2001; J. M. Rodriguez et al., manuscript in preparation, 2001]. Increasing the rate of the CH3O2 + HO2 reaction at low temperatures, as was done in our standard simulation following Ravetta et al.  decreases OH and HO2 concentrations in the upper troposphere by 4% and 10%, respectively.
 We find that mineral dust and sulfate aerosols exported from Asia across the North Pacific [Ginoux et al., 2001; Moore et al., 2003; Clark et al., 2001] have only a modest effect on concentrations of NO, HNO3, PAN, and O3. Including mineral dust in the model decreases NOx and PAN by up to 5 pptv in the middle and upper troposphere over the northeastern tropical Pacific, bringing the model into better agreement with the observations. Nitric acid in this region increases by 5–10 pptv because conversion from NOx to HNO3 is enhanced. The model HNO3 is higher than the observations even after factoring in the contribution of NO3− observed in aerosols to the total observed HNO3 (Figure 6). In the lower troposphere, NOx, PAN, and HNO3 decrease by 5–10 pptv when dust is added.
 The simulated concentrations of NO over the equatorial and southern Pacific (Figures 2 and 6) are generally within the range of the observed variability and reproduce the observed increase with altitude. The model underestimates NO in the equatorial marine boundary layer, where point photochemical model calculations constrained with the aircraft observations indicate that photolysis of methyl nitrate emitted from the ocean (not included in the 3-D model) can supply a few pptv NOx. Simulated and observed concentrations of HNO3 and PAN are minimum near the surface and in the equatorial upper troposphere where convection is active, resulting in midtropospheric maxima.
 The simulated concentration ratios for HNO3/NOx and PAN/NOx generally capture observed concentrations (Figure 6). Discrepancies for HNO3/NOx are in the middle and upper troposphere south of 15°S, where the model is too low, and above 4 km altitude north of the equator, where the model is too high. Photochemical equilibrium models generally overestimate greatly the HNO3/NOx ratio in the remote troposphere, which has led to speculation about a missing chemical mechanism to convert HNO3 to NOx [Chatfield, 1994; Hauglustaine et al., 1998; Lary et al., 1997]. We find no support here for such a mechanism.
5. Sensitivity to Convective Mixing and Scavenging
 Concentrations of NOy species (NOx, HNO3, and PAN) in the tropics are very sensitive to convection which provides an efficient sink through two pathways: (1) by bringing air rapidly down to the marine boundary layer where HNO3 deposits to the surface and PAN thermally decomposes, and (2) by scavenging of HNO3 via precipitation in the convective updrafts. If either of these processes is too strong in the model, then the simulated concentrations of HNO3 and PAN and the simulated concentration ratio HNO3/NOx will be too low. We find that when scavenging of HNO3 by ice in the model is allowed to proceed at the same rate as scavenging by liquid precipitation, HNO3 above 6 km altitude is underestimated by 10–20 pptv over the South Pacific (Figure 7). Decreased scavenging of HNO3 by ice improves the simulation over the South Pacific by increasing the concentrations of the NOy species and the concentration ratio of HNO3/NOx throughout the column. When scavenging by ice is more rapid, HNO3 concentrations are up to 50 pptv lower in the middle and upper troposphere over the North Pacific. Although the large overestimates of HNO3 make it difficult to determine the appropriate level of scavenging, the large sensitivity of simulated HNO3 to the scavenging parameterization points to a need to improve our understanding of the relative roles of ice and liquid scavenging. Indeed, adjusting the rate of scavenging by ice by 25% has a similar impact on the HNO3 concentrations as decreasing the lightning source by 50% as discussed in section 7.
 We find in the standard version of the model (in which the convective mass fluxes are reduced by 25% from Staudt et al. ) that the overturning of air in the 7.5–17 km altitude band over the South Pacific (2°N–30°S, 165°E–105°W) by convection-driven subsidence is 6% d−1. The corresponding residence time of 15 days is consistent with other models of the tropical troposphere, providing support for the 25% reduction in convective mass fluxes. Using CH3I observations, Wang et al.  estimated a 10-day convective turnover timescale for the tropospheric column during PEM-Tropics B and a 20-day convective turnover timescale for PEM-Tropics A. Prather and Jacob  derived a 20-day convective turnover timescale for 8–17 km altitude in the tropics using a general circulation model.
 In simulations where the rate of convective turnover is not reduced by 25%, simulated HNO3 concentrations are depressed by 10–20 pptv throughout the tropospheric column over the South Pacific and up to 50 pptv over the North Pacific, causing a corresponding decrease in simulated HNO3/NOx (Figure 7). The concentration ratio of CH3OOH/H2O2 (not shown) is another indicator of convection in the upper troposphere because CH3OOH is elevated in convective outflow while H2O2 is depleted due to scavenging [Cohan et al., 1999]. With more rapid convective turnover, CH3OOH/H2O2 is overestimated by 25% over the South Pacific and by as much as a factor of 2 over the North Pacific.
6. Sensitivity to Oxygenated Organics
 Because the model does not include an oceanic source for acetone, it simulates acetone concentrations over the South Pacific (100–200 pptv) that are much lower than observed (400 pptv) [Singh et al., 2001]. An ad hoc correction was applied in the standard simulation by imposing a minimum concentration of 400 pptv everywhere. When acetone is not constrained in this way, PAN and PAN/NOx are underestimated over the South Pacific, as shown in Figure 8.
 Despite improvements achieved with fixed acetone, the model still underestimates the PAN/NOx concentration ratio in the midtroposphere and near the surface over the South Pacific region. Because the majority of PAN measurements near the surface were below the 1 pptv detection limit, the mean observed PAN and PAN/NOx shown in Figure 8 are too high. The model underestimate of PAN/NOx in the midtroposphere, however, may indicate that simulated concentrations of the peroxy acetyl radical (CH3C(O)OO) are too low there. As mentioned previously, the model substantially underestimates the PEM-Tropics B observations of CH3CHO, an important PAN precursor, while producing concentrations in the continental boundary layer comparable to prior observations.
 To test the impact of higher levels of CH3CHO on the NOx chemistry, we conducted a sensitivity simulation in which CH3CHO was set to a minimum concentration of 80 pptv to match the PEM-Tropics B observations of Singh et al.  without perturbing the simulated distribution over continents where no bias with observations is apparent. When CH3CHO concentrations are constrained, the simulated PAN concentrations and PAN/NOx concentration ratios are 2–6 times higher than observed in the middle and upper troposphere, while still underestimating PAN/NOx near the surface (Figure 8). In addition, the HNO3/NOx ratio in the lower troposphere is decreased by up to a factor of 2 reflecting the enhanced release of NOx from PAN transported into the region. These large discrepancies between observed and simulated PAN indicate that our current understanding of NOy chemical cycling appears to be incompatible with the free tropospheric observations of CH3CHO during PEM-Tropics B. Only near the surface does increased CH3CHO appear consistent with the NOx and HOx chemistry, suggesting an oceanic source may be present.
7. Constraining the Lightning NOx Source
 Considerable uncertainty is associated with the magnitude, location, vertical distribution, and timing of NOx emissions from lightning. We conducted two simulations with lightning emissions reduced by 50% to 3 Tg N yr−1 (dashed lines in Figure 6) and increased by 50% to 9 Tg N yr−1 (dotted lines) to see if the PEM-Tropics B observations could be used to constrain the magnitude of this source. Since the lightning source is located primarily over tropical continents, the changes in concentrations simulated over the tropical Pacific reflect the impact of aged emissions. Decreasing lightning improves the simulation of upper tropospheric OH and of upper tropospheric HNO3 in the Northern Hemisphere. On the other hand, increasing lightning improves the simulation of HNO3 in the upper troposphere south of 20°S, while exacerbating the HNO3 overestimate above the North Pacific. Other species are not much impacted by changing the lightning NOx emissions.
 Lightning emissions have a rather small effect on HNO3/NOx. Increasing lightning reduces the discrepancy between model and observed HNO3/NOx over the South Pacific, but aggravates model overestimates over the North Pacific. We find that increasing the lightning source actually causes a slight increase in the simulated HNO3/NOx above the tropical Pacific, contrary to what might have been expected. The additional NOx from lightning increases OH (Figure 6), which shortens the NOx lifetime. The HNO3/NOx ratio is more sensitive to the rate of convective overturning and associated scavenging than to the source of NOx from lightning (see section 5).
 The fact that increased emission of NOx from lightning improves the simulation above the tropical South Pacific while degrading the simulation above the tropical North Pacific might suggest a model error in the distribution of lightning emissions. However, we find that the geographic distribution (Figure 5) is consistent with satellite observations. Observations of lightning strikes from the satellite-based Lightning Imaging Sensor (LIS) indicate maximum lightning in the southern tropics during March and a northward shift during April [Fuelberg et al., 2001]. Using observations from the satellite-based Optical Transient Detector (OTD) to derive NOx emissions from lightning, Nesbitt et al.  similarly find maximum emissions over the southern tropics during December–January–February, shifting to a maximum extending from 10°S to 40°N during March–April–May. Maximum emissions of NOx from lightning in our model for 1 March 1999 to 15 April 1999 are likewise found between 20°S and 10°N.
8. NOx Budget for the Tropical Pacific Troposphere
 Primary sources for NOx in the Southern Hemisphere during the PEM-Tropics B period are dominated by lightning followed by fossil fuel combustion and soils (Table 3). The impact of these sources on concentrations of NOx simulated over the tropical South Pacific (region a in Figure 1) is illustrated in Figure 9 using results from sensitivity simulations with individual sources shut off. At 8–12 km altitude, shutting lightning off reduces NOx concentrations over the tropical South Pacific by 76%, while HNO3 and PAN (not shown) decrease by 86% and 65%, respectively. Ozone and OH likewise show reductions in the upper troposphere of 48% and 45%, respectively, when lightning is turned off. Lightning remains the largest primary source for NOx, HNO3, and PAN at 2–8 km altitude. Near the surface, fossil fuel makes a larger contribution to simulated NOx and HNO3 concentrations, although lightning still accounts for about half of the simulated HNO3.
 The primary sources of NOx in the Northern Hemisphere during PEM-Tropics B are estimated to be a factor of 5 greater than those in the Southern Hemisphere (Table 3). Anthropogenic emissions from fossil fuel, biofuel, and aircraft contribute 68% of the northern hemispheric source, while lightning contributes only 5%. Nevertheless, as illustrated in Figure 9, lightning is the most important source of NOx in the upper troposphere over the northern tropical Pacific. Concentrations of HNO3 and, to a lesser extent, PAN are also enhanced by NOx emissions from lightning (not shown). This result is robust because the excessive transport from the high northern latitudes would tend amplify the influence of Northern Hemispheric sources. Concentrations of PAN in the upper and midtroposphere are more sensitive to biomass burning emissions, which only affect NOx and HNO3 in the 2–8 km altitude range. Below 4 km, the influence of fossil fuel emissions is dominant (Figure 9), consistent with past studies that found boundary layer air observed above the North Pacific during PEM-Tropics B to have strong Asian and European pollution influence due to circulation around the Pacific High and into the trade winds [Moxim et al., 1996; Blake et al., 2001; Staudt et al., 2001]. We find here that this pollution influence extends to NOx.
 The NOx budget above the South Pacific is given in Table 4 for the standard simulation. The model limitations over the North Pacific and the smaller area spanned by the aircraft make it imprudent to consider the budget of that region. For the tropical South Pacific, the NOx source to the lower troposphere is mostly from decomposition of PAN (74%), with secondary contributions from decomposition of HNO3 (11%) and transport of NOx into the region (11%). The chemical sources and sinks of NOx in the lower troposphere are close to steady state because the budget is dominated by cycling with HNO3 and PAN.
Table 4. NOx Budget for the Southern Tropical Pacific Tropospherea
Lower Troposphere 0–6 km
Upper Troposphere 6–12 km
Budget terms are model averages from the standard simulation for a region defined as 2°N–30°S and 165°E–100°W (an area of 3.4 × 1013 m2) and for the PEM-Tropics B time period of 1 March to 19 April 1999. For the purposes of this budget, we use an expanded chemical family definition of NOx (NOx = NO + NO2 + NO3 + HNO2 + HNO4 + 2*N2O5).
Includes the products of nonmethane hydrocarbon oxidation but no contribution from oceanic emissions.
Photochemical point model calculation by Ravetta et al.  for the ensemble of PEM-Tropics B flight tracks.
 The NOx sources to the upper troposphere above the South Pacific are more distributed, with contributions from transport into the region (34%), emissions from lightning (19%), and chemical recycling from HNO3 (23%) and PAN (23%). The chemical imbalance in the upper troposphere, diagnosed by the ratio of NOx chemical losses (LNOx) to NOx chemical sources (PNOx) in the 3-D model, is equal to 2.0, compared to an imbalance of 1.6 computed in the 3-D model for the PEM-Tropics A time period [Staudt et al., 2002]. The larger chemical imbalance during PEM-Tropics B is confirmed by a lower HNO3/NOx concentration ratio in the observations (Table 1) and the model (1.9 for PEM-Tropics A, 1.5 for PEM-Tropics B). Similar to PEM-Tropics A, the imbalance during PEM-Tropics B is largely maintained by the input of fresh NOx from lightning into the upper troposphere, where the NOx lifetime is about 10 days. Turning off the source of NOx from lightning reduces LNOx/PNOx from 2.0 to 1.3. For the same region during PEM-Tropics A, Staudt et al.  found that recycling of HNO3 and PAN contributed 63% of the NOx source. Chemical recycling was more important during PEM-Tropics A because the NOx source from biomass burning, which had aged 1–2 weeks by the time it reached the Pacific, was much larger while the NOx source from in situ lightning emissions was a factor of 2 less. In addition, more rapid convective overturning during PEM-Tropics B [Wang et al., 2001] created a larger chemical imbalance in the NOx budget by enhancing HNO3 loss.
 We examined the sources and chemistry affecting NOx over the tropical Pacific during the PEM-Tropics B aircraft campaign March–April 1999 by using a global 3-D model of tropospheric chemistry driven by assimilated meteorological data. The model reproduces the particularly low concentrations of NO (1–5 pptv), HNO3 (20–50 pptv), and PAN (<1–10 pptv) observed in the lower troposphere above the equatorial and southern Pacific. Low concentrations of HNO3 (10–50 pptv) and PAN (10–30 pptv) observed in the upper equatorial troposphere, because of convection linking it to the lower troposphere, are also reproduced by the model, resulting in midtroposphere maxima for these species. Simulated and observed NO concentrations are maximum in the upper troposphere (10–100 pptv), due to emissions from lightning. The model is less successful over the North Pacific, where excessive vertical mixing in the high latitudes brings air with elevated O3, HNO3, and PAN into the lower tropospheric flow around the North Pacific High.
 Concentrations of NO, HNO3, and PAN observed during PEM-Tropics B are generally lower than those observed over the South Pacific during PEM-Tropics A over the same region but in September–October 1996. PEM-Tropics A was conducted in the biomass burning season of the southern tropics and long-range transport of combustion effluents elevated NOx, HNO3, and PAN concentrations throughout the South Pacific atmosphere. PEM-Tropics B was conducted in the wet season of the southern tropics but the biomass burning season of the northern tropics, and this combustion influence was observed over the North Pacific superimposed to that from fossil fuel combustion.
 Lightning was the dominant source of NO, HNO3, and PAN above the equatorial and southern Pacific during PEM-Tropics B. Turning off the source of NOx from lightning in the model decreases the NOx source to the region by a factor of 7 in the upper troposphere and a factor of 3 in the lower troposphere. Lightning contributed most of the NOx in the upper troposphere; lightning, biomass burning, and fossil fuels all made significant contributions in the middle troposphere, while fossil fuel combustion was the most important source of NOx to the lower troposphere.
 We find that the simulation of HNO3 and HNO3/NOx over the tropical Pacific is very sensitive to the rate of convective overturning and convective scavenging of HNO3. The simulation of HNO3 and HNO3/NOx is improved when the scavenging efficiency of HNO3 by ice in the convective updrafts is less than that by liquid precipitation. Increasing the convective turnover rate by 25% causes HNO3 to decrease by 10–20 pptv above the South Pacific and by up to 50 pptv above the North Pacific. The HNO3/NOx concentration ratio is lower when convection is more vigorous.
 PEM-Tropics B provided the first observations of oxygenated organics in the tropical marine atmosphere. Average concentrations observed for acetone (431 pptv) and CH3CHO (78 pptv) were much higher than simulated by the model without an oceanic source. To account for such an oceanic source we constrained acetone in the standard simulation to a minimum value of 400 pptv, which provided an improved simulation of PAN and PAN/NOx. When we also constrained CH3CHO to a minimum value of 80 pptv, simulated PAN concentrations and PAN/NOx concentration ratios were several times higher than observed. The high observed CH3CHO concentrations above the South Pacific can not be reconciled with our understanding of PAN chemistry.
 During PEM-Tropics B, chemical losses of NOx exceeded chemical sources by a factor of 2 at 6–12 km altitude above the equatorial and southern Pacific; a chemical imbalance also apparent in the observed HNO3/NOx ratios. Lightning emissions and convective overturning were the main causes for this chemical imbalance. There was no evidence for a missing chemical pathway recycling HNO3 to NOx. During PEM-Tropics A, NOx chemical sinks exceeded chemical sources in the South Pacific upper troposphere by only 60%, and HNO3/NOx concentration ratios were correspondingly higher than in PEM-Tropics B, because the large biomass burning contribution was close to chemical steady state and convection was less active. Observed HNO3 concentrations and HNO3/NOx concentration ratios in this region were a factor of 3 less during PEM-Tropics B than during PEM-Tropics A.
 This research was funded by the NASA Tropospheric Chemistry Program (TCP).