Ozone production in the upper troposphere and the influence of aircraft during SONEX: Approach of NO_x-saturated conditions

L. Jaeglé and  D.J. Jacob
Harvard University, Cambridge, Massachusetts

W.H. Brune, I.C. Faloona, and D. Tan
Pennsylvania State University, University Park

Y. Kondo
Nagoya University, Japan

G.W. Sachse, B. Anderson, G.L. Gregory, and S. Vay
NASA Langley Research Center, Hampton, Virginia

H.B. Singh
NASA Ames Research Center, Moffett Field, California

D.R. Blake
University of California, Irvine, California

R. Shetter
National Center for Atmospheric Research, Boulder, Colorado
 

Abstract. During October/November 1997, simultaneous observations of NO, HO₂ and other species were obtained as part of the SONEX campaign in the upper troposphere. We use these observations, over the North Atlantic (40-60°N), to derive ozone production rates, P(O₃), and to examine the relationship between P(O₃) and the concentrations of NO_x (= NO + NO₂) and HO_x (= OH + peroxy) radicals. A positive correlation is found between P(O₃) and NO_x over the entire data set, which reflects the association of elevated HO_x with elevated NO_x injected by deep convection and lightning. By filtering out this association we find that for NO_x>70 pptv, P(O₃) is nearly independent of NO_x, showing the approach of NO_x-saturated conditions. Predicted doubling of aircraft emissions in the future will result in less than doubling of the aircraft contribution to ozone over the North Atlantic in the fall. Greater sensitivity to aircraft emissions would be expected in the summer.

• Introduction
• Calculation of ozone production
• Relationships between NO_x, HO_x, and ozone production
• Chemical regime for ozone production
• Conclusions

Introduction

In recent years, considerable attention has been given to the potential role of aircraft emissions of nitrogen oxides (NO_x = NO + NO₂) on the concentration of upper tropospheric ozone, an effective greenhouse gas [NASA, 1997]. Ozone is produced in the troposphere by the photochemical oxidation of CO and hydrocarbons which is catalyzed by NO_x radicals and hydrogen oxide radicals (HO_x = OH + peroxy). Oxidation of CO dominates in the upper troposphere, and the rate-limiting step for ozone production is the reaction of HO₂ with NO (R2):
 
 

The sensitivity of the ozone production rate, P(O₃), to increasing NO_x is critical to the assessment of aircraft effects. Photochemical models [e.g., Brasseur et al., 1996] predict that P(O₃) should increase with increasing NO_x (NO_x-limited regime) up to a turnover point of a few hundred pptv NO_x, beyond which further increases in NO_x cause P(O₃) to decrease (NO_x-saturated regime). These two regimes result from the dual role of NO_x in regulating the chemistry of HO_x radicals. On the one hand, NO_x drives the ozone production cycle (R1)-(R3). On the other hand, NO_x promotes the removal of HO_x through reactions of OH with HO₂, HNO₄, and NO₂ [Wennberg et al., 1998]. The chemical regime for ozone production is largely determined by the relative magnitudes of the sources of HO_x and NO_x [Jaeglé et al., 1998].

Ozone production rates in the upper troposphere have been previously determined from simultaneous measurements of HO₂ and NO during three recent aircraft campaigns: ASHOE/MAESA, STRAT and SUCCESS [Folkins et al., 1997; Wennberg et al., 1998; Brune et al., 1998; Jaeglé et al., 1998]. Examination of the P(O₃) versus NO_x relationships indicated a much greater prevalence for NO_x-limited conditions than expected from models [Folkins et al., 1997; Jaeglé et al., 1998]. However, as we will see, interpretation of this relationship in terms of the chemical regime for ozone production can be biased by a commonality of sources for NO_x and HO_x.

We present here the first direct evidence of NO_x-saturated conditions for ozone production in the upper troposphere and show that a simple interpretation of the observed P(O₃) versus NO_x relationship as a partial derivative dP(O₃)/dNO_x overestimates the actual sensitivity of ozone concentrations to emissions from aircraft. We use concurrent observations of HO₂ and NO obtained during the Subsonic assessment: Ozone and NO_x Experiment (SONEX) DC-8 aircraft campaign. SONEX took place in October and November 1997 in the North Atlantic flight corridor, a region of dense aircraft traffic in the upper troposphere [Singh et al., this issue]. Companion papers use the SONEX data to improve our understanding of the chemistry and sources of NO_x [Kondo et al., this issue; Thompson et al., this issue], and HO_x [Brune et al., this issue].

Calculation of ozone production

Singh et al. [this issue] describe the flight tracks and the instruments aboard the DC-8 aircraft during SONEX. We focus here on observations made in the upper troposphere (8-12 km) between 40 and 60° N latitude, the main theater of operations. We exclude observations made in clouds (diagnosed by an abundance of particles larger than 3 m m), in fresh aircraft exhaust plumes (short duration peaks of elevated NO_x and condensation nuclei), at high solar zenith angles (>80°), and in air masses with stratospheric influence (O₃>90 ppbv and CH₄<1760 ppbv).

We define the budget of ozone as that of the odd-oxygen family, O_x (O_x=O₃ + O + O(¹D) + NO₂ + HNO₄ + HNO₃ + 2NO₃+ 3N₂O₅), to account for rapid chemical cycling within this family. Ozone typically accounts for over 99% of O_x, so the budgets of ozone and O_x can be viewed as equivalent. In addition to reaction (R2), ozone can be produced by the reaction of organic peroxy radicals, RO₂, with NO,

Ozone chemical loss is almost exclusively due to:

The ozone production and loss rates, P(O₃) and L(O₃), can thus be expressed as:
 
 

where k_i is the rate constant for reaction (Ri), and k₄ is weighted over the different RO₂ radicals contributing to ozone production. In the above expressions, we neglect the role of HNO₃+OH, which is a minor source of ozone (as per our definition of the O_x chemical family).

We use equations (1) and (2) with 1-minute averages of concurrent observations of NO, HO₂, OH, H₂O, O₃, UV actinic flux, aerosol surface area, temperature and pressure, to calculate instantaneous values of P(O₃) and L(O₃) along the flight tracks of the DC-8. For species which are not observed (RO₂, O(¹D), N₂O₅) we use calculations from a diel steady state model [Jaeglé et al., 1998] constrained with local observations of O₃, H₂O, NO, HNO₃, PAN, acetone, CO, CH₄, C₂H₆, C₃H₈, C₄ alkanes, UV actinic flux, aerosol surface area, temperature and pressure. The resulting instantaneous P(O₃) and L(O₃) rates are then scaled to 24-hour average values by using local results from the diel steady state model:
 
 

A more detailed description of the model as applied to SONEX observations can be found in Jaeglé et al. [Photochemistry of HO_x in the upper troposphere at northern midlatitudes, submitted to J. Geophys. Res., 1999, hereafter referred to as J99].

Relationships between NO_x, HO_x, and ozone production

Figure 1a shows the relationship between the 24-hour average values of P(O₃) derived from observed HO₂ and NO, and the local NO_x concentrations in the upper troposphere (8-12 km) between 40ºN and 60ºN latitude. The calculated median P(O₃) for SONEX was 0.57 ppbv/day. The median L(O₃) was 0.13 ppbv/day (not shown here), resulting in a net ozone production of 0.44 ppbv/day. Reaction of NO with HO₂ dominates ozone production; reaction (R4) contributes on average less than 15% of the total P(O₃). Reactions of O₃ with HO₂ and OH contribute more than 80% of L(O₃), while reactions (R7) and (R8) contribute 10% and 5% respectively.

The line in Figure 1a shows the expected dependence of ozone production on NO_x, for diel steady state model calculations where model input variables are specified from median background conditions observed during SONEX at 10 km [J99]: 55 ppbv O₃, 120 ppmv H₂O, 120 pptv HNO₃, 64 pptv PAN, 510 pptv acetone, 90 ppbv CO, 1761 ppbv CH₄, 670 pptv C₂H₆, 79 pptv C₃H₈, 55 pptv C₄ alkanes, 50ºN latitude, 285 DU ozone column, 8 µm²cm^-3 aerosol surface area, 227 K temperature, on November 1. In the model, P(O₃) becomes relatively independent of NO_x above 70 pptv; and the turnover to the NO_x-saturated regime (dP(O₃)/dNO_x=0) takes place at 300 pptv. The bulk of the observations (NO_x<300 pptv) shows indeed a leveling off of the dependence of P(O₃) on NO_x as NO_x increases above 70 pptv, in accordance with the expected behavior. However, for the highest NO_x concentrations, the values of P(O₃) computed from observed HO₂ and NO continue to increase with increasing NO_x, suggesting a consistently NO_x-limited regime which is at odds with model results.

The largest NO_x concentrations (>300 pptv) shown in Figure 1 correspond to relatively fresh convective outflows sampled close to the U.S. East coast [Thompson et al., this issue]. Elevated NO_x/NO_y ratios (>0.5 mol/mol), backtrajectory calculations and satellite lightning imagery all support a strong source of NO_x from lightning associated with this convection [Pickering et al., 1999]. These air masses were also characterized by an enhanced HO_x source resulting from convective transport of surface air with elevated concentrations of HO_x precursors such as peroxides and CH₂O [J99]. Based on (1), comparison between observed and modeled P(O₃) for a given NO_x concentration is roughly equivalent to comparison of observed and modeled HO₂ concentrations. We see from Figure 1b that the model constrained with background conditions for SONEX underestimates HO₂ by a factor of two or more when NO_x>300 pptv. Using the locally observed concentrations of HO_x precursors (H₂O, acetone, peroxides and CH₂O) for these high-NO_x points improves the agreement but still comes short of the observed levels. The discrepancy suggests the presence of other unmeasured sources, such as higher aldehydes, possibly also resulting from convection [Müller and Brasseur, 1999]. It could also reflect flaws in our understanding of HO_x chemistry in the high-NO_x regime [Faloona et al., 1999], or other unknown HO_x sources [Chatfield et al., 1999]. For the remaining observations (NO_x<300 pptv), the dependence of HO₂ and OH on NO_x is generally well reproduced (Figure 1b and Figure 1c). The scatter around the model lines in Figure 1 can be explained by variations in the magnitude of the local HO_x sources [J99].

In models of the upper troposphere, P(O₃) is largely determined by two variables: NO_x mixing ratios and the strength of the primary HO_x source, P(HO_x) [Jaeglé et al., 1998]. Figure 2 shows the variations of P(O₃) as a function of NO_x and P(HO_x). We separate primary sources (i.e. sources independent of HO_x) from secondary sources (i.e. sources dependent on a preexisting pool of HO_x), and define the primary HO_x source as:

where J_acet and y_acet are the photolysis rate constant and HO_x yield for acetone, and J_Xi and y_Xi are the photolysis rate constants and HO_x yields for other convected HO_x precursors such as peroxides and aldehydes (see Müller and Brasseur [1999] for this definition of the primary HO_x source). The value of y_acet is about three [Singh et al., 1995]. Methane oxidation by OH, and the subsequent photolysis of CH₂O was an important HO_x source during SONEX [J99]. We do not include this source in our definition of P(HO_x) because it is a secondary HO_x source.

In Figure 2a, we calculate P(HO_x) based on the observed H₂O and acetone. P(HO_x) is averaged over 24 hours using (3). In Figure 2b, in addition to H₂O and acetone, we include an additional HO_x source (Sy_XiJ_Xi[X_i]) as required to match the observed HO₂ concentrations. This source might include contributions from convected peroxides and aldehydes, which we cannot easily quantify from our model. Its impact on P(HO_x) is small (less than 50%) except for observations in continental convective outflows with elevated NO_x (15% of the points Figure 2b). As seen in Figure 2, model calculations of P(O₃) for median background SONEX conditions with varying NO_x and P(HO_x) (contour lines) generally reproduce the observations (square symbols).

The primary HO_x source in the SONEX data displays large variations, with 24-hour average values ranging from 10 pptv/day to 700 pptv/day (Figure 2). Elevated P(HO_x) values (200-700 pptv/day) are sometimes associated with low NO_x concentrations (<30-40 pptv). These air masses were influenced by recent marine convection and high concentrations of water vapor; despite the enhanced source of HO_x, the low levels of NO_x result in relatively low P(O₃) (0.1-0.5 ppbv/day). Elevated P(HO_x) is also found in association with high NO_x concentrations (>300 pptv). As noted above, the elevated NO_x was the result of recent lightning and convection, and concurrent enhancement of P(HO_x) would be expected from the convective injection of HO_x precursors. The positive correlation between high NO_x and P(HO_x) in the observations is particularly apparent in Figure 2b, but can also be seen in Figure 2a. This correlation results in the highest levels of P(O₃) observed. By supplying HO_x precursors together with NO_x, deep convection extends the NO_x-limited regime to higher concentrations of NO_x.

Chemical regime for ozone production

To diagnose the actual dependence of P(O₃) on NO_x in the SONEX observations, the additional sensitivity to P(HO_x) must be resolved. We therefore examined the dependence of P(O₃) on NO_x for similar primary HO_x production rates. Figure 3 illustrates this dependence for three ranges of P(HO_x). The P(HO_x) values used to segregate the observations are those required in order to match the observed HO₂ (Figure 2b). Choosing instead the P(HO_x) values computed from H₂O and acetone only (Figure 2a) results in some small differences in Figure 3c.

Figure 3 shows that P(O₃) derived from observations increases nearly linearly with NO_x for NO_x<70 pptv. In Figure 3a and Figure 3b, P(O₃) shows very little dependence on NO_x between 70 pptv and 300 pptv, approaching the NO_x-saturated regime. The bin with the highest levels of P(HO_x) (Figure 3c), shows a positive dependence of P(O₃) on NO_x extending to a higher NO_x concentration (200 pptv) but there is still clear evidence of NO_x-saturated conditions beyond this. The median NO_x mixing ratio was 93 pptv in the upper troposphere (8-12 km) during SONEX, and the median P(HO_x) was 50 pptv/day. These conditions correspond to a regime where ozone production is less sensitive to changes in NO_x.

Because of the slow photochemistry in October-November and the elevated levels of NO_x, the conditions during SONEX allowed extensive sampling of the transition region between the NO_x-limited and NO_x-saturated regimes as illustrated in Figure 3. Previous aircraft campaigns (ASHOE/MAESA, STRAT, SUCCESS), where ozone production was consistently NO_x-limited, featured lower NO_x concentrations and more active photochemistry. In the tropical upper troposphere during STRAT, NO_x mixing ratios were generally less than 100 pptv [Wennberg et al., 1998]. Over the central United States during SUCCESS, springtime conditions resulted in more rapid photochemistry compared to SONEX and thus a higher transition from NO_x-limited to NO_x-saturated regimes (NO_x~500 pptv) [Jaeglé et al., 1998].

Delineation of the regimes for ozone production is critical when assessing the effect of aircraft emissions. Aircraft, unlike deep convection and lightning, inject NO_x into the upper troposphere without injecting HO_x precursors (the aircraft sources of H₂O and HONO are negligibly small [NASA, 1997]). As summarized in Singh et al. [this issue], aircraft emissions might have contributed 20-70% of the observed NO_x in the upper troposphere during SONEX. A 40% aircraft effect corresponds to a 37 pptv contribution to the median observed NO_x concentration of 93 pptv. Based on the dependence shown in Figure 3b, and assuming a SONEX median value for P(HO_x) of 50 pptv/day, such a NO_x increase results in an increase of P(O₃) from 0.45 to 0.6 ppbv/day. For a 2-week residence time of air in the upper troposphere at midlatitudes, this increase of 0.15 ppbv/day adds about 2.1 ppbv of ozone, resulting in a 4% increase in upper tropospheric ozone in the north Atlantic flight corridor in the fall. For a larger range of P(HO_x) values (20-100 pptv/day), the increase in ozone is 2-6%.

A further doubling of NO_x concentrations due to future aircraft emissions (NO_x=93+37 pptv) would only result in an additional 0.1 ppbv/day ozone production under SONEX conditions (northern midlatitudes in the fall) because of the NO_x-saturated regime. However, if the primary source of HO_x were to rise in the future, P(O₃) would become more sensitive to increases in NO_x from aircraft emissions. As noted above, during summer the transition to NO_x-saturated regime occurs at higher levels of NO_x, and thus increases in aircraft emissions should continue to result in O₃ increases in the foreseeable future [NASA, 1997].

Conclusions

We computed ozone production rates P(O₃) in the upper troposphere at northern midlatitudes, using simultaneous observations of HO₂ and NO during SONEX (October-November 1997). High levels of NO_x due to lightning and convection were associated with high concentrations of HO_x precursors also supplied by convection. The observed correlation between elevated NO_x and HO_x sources resulted in a positive relationship between P(O₃) and NO_x extending over the full range of NO_x concentrations observed (up to 1 ppbv). By segregating the data according to the primary HO_x production rate, P(HO_x), we find that ozone production in fact approached NO_x-saturated conditions for NO_x concentrations larger than 70 pptv. This result implies little sensitivity of P(O₃) to future increases in NO_x emissions from aircraft (which unlike convective injection are not associated with a large source of HO_x) during the fall at northern midlatitudes. A greater sensitivity of P(O₃) to NO_x would be expected under summer conditions.

Acknowledgments. This work was supported by the Subsonic Assessment Program (SASS) of the National Aeronautics and Space Administration (NASA).

References