Impact of aircraft emissions on reactive nitrogen over the North Atlantic Flight Corridor region

. The impact of aircraft emissions on reactive nitrogen in the upper troposphere (UT) and lowermost stratosphere (LS) was estimated using the NOy-O3 correlation obtained during the Subsonic Assessment (SASS) Ozone and Nitrogen Oxide Experiment (SONEX) carried out over the U.S. continent and North Atlantic Flight Corridor (NAFC) region in October and November 1997. To evaluate the large-scale impact, we made a reference NOy-O 3 relationship in air masses, upon which aircraft emissions were considered to have little impact. For this purpose, the integrated input of NO x from aircraft into an air mass along a 10-day back trajectory (ANOv) was calculated based on the Abatement of Nuisance Caused by Air Traffic/European Commission (ANCAT/EC2) emission inventory. The excess NOy (dNOy) was calculated from the observed NOy and the reference NOy-O 3 relationship. As a result, a weak positive correlation was found between the dNOy and ANOy, and dNOy and NOx/NOy values, while no positive correlation between the dNOy and CO values was found, suggesting that dNOy values can be used as a measure of the NOx input from aircraft emissions. The excess NOy values calculated from another NOy-O 3 reference relationship made using in situ condensation nuclei data also agreed with these dNOy values, within the uncertainties. At the NAFC region (45øN-60øN) the median value of dNOy in the troposphere increased with altitude above 9 km and reached 70 parts per trillion by volume (pptv) (20% of NOy) at 11 km. The excess NOx was estimated to be about half of the dNOy values, corresponding to 30% of the observed NOx level. Higher dNOy values were generally found in air masses where O 3 = 75-125 ppbv, suggesting a more pronounced effect around the tropopause. The median value of dNO v in the stratosphere at the NAFC region at 8.5-11.5 km was about 120 pptv. The higher dNOy values in the LS were probably due to the accumulated effect of aircraft emissions, given the long residence time of affected air in the LS. Similar dNOy values were also obtained in air masses sampled over the U.S. continent.

Details of the experiment are described by Singh et al. [1999]. In this paper, the impact of aircraft emissions on the amount of reactive nitrogen was estimated using the NOy-O3 correlation. This is because a positive correlation between NOy and 03 was generally observed in the UT and LS [Murphy et al., 1993;Ridley et al., 1994;Kondo et al., 1996] and aircraft emissions can alter the correlation . In Plate 1 a correlation plot between NOy and 03 using all of the 10-s SONEX data obtained above 8.5 km are shown. In this plate we excluded the data obtained within aircraft plumes, influenced by lightning, or affected by recent convective transport, as described in detail in the following section. As seen in this plate, the NOy values at 45øN-60øN and those over the U.S. continent are generally higher than those obtained over the west coast of the United States and at 30øN-45øN and 60øN -70øN, where air traffic density is lower. As described in the following sections, a reference NOy-O3 relationship will be derived using data from air masses, in which aircraft emissions are considered to have little impact, that is, "background air." The excess NOy (dNOy) will be estimated using this reference relationship. The statistical analyses of these dNOy values will be made for the data obtained at 45øN-60øN over the Atlantic (corresponding to the NAFC) and the U.S. continent. The excess NOx (dNOx) will also be estimated using the reference NOx/NOy ratio in the background air.  [Rinsland et al., 1982;Mahieu et al., 1995]. The conversion efficiencies of NH 3 and CH3CN were also measured in the laboratory at 1-3 and 0.5-2%, respectively. Assuming mixing ratios of NH 3 and CH3CN in the UT of 1 pptv [Ziereis and Arnold, 1986] and 200 pptv [Hamm et al., 1989], the interference from these species is considered to be negligible. During the SONEX experiment the 15 s averaged actinic flux was measured using grating monochromators to calculate photolysis rates of NO 2 (J(NO2)) [Shetter and Mfiller, 1999]. The photostationary state NO2fNO ratios were calculated using the measured values of 03, temperature, pressure, and J(NO2) for periods when the solar zenith angles were lower than 87 ø. The NOx mixing ratios were calculated from these NO2/NO ratios and the observed NO mixing ratios. The calculated NO2/NO ratios and NOx values were compared with those calculated using a box model in which a more complete photochemistry was included [Jaegle et al., 1998]. At altitudes above 7 km they agreed within 30 and 10%, respectively. Simplified chemistry was used in this study to maximize the amount of NOx data.
In this study, 10-s averaged data of 03, N20, and CO were also used. The 03 measurements were made using a chemiluminescence technique, and N20 and CO were measured using a tunable diode laser system. The concentration of condensation nuclei (CN) with a diameter larger than 15 nm (CN fine particle; TSI model 3760) [/lnderson et al., 1999]

Data Selection
During the SONEX experiment, most of the data were obtained at altitudes between 8 and 11.5 km. Air masses with 03 values lower and higher than 100 ppbv were defined as tropospheric and stratospheric, respectively. The stratospheric air masses were sampled at potential temperatures between 310 and 360 K, within the lowermost stratosphere (LS), the region above the midlatitude tropopause and below the isentrope touching the tropical tropopause, roughly 385 K, defined by Holton et al. [1995]. On the basis of the Microwave Temperature Profiler (MTP) tropopause height data the LS data were obtained within 2 km from the tropopause. The MTP data also showed that air masses with 03 values higher than 100 ppbv were sometimes obtained below the tropopause even though a clear signature of stratospheric air intrusion was not seen in the airborne differential absorption lidar (DIAL) 03 data. However, in general, air masses with 03 values higher than 100 ppbv were considered to have the same chemical characteristics as those from the LS. Similarly, air masses in which 03 values lower than 100 ppbv observed above the MTP-derived tropopause height were regarded as the UT ones.
In the present study, an increase in NOx and NOy from aircraft emissions was estimated for two study areas. The first area chosen was at latitudes between 45øN and 60øN and at longitudes between 0øW and 80øW. This data set was obtained mostly during the intensive flights from Bangor and Shannon (Figure 1). Air traffic density is quite high in this area and will be referred to as the "NAFC region" in the following. The other area was located over the U.S. continent at latitudes between 35øN and 45øN and at longitudes between 70øW and 120øW. The data set in this area was obtained during the two transit flights made on October 13, 1997, and November 12, 1997, between NASA Ames on the U.S. west coast and Bangor on the east coast. This area will be referred to as the "U.S.  , 1999]. In this study, the data obtained within the plumes were excluded in the evaluation of the large-scale impact from the aircraft emissions. This was because the sampling frequency of the plumes depends on the location and time of the measurement (many plumes were observed when the measurements were made in air masses just after heavy air traffic) and an inclusion of these data could be a potential cause of a bias in overestimating the aircraft impact. The excluded data corresponded to 6% of the entire data set at the 8.5 to 11.5 km altitude range. Because they were relatively infrequent, the results of this study would not change significantly even if the plumes were included in the analyses. A clear signature of NO production by lightning was seen during four flights on October 13 and 29, 1997, and November 3 and 9, 1997 [Allen et al., this issue; A.M. Thompson et al., unpublished manuscript, 1999]. In these cases the NOx and NOy mixing ratios increased more than 1 ppbv, and NOx/NOy ratios ranged between 0.4 and 1.0 suggesting that NO production by lightning had occurred within a few days prior to the measurements. These results were generally consistent with air mass trajectory analysis, convective activities using cloud images, and lightning activities detected by the U.S. ground-based lightning network [Allen et al., this issue; A.M. Thompson et al., unpublished manuscript, 1999]. Possible influences from lightning were also observed on October 20 and 28, 1997. In this study all of the data likely to have been influenced by NO production by lightning were excluded. It should be noted, however, that the influence of lightning cannot be identified in an air mass once the NOx in the air mass is diluted down near the background level. Consequently, we could not completely remove from our analysis the data which might have been affected by NO production from lightning. The contribution from lighting will be revisited in section 3.2.3.
The vertical profile of the CO in the troposphere is shown in Figure 2 using all of the data except for those obtained from aircraft plumes and clearly influenced by lightning. The median values and 67% ranges in the NAFC region are also shown. As seen in this figure, the CO mixing ratio generally decreased with altitude, and the median values were between 76 and 82 ppbv at 9 to 11 km. These values were systematically lower than the median CO values of 100 to 105 ppbv obtained in the continental or maritime air masses in the UT over the middle to high latitude western Pacific in September and October [Kondo et al., 1996]. In Figure 3

a correlation plot between
NOy and CO is shown using the data obtained at altitudes above 8.5 km. Different symbols are used for the UT and LS data. The NOy mixing ratio did not generally have a positive correlation with the CO mixing ratio in the UT. These results suggest that NO•, values in most of air masses in the UT obtained during SONEX had not been influenced by recent convective transport of polluted continental surface air. To eliminate the possible influence from recent convection, the data with CO values higher than 100 ppbv were excluded in this study. In addition, data showing a clear positive correlation between CO and NO v were also excluded. The excluded data by these criteria corresponded to 19% of the entire UT data set obtained above 8.5 km.
The correlations of NO v with nonmethane hydrocarbons such as ethyne (C2H2), ethane (C2H•,), and propane (C3Hs), and halocarbons such as tetrachloroethylene (C2C14) , methyl iodide (CH3I), and CHBr 3 were also examined. No clear positive correlation was seen in the UT. In fact, NOy decreased with increasing mixing ratios of these hydrocarbons and halocarbons for NO v mixing ratios higher than 300 pptv (not shown). C2H2, C2H6, C3Hs, and C2Cl 4 are emitted in urban areas. CH3I and CHBr3 are considered to be good indicators of maritime air mass. The photochemical lifetimes of these species are comparable or shorter than that of CO. Consequently, the lack of a clear correlation of NOv with these species further confirms that the data selected in this study were generally free from the recent convective transport of air influenced by urban and other surface sources.

Reference NOy-O3 Relationship
To estimate the increase in the NO,, mixing ratio due to aircraft emissions, we utilized a method that uses the NOv-O 3 relationship. In this method a reference NOy-O3 relationship was estimated using data in air masses upon which aircraft emissions were believed to have had little impact, that is, "background air masses." To select the background air masses at 8.5-11.5 kin, two independent approaches were taken. As described below, the agreement in the estimates on the excess NOy derived from these two approaches was evaluated, and the first approach described below was used for the further analyses.
For the first approach, 10-day back trajectories were calculated using a kinematic method for air masses sampled every 1 min on board the NASA DC-8. The European Centre for Medium-Range Weather Forecasts (ECMWF) data were used for this calculation. Then an integrated value of expected NO x input from aircraft emissions along each trajectory was calculated using the monthly mean values of the three-dimensional NOx emission distribution for October 1992, compiled in the Abatement of Nuisance Caused by Air Traffic/European Commission (ANCAT/EC2) emissions inventory [Gardner, 1998]. No chemical loss or diffusion process and no diurnal variation in the NOr emission rate were taken into account for this calculation, although some initial dilution effect was included because the emission rate was provided for each 1 ø x 1 ø in latitude and longitude and every 1 km in altitude. The typical emission rate in the corridor region was 2-5 pptv h -•. The calculated value is denoted as ANOy. For the present analysis, and 47% of the UT and LS data. As for the "low ANOy reference," a median NOy value was calculated for each 10 ppbv 03 range, and a linear relationship was calculated for the UT and LS data separately (Figure 4, denoted as the "low CN reference" hereafter). The dNOy values were calculated for every 10-s data using this relationship. Note that data obtained from background air masses were not excluded and were used in conjunction with other data for the analyses described below.
It should be noted that products from three-dimensional    Table 1  of the dNOy/NOy ratio at these two altitudes were 16 and 21%; that is, about 20% of the NOy in the UT likely originated from aircraft emissions. These results are also summarized in Table   1.

Validity and Uncertainty in the dNO), Estimate 3.2.1. Reference values. The median values of the dNO• calculated using the two references are shown in
As described above, the model input NOx value along the air mass trajectory, zXNOy, was calculated for each data point. Consequently, a statistical analysis can be made on the zINO•, value for the same data set from which the dNO• values were calculated. In Figure 7 the median values of ANO v for the 5and 10-day trajectories are shown. The median zXNOy (10 days) was 25 pptv at 9 km and increased with altitude reaching 160 pptv at 11 km. The shape of the profile is consistent with that of the dNOy value, although the dNOy values were systematically lower than the zXNOy values, a likely result of the dilution effect.
The median value of dNOy in the UT over the U.S. continent region was calculated using all of the data obtained at 8.5-11.5 km range (Figure 11 and Table 1). This was because the amount of data was not sufficient to derive a vertical profile of the dNOy values. The median value was 32 pptv, comparable to the 28 pptv in the NAFC region. Although the data over the U.S. continent were obtained from only two flights and the amount of data was smaller (4790 versus 1539 10-s data), the results here suggest a similar effect from aircraft emissions in the two regions.
The median value of dNOy in the LS was calculated for the 8.5-11.5 km range ( Figure 11 and Table 1

Relationship of dNOy With NOx/NOy, CN, and O3 In Figures 12a-12d the correlation plots between the dNOy and NOx/NOy values and between the dNOy and CN values
are shown for the UT and LS using the data from both the NAFC and the U.S. continent regions. For these figures we used only the data in which the model input NO• values during the last 5 days were greater than 80% of the 10-day values (i.e., zXNOy (5 days)/ANOy (10 days) > 0.8). This is because NO• molecules emitted from aircraft more than 5 days prior to the measurements could have already been converted to higher oxidized species, such as HNO 3. In fact, this data selection

Summary
The large-scale impact of aircraft emissions on reactive nitrogen in the UT and LS was estimated using the NOy-O 3 correlation observed during SONEX, which was carried out over the U.S. continent and NAFC region in October and November 1997. For the present analysis we excluded data obtained within aircraft plumes, data clearly influenced by NO production from lightning, and data influenced by recent convective transport of air affected by surface sources. Two reference NOy-O 3 relationships in the air masses, which were considered to be affected little by aircraft emissions, were made using the data obtained above 8.5 km. First, the integrated input of NOx into an air mass along a 10-day back trajectory The ratio between excess NOx (dNOx) and dNO v was cal-culated using the reference NOx/NOy ratio. In the UT (NAFC region) it was estimated that about half of dNOy was in the form of NOx (20 and 33 pptv at 10 and 11 km) and about 30% of the NOx had originated from aircraft emissions. These estimates were generally consistent with the results from the model calculations.
In this study, the degree of the large-scale impact of aircraft emissions on reactive nitrogen was estimated from the observed data. A more quantitative estimate will be made using sophisticated three-dimensional models when they can consistently reproduce most of the observed features of various species.