Total reactive nitrogen (NOy) in the Arctic lower stratosphere was measured from the NASA DC-8 aircraft during the SAGE III Ozone Loss and Validation Experiment (SOLVE) in the winter of 1999/2000. NOy-N2O correlations obtained at altitudes of 10–12.5 km in December 1999 and January 2000 are comparable to the reported reference correlation established using the MkIV balloon measurements made during SOLVE prior to the onset of denitrification. Between late February and mid-March, NOy values obtained from the DC-8 were systematically higher than those observed in December and January by up to 1 part per billion by volume, although a compact correlation between NOy and N2O was maintained. Greater increases in NOy were generally observed in air masses with lower N2O values. The daily minimum temperatures at 450–500 K potential temperature (∼20–22 km) in the Arctic fell below the ice saturation temperature between late December and mid-January. Correspondingly, intense denitrification and nitrified air masses were observed from the ER-2 at 17–21 km and below 18 km, respectively, in January and March. The increases in NOy observed from the DC-8 in late February/March indicate that influence from nitrification extended as low as 10–12.5 km over a wide area by that time. We show in this paper that the vertical structure of the temperature field during the winter was a critical factor in determining the vertical extent of the NOy redistribution. Results from the Reactive Processes Ruling the Ozone Budget in the Stratosphere (REPROBUS) three-dimensional chemistry transport model, which reproduced the observed general features only when the NOy redistribution process is included, are also presented.
 During the recent cold winters of the 1990s, such as 1994/1995, 1995/1996, and 1996/1997, significant ozone (O3) loss was observed in the lower stratosphere in the Arctic. At temperatures below the nitric acid trihydrate (NAT) particle saturation point (TNAT), liquid or solid particles (polar stratospheric clouds, PSCs) form and chlorine activation occurring on them results in subsequent catalytic photochemical O3 destruction [World Meteorological Organization (WMO), 1999, and references therein]. Formation of PSC particles also causes nitric acid (HNO3) uptake, leading to significant decreases in gas phase total reactive nitrogen (NOy) concentrations [Carslaw et al., 1994, 1997; Drdla et al., 1994; Tabazadeh et al., 1994; Del Negro et al., 1997]. Denitrification, defined as the permanent removal of NOy from an air mass by the sedimentation of PSCs, maintains high chlorine levels after temperatures warm to above the threshold temperature of PSC formation. Consequently, it delays deactivation of chlorine, extending ozone depletion through the late winter and into early spring [Salawitch et al., 1993; Rex et al., 1997; Tabazadeh et al., 2000].
 Denitrification was observed in the Arctic stratosphere during the cold winters of 1989/1990 [Fahey et al., 1990], 1994/1995 [Sugita et al., 1998; Waibel et al., 1999], 1995/1996 [Hintsa et al., 1998], and 1996/1997 [Kondo et al., 2000; Irie et al., 2001]. The degree of denitrification in the Arctic vortex is generally less intense and has greater interannual variability than that in the Antarctic vortex. When large PSC particles fall to lower altitudes and encounter higher temperatures, they evaporate or sublime, producing local increases in HNO3 (nitrification). Nitrification has been observed in the Arctic lower stratosphere from aircraft [Hübler et al., 1990; Kawa et al., 1992; Hintsa et al., 1998] and satellite [Kondo et al., 2000; Irie et al., 2001] instruments. Nitrification can result in an enhancement of NO2 through the photolysis of the enhanced HNO3 and therefore reduce the ozone loss rate at these altitudes. Consequently, measurements of nitrification provide not only additional evidence of redistribution of reactive nitrogen but also insight into the photochemistry at altitudes below the denitrified layer. Limited measurements of nitrification, however, have left significant uncertainties in the degree of nitrification, its horizontal extent, and its seasonal evolution.
 The Arctic stratospheric vortex during the 1999/2000 winter, studied by the SAGE III Ozone Loss and Validation Experiment (SOLVE)/Third European Stratospheric Experiment on Ozone (THESEO) 2000 campaign, was characterized by unusually low temperatures, especially in December and January [Manney and Sabutis, 2000]. Several measurements conducted during this campaign show that significant ozone loss occurred in the Arctic stratosphere [Richard et al., 2001; Salawitch et al., 2002; Newman et al., 2002, and references therein]. By the end of March 2000, O3 decreased by up to 2.5 parts per million by volume (ppmv), corresponding to 70% O3 loss in the lower stratosphere [Sinnhuber et al., 2000; Santee et al., 2000]. The larger than normal loss of O3 in March 2000 was likely due to widespread denitrification [Gao et al., 2001; Sinnhuber et al., 2000; Santee et al., 2000]. In fact, severe denitrification was observed from the ER-2 at 16–21 km between late January and mid-March [Popp et al., 2001]. An average removal of more than 60% was observed in air masses throughout the core of the Arctic vortex. The severity of denitrification was comparable to that inferred from previous measurements using the ER-2 in the Antarctic polar vortex in 1987. Fahey et al.  observed large PSC particles (10–20 μm in diameter) containing nitric acid over a large altitude range (16–21 km) and horizontal extent from the end of January to the beginning of March. These large particles could fall 1.5 km/day and they are considered to be responsible for denitrification in the 1999/2000 Arctic winter.
 During the SOLVE campaign, a series of aircraft measurements in the lowermost part of the stratosphere at altitudes between 10–12.5 km (potential temperature of 300–360 K) were made using the NASA DC-8 aircraft from Kiruna, Sweden. Measurements were made during three phases from December 1999 to March 2000 (Table 1). Summaries of individual flights are given in the overview paper by Newman et al. . In situ measurement of NOy was made along with various other species and physical parameters. In this paper, NOy-N2O correlations obtained in these three deployments are presented, and the causes of changes in the correlations are discussed.
Table 1. Summary of SOLVE DC-8/ER-2 Deployments
Number of Flights
30 Nov. 1999
16 Dec. 1999
14 Jan. 2000
29 Jan. 2000
27 Feb. 2000
15 March 2000
14 Jan. 2000
3 Feb. 2000
26 Feb. 2000
16 March 2000
2. Measurements and Chemical Transport Model
 A schematic diagram of the DC-8 NOy instrument used during SOLVE is shown in Figure 1. NOy concentration was measured using an NO-O3 chemiluminescence technique after NOy compounds were catalytically converted into NO on the surface of a heated gold tube with the addition of CO. The NOy instrument used for SOLVE was very similar to that used during NASA's Subsonic Assessment (SASS) Ozone and NOx Experiment (SONEX) conducted in 1997 [Koike et al., 2000; Kondo et al., 1997]. The main difference was the mounting of the NOy catalytic converter. During SOLVE, the NOy converter unit was placed outside the aircraft cabin so that the length of the Teflon tube between the air sample inlet and the converter was minimized. The Teflon tube upstream of the converter was heated to 40°C. The inlet tube for air sampling faced rearward, discriminating against particles of diameters larger than about 1 μm. The conversion efficiency of NO2 in ambient air was checked approximately every hour during each flight to achieve an efficiency of 99 ± 1%. The conversion efficiency of HNO3 was checked in the laboratory before and after each deployment and values higher than 95% were found. The data were recorded every 1 s and 10-s averages were used in this study. The precision of the 10-s NOy measurements at 10 km estimated from photon count fluctuations (1-σ) was 4 parts per trillion by volume (pptv) for an NOy value of 1000 pptv. The absolute accuracy was estimated to be 10% for the given NOy values.
 In addition to the gas phase NOy measurement, measurement of NOy with the forward facing inlet was simultaneously made using an independent gold tube converter (Figure 1). With this configuration, HNO3-containing particles were detected in addition to gas phase NOy and this measurement is denoted as the total NOy measurement. Except for the very limited cases in which PSC particles existed, NOy and total NOy measurements agreed to within the combined uncertainties throughout the three deployments. This good agreement between the two NOy measurements ensure the overall performance of the NOy measurements during the SOLVE experiment.
 Intercomparisons between the DC-8 and ER-2 aircraft measurements were made on 23 January 2000 during the SOLVE campaign. The two aircraft flew very close-track at an altitude of 11.3 km in the upper troposphere for about 25 min. The results of the intercomparison for 10-s NOy data are shown in Figure 2. The time axis of the ER-2 data was adjusted so that data obtained at the same locations are overlaid. The average horizontal distance between the two measurements was 4.7 ± 2.1 km after this adjustment. As seen in this figure, the agreement between the two aircraft measurements was very good. On average, they agreed to within 35 ± 42 pptv or 8.7 ± 9.8% (the DC-8 measurements were higher), which was within the combined uncertainties of the two measurements.
 In this paper, N2O data simultaneously obtained on board the DC-8 aircraft using a tunable diode laser absorption technique [Vay et al., 1998] were used. Data were recorded at 1 Hz with a precision (1-σ) of 0.1% and an absolute accuracy better than 1% (2-σ). The O3 measurements were made using a chemiluminescence technique with a precision (1-Hz measurement) of 2% or 1 parts per billion by volume (ppbv), whichever is larger, and an absolute accuracy of 2 ppbv. Data from the ER-2 aircraft measurements of NOy [Fahey et al., 1989] and N2O are also used. Because three independent N2O measurements were made on board the ER-2 aircraft, we used a combined N2O data set (unified N2O data) [Hurst et al., 2002]. Typical agreement ranged from 3.6 to 7.3 pptv (1.8 to 3.7%) for the different instrument pairs, while the unified N2O data agreed with whole air sample measurements within 2.9 ppbv (1.5%).
2.2. Chemistry Transport Model
 We used in our analysis the REPROBUS three-dimensional (3-D) Chemical Transport Model (CTM), which computes the time evolution of 55 stratospheric species [Lefevre et al., 1994, 1998]. The chemistry package includes a comprehensive treatment of gas-phase chemistry (147 reactions and photolysis rates from JPL [DeMore et al., 1997; Sander et al., 2000]), as well as heterogeneous reactions taking place at the surface of liquid [Carslaw et al., 1995] and solid PSCs. The model was initialized on October 15, 1999. The wind and temperature fields used to drive the transport of chemical species and to compute the chemical reaction rates were prescribed every six hours by the European Centre for Medium-Range Weather Forecasts (ECMWF) analysis. In the configuration used in our study, REPROBUS extends over 42 vertical levels from the ground up to 0.1 hPa (about 65 km altitude), with a horizontal resolution of 2° × 2° and a chemical time step of 15 min.
 The process leading to denitrification is also included in the REPROBUS model. When the temperature drops below TNAT and a supersaturation ratio of 10 is achieved, NAT particles are allowed to form on a limited number of condensation nuclei (5 × 10−3 particles cm−3), and they coexist with supercooled ternary solution (STS) droplets in much larger concentrations. This highly selective nucleation process leads to large NAT particles with a maximum diameter of about 10 μm, resulting in denitrification. Various sensitivity tests were performed to evaluate the impact of the assumed NAT number density, and the results will be presented in section 4.3. Note that because of the assumption of a supersaturation ratio of 10, NAT particle are allowed to form only when temperatures drop below TNAT by 3 K at 50 hPa, which is about 192 K. This temperature is the edge of the “nucleation window” of 191 ± 1 K at 50 hPa, where production rates of nitric acid dihydrate (NAD) and NAT particles through the homogeneous freezing process are largest [Tabazadeh et al., 2001]. When these large NAT particles fall to lower altitudes and encounter temperatures higher than TNAT, they evaporate or sublime, causing nitrification.
3. Meteorological Conditions
 The 1999/2000 Arctic stratospheric vortex was characterized by unusually low temperatures, especially in December and January [Manney and Sabutis, 2000]. Figure 3 shows the minimum temperatures north of 45°N at various potential temperature levels from the ECMWF analyses. The threshold temperatures for the formation of NAT particles, TNAT, and ice particles, TICE, are also shown in this figure. These values were calculated using the formula derived by Hanson and Mauersberger  and Marti and Mauersberger  and assuming the HNO3 and H2O mixing ratios used by Irie et al. . Figures 4a and 4b show the development of areas with temperatures below TICE and TNAT, respectively. At 475 K (∼21 km), where the ER-2 measurements were intensively made, the minimum temperature fell below TNAT from the end of November to the beginning of March and below TICE at the end of December and the beginning of January. Between 450 and 500 K, the area where temperatures were below TNAT covered close to half of the total Arctic vortex area of about 19 × 106 km2 through January. At 350 K (∼13.5 km), which is slightly higher than the altitude where the DC-8 measurements were intensively made, the minimum temperature fell well below TNAT from the beginning of February to the beginning of March. At this altitude, areas in which temperatures occasionally fell below TNAT in February correspond to 5% of the total vortex area. The time when the lowest temperatures appeared was generally observed later at lower altitudes.
 Corresponding to these low temperatures in entire Arctic stratosphere, extensive PSCs were observed over the course of the SOLVE campaign [Newman et al., 2002, and reference therein]. These results suggest that large-scale low-temperature features were generally responsible for PSC formation during the 1999/2000 Arctic winter.
4. Results and Discussion
4.1. N2O-NOy Correlation Observed from DC-8
 Scatterplots between NOy and N2O obtained from DC-8 measurements during the December, January, and March deployments are shown in Figure 5. Median values of NOy in each N2O range are shown in Figure 6, and a linear regression line calculated for these median values is given in Table 2. In these figures, data obtained in stratospheric air masses (O3 > 100 ppbv) sampled north of 50°N and above 5 km were used. Popp et al.  established a reference N2O-NOy relationship for the 1999/2000 Arctic vortex using remote measurements obtained by the MkIV balloon-borne interferometer on December 3, 1999, prior to the onset of denitrification:
where both [NOy*] and [N2O] are expressed in ppbv. The reference relationship was derived using data with N2O values between 33 and 321 ppbv. The NOy values measured from the ER-2 aircraft agreed well with the NOy* values in air masses that had not been perturbed by denitrification or nitrification processes, and we will therefore refer to the above equation as the MkIV/ER-2 reference relationship. This reference relationship is shown in Figures 5 and 6. M. Loewenstein et al. (An NOy* algorithm for Arctic winter 2000, submitted to Journal of Geophysical Research, 2001) also established a reference N2O-NOy relationship for the 1999/2000 Arctic winter for the entire N2O range. Their NOy* values for the N2O = 60–320 ppbv range were quite similar to those derived using the equation (1).
These linear regression lines were calculated for median values, shown in Figures 6 and 8 for the case of the NOy-N2O and O3-N2O relationships. In the same way, the relationship between NOy and O3 was calculated for median values of NOy for various O3 ranges (50 ppbv intervals). When the regression lines were calculated, the value of 100 or 316.6 in the expression was kept fixed. The O3 mixing ratio of 100 ppbv was chosen as a typical value at the tropopause. The N2O mixing ratio of 316.6 ppbv is an average value in the troposphere during SOLVE.
[NOy(pptv)] = −70.29 ([N2O(ppbv)] −316.6) + 609
[NOy(pptv)] = −76.35 ([N2O(ppbv)] −316.6) + 653
[NOy(pptv)] = −122.47 ([N2O(ppbv)] −316.6) + 249
[O3(ppbv)] = −24.42 ([N2O(ppbv)] −316.6) + 120
[O3(ppbv)] = −23.13 ([N2O(ppbv)] −316.6) + 99
[O3(ppbv)] = −20.63 ([N2O(ppbv)] −316.6) + 126
[NOy(pptv)] = −2.603 ([O3(ppbv)] −100) + 655
[NOy(pptv)] = −3.558 ([O3(ppbv)] −100) + 536
[NOy(pptv)] = −5.534 ([O3(ppbv)] −100) + 288
 During the SOLVE campaign, the N2O mixing ratios measured from the DC-8 ranged between 280 and 320 ppbv. The N2O–NOy relationship in December was in good agreement with the MkIV/ER-2 reference relationship. In January, although NOy values increased slightly in air masses with N2O mixing ratios of 290–300 ppbv, they were still in good agreement with the MkIV/ER-2 reference values. In contrast, NOy values obtained in late February/March (March deployment) were systematically higher than those in December and January. The increase in NOy values (ΔNOy = NOy − NOy*) was especially evident in air masses with N2O values of 280–290 ppbv (altitudes of 11.0–12.5 km and potential temperatures of 340–350 K), where ΔNOy values were 0.5 to 1 ppbv. These air masses were sampled during the five flights made from 27 February to 15 March. It should be noted that the linear correlation between NOy and N2O is quite compact in late February/March, as confirmed by small standard deviations in NOy values (Figures 5 and 6). This indicates that air masses sampled in late February/March were generally well mixed with surrounding air masses.
 The highest values of NOy and ΔNOy, up to 5 and 1 ppbv, respectively, were observed on 27 February during the transit flight from the west coast of United States to Kiruna, Sweden. The locations in which these high NOy values were observed are shown in Figure 7 with the Ertel's potential vorticity (EPV) field at the 217-hPa surface. As seen in this figure, these high-NOy air masses were sampled in the regions where EPV values were relatively high on that pressure altitude.
 Possible causes that could alter the N2O-NOy correlation are now examined. In Figure 6, a scatterplot between the REPROBUS model-calculated NOy and N2O is shown for the 16 January and 27 February cases. In these calculations, the process leading to denitrification was disabled (model denitrification OFF), and model values along the DC-8 flight track were extracted. These two days were selected because 16 January (January deployment) was the day on which air masses with N2O = 280–290 ppbv were observed with no increase in NOy from the DC-8, while 27 February (March deployment) was the day on which air masses in the same N2O range were observed with a clear increase in NOy (∼1 ppbv). The model-calculated NOy-N2O relationships are quite similar between the January and March cases. This result indicates that a seasonal change in the transport processes is not responsible for the observed change in the NOy-N2O relationship. The earlier ER-2 measurements also show that the NOy-N2O correlation at northern midlatitudes over the wider N2O range (N2O = 170–300 ppbv) changed little with season when intensive denitrification did not take place [Keim et al., 1997].
 In Figure 8, median values of O3 are shown versus N2O using DC-8 data. A linear regression line calculated for these median values is given in Table 2. As seen in this figure and table, the relationship changed little between December and March and did not show any corresponding increase in O3 in March as seen in the NOy-N2O relationship. This result further confirms that there was no change in the meridional transport, which can alter the NOy-N2O or O3-N2O relationship, in the DC-8 sampling area during the SOLVE campaign. Slight decreases in O3 amount from December to March could be due to photochemical O3 loss. In fact, the decrease of 63 ppbv (7.1%) in air masses with N2O = 280 ppbv between the January and March deployments (44 days apart) is generally consistent with the photochemical O3 loss of about 80 ppbv (9%) within the same period calculated using the REPROBUS model for N2O = 280 ppbv air masses. An O3 reduction of 60–130 ppbv (9–16%) was also observed during the AASE 2 experiment from the DC-8 between January and March 1992 in air masses with N2O = 270–290 ppbv (at altitudes around 11.5 km) when the N2O-O3 correlation was used in the estimation [Collins et al., 1993]. In addition to the N2O-O3 and NOy-N2O relationships, a relationship between NOy and O3 observed during SOLVE is given in Table 2 for reference.
 There are other possible causes that could alter the NOy-N2O correlation. First, it is very unlikely that the observed increase in NOy in March was due to errors in the observations. The absolute values of the NOy measurements were well calibrated with a NO standard gas, and no systematic change in the sensitivity was found throughout the three deployments. The conversion efficiency of HNO3 was higher than 95% and, in fact, NOy measurements agreed with independent measurements of total NOy within the combined uncertainties, as described in section 2.1. Furthermore, the changes in sensitivity and HNO3 conversion efficiencies cannot explain the observed N2O dependence in the ΔNOy values. Second, recent mixing of tropospheric air into the lower stratosphere cannot explain the observed increase in NOy, because the NOy values in the troposphere were mostly below 1 ppbv. Third, aircraft emissions can increase the NOy level in the lower stratosphere; however, its contribution was estimated to be only 0.1–0.2 ppbv, even near the North Atlantic flight corridor [Koike et al., 2000]. It was also suggested that their impact was largest near the tropopause region (O3 = 75–125 ppbv), and it was generally smaller deeper inside the stratosphere. Therefore the increase in NOy observed in late February/March was quite likely due to the vertical redistribution of NOy in the Arctic vortex caused by the gravitational sedimentation of PSC particles. As seen in Figures 5 and 6, the NOy value increased by changing the dNOy/dN2O gradient rather than by a constant offset. The N2O dependence of the NOy increase was reasonable, because air masses with lower N2O values were expected to be affected more strongly by higher altitude air masses where nitrification was more intense. In the following sections (sections 4.2 and 4.3), redistribution of NOy observed from the ER-2 and results from the 3-D REPROBUS model calculations are presented to confirm our interpretation.
4.2. N2O-NOy Correlation Observed from ER-2
 In Figure 9a, a scatterplot between N2O and NOy measured from the ER-2 aircraft is shown. For this plot, we used only data obtained at temperatures higher than TNAT, to exclude measurements made when NOy-containing PSC particles were present in the sampled air mass. For the TNAT calculation, observed values of NOy and H2O were used, and all of NOy was assumed to be HNO3. In this figure, data obtained in the January and March deployments are shown using different colors (blue and red). For measurements made at altitudes higher than 15 km, data obtained outside the outer edge of the Arctic vortex, as determined using the definition of Nash et al. , are shown using another color (green). These extra-vortex data agreed well with the MkIV/ER-2 reference relationship. Severe denitrification was observed between late January and mid-March 2000 at altitudes of 17–21 km and for N2O mixing ratios between 40 and 170 ppbv, as reported by Popp et al. . They also reported that an average removal of NOy of more than 60% was observed in air masses throughout the core of the Arctic vortex. Corresponding to the denitrification, nitrification was observed from the ER-2 both in January and March (Figure 9a). Increases in NOy were as high as 10 ppbv in January.
 In Figure 9b, the vertical profile of redistributed NOy (ΔNOy = NOy–NOy*) derived from ER-2 measurements using the MkIV/ER-2 reference relationship (equation (1) given above) is shown using the same data set used for Figure 9a. As seen in this figure, denitrification was observed between 16 and 21 km, and maximum value of the NOy removal in each altitude increased with altitude partly because the amount of available NOy increased with altitude [Popp et al., 2001]. The colder temperatures at higher altitudes (Figures 3 and 4) were also favorable for the efficient removal of NOy. The vertical profile of maximum removal value shifted downward by 0.5–1 km from January to March between 18 and 21 km, although the number of measurements in January was small. This vertical shift was partly due to diabatic descent of air inside the Arctic vortex. In fact, vertical profiles of N2O obtained from the ER-2 measurements shifted downward by 0.5–1 km (10–20 K potential temperature) from January to March at altitudes above 14 km, with higher descent rates at higher altitudes (not shown). This descent rate was generally consistent with the estimates from a more comprehensive study for the 1999/2000 Arctic winter [Greenblatt et al., 2002], earlier measurements [Schoeberl et al., 1992; Abrams et al., 1996], and model calculations [Rosenfield et al., 1994]. In addition to the descent of air masses, further denitrification, which occurred at lower altitudes in February, could also cause negative ΔNOy values at 380–400 K. As seen in Figures 3 and 4, the minimum temperatures at 380–400 K (∼15.5–17 km) were lower in February than in January and were close to TICE until the beginning of March. In fact, large HNO3-containing PSC particles, which contributed significantly to denitrification, were detected until 7 March [Fahey et al., 2001].
 Air masses with positive ΔNOy values due to nitrification were occasionally observed from the ER-2. The number of measurements below 16 km was quite small because they were made only during ascent and descent from the Kiruna airport, and therefore the location of air mass sampling was also limited to the region over northern Scandinavia. Within this limited data set, nitrification was clearly seen at 15–18 km and 14–16.5 km in January and March, respectively. Increases in NOy of 0.5–1 ppbv were also seen in March in air masses with N2O = 270–320 ppbv sampled at DC-8 altitudes. (Only a few data were obtained in January in air masses with N2O = 270–320 ppbv at DC-8 altitudes). The downward shift of nitrification altitudes is attributed to the same two factors affecting the denitrification profiles, i.e., the diabatic descent of the Arctic air mass and colder temperatures in February at lower altitudes. Minimum temperatures at 350–380 K (∼13–15.5 km) were lower in February than in January, and fell below TNAT continuously only in February and the beginning of March. As a result, PSC particles could efficiently fall to lower altitudes only in February. Considering that the vortex edge was often located near Kiruna, more intensive nitrification than that observed by the ER-2 was possible in the center of the vortex, where atmospheric temperature was lower, as in the case for the winter of 1996/1997 [Irie et al., 2001]. The observations from the DC-8 indicate that nitrification extended to altitudes as low as 10–12.5 km over a wide area in late February/March 2000. This is further confirmed by model calculations, as described below (section 4.3).
4.3. Results From Model Calculations
 In Figures 10a and 10b, vertical profiles of redistributed NOy calculated with the model are shown for 16 January and 27 February, respectively. These two days were selected because 16 January (January deployment) was the day on which air masses with N2O = 280–290 ppbv were observed with no increase in NOy from the DC-8, while 27 February (March deployment) was the day on which air masses in the same N2O range were observed with a clear increase in NOy (∼1 ppbv). NOy* was calculated using the MkIV/ER-2 reference relationship. In these figures, results from the two model cases are shown, i.e., the case in which the process leading to denitrification is included (denitrification ON) and the case in which the process is disabled (denitrification OFF). In the 16 January case, denitrification is seen at altitudes above 20 km and nitrification is seen between 14 and 20 km. The increase was limited to altitudes above 14 km because high temperatures prevent PSC particles from descending to lower altitudes. In contrast, nitrification is seen down to 10 km in the 27 February case. This was caused by two factors, as described in section 4.2. First, low temperatures at 12–15 km in February allow PSC particles to fall to lower altitudes more efficiently than they did in January. Second, diabatic descent in the Arctic vortex brought air masses influenced by nitrification to slightly lower altitudes. It is noted that the altitude range of nitrification should generally be well simulated because nitrification occurs when PSC particles encounter temperatures above TNAT, and this process is reliably reproduced by the model. Negative ΔNOy values in the denitrification-OFF case were due to the mixing of air masses in the region where NOy and N2O had a nonlinear relationship [Michelsen et al., 1998; Kondo et al., 1999; Rex et al., 1999].
 In Figures 11a and 11b, scatterplots between the model-calculated NOy and N2O are shown for the 16 January and 27 February cases. For these plots, model values along the DC-8 flight track were extracted. As already described above (section 4.1), the NOy-N2O relationships are quite similar between the January and March cases when denitrification is switched off. When the results from the denitrification-ON and -OFF cases are compared, it is found that denitrification caused only a small change in the NOy values in the 16 January case at DC-8 altitudes. In contrast, in the 27 February case, an increase in the NOy values is seen when the denitrification process is included. The ΔNOy value is 0.5–1.0 ppbv in air masses with N2O = 280–290 ppbv, which is in good agreement with the measurements (Figure 5 and Figure 6), although the scatter in the NOy values is greater than the observations (see discussion in section 4.4). The good agreement of the model-calculated NOy-N2O correlation with the observations both in the January and March deployments further confirms our interpretation that the increase in NOy observed from the DC-8 in late February/March was due to NOy redistribution.
 In our model calculations, the number density of NAT particles, which were allowed to form on condensation nuclei, was limited to 5 × 10−3 particles cm−3. This density was chosen because it gave the best agreement with observations in the similar large-scale, three-dimensional model calculations for the Arctic winter stratosphere by Waibel et al. . We performed various sensitivity tests to evaluate the impact of the assumed NAT number density. When a factor of 10 greater NAT particle density is assumed (5 × 10−2 particles cm−3), the degree of nitrification becomes too small (one third of the observed NOy increase) for the 27 February case; when a factor of 10 smaller density is assumed (5 × 10−4 particles cm−3), the degree of nitrification becomes too large (0.5–0.8-ppbv increase in air masses with N2O = 280–290 ppbv) for the 16 January case. This was because the greater (smaller) number density of NAT particles resulted in smaller (larger) particle diameters owing to the limited amount of condensable material, HNO3 and H2O. Particles with large diameters (maximum of about 10 μm in our calculation) resulted in effective denitrification and nitrification. It is noted that in both this study and study by Waibel et al. , the best agreement between the large-scale model calculations and observations was achieved when a NAT density of 5 × 10−3 particles cm−3 was assumed.
Fahey et al.  showed that the HNO3-containing particles in the 1999/2000 Arctic winter stratosphere likely had two particle size modes: the larger mode with a mean diameter of 14.5 μm and a number density of 2.3 × 10−4 particles cm−3, and the smaller mode at 3.5 μm with a number density of 2 × 10−3 particles cm−3 (uncertainty of ±30%). The best guess NAT density (5 × 10−3 particles cm−3) in the present study is therefore systematically greater than these observed particle densities. Because of the rather simple treatment of particle growth in the REPROBUS model, the NAT density used in the model may not be directly comparable to the observations. However, the difference between the observations and model calculations provides a range of uncertainties in the NAT particle density causing denitrification and nitrification.
4.4. Processes of NOy Redistribution
 In this section, we explain two observed features: first, a difference in time between the first observation of denitrified air from the ER-2 (20 January) and the first observation of nitrification influence from the DC-8 (27 February), and second, the compactness of the NOy-N2O correlation observed from the DC-8 in late February/March, in spite of the sporadic nature of the nitrification processes.
 The first ER-2 measurement within the Arctic vortex in the 1999/2000 winter was made on 20 January, in which severely denitrified air masses were observed, indicating that denitrification started prior to this measurement. The nucleation of NAT and NAD at temperatures near TICE was indicated by laboratory experiments [Worsnop et al., 1993; Salcedo et al., 2001]. Selective nucleation of these particles and subsequent particle growth can cause denitrification [Waibel et al., 1999; Tabazadeh et al., 2001]. In fact, recent satellite measurements have shown that denitrification occurred only in air masses that experienced temperatures near TICE [Kondo et al., 2000; Irie et al., 2001]. The minimum temperature in the Arctic at 475 K potential temperature was near or below TICE from the end of December (Figures 3 and 4). Consequently, denitrification must have started at that time.
 The January DC-8 measurements made between 14 and 29 January (Table 1) showed no clear increase in NOy in this period (Figures 5 and 6). The clear increase in NOy was observed for the first time during the 27 February flight (transit into Kiruna), which was the first flight during the March deployment. No DC-8 data are available between 29 January and 27 February.
 The absence of significant nitrification at DC-8 altitudes in January was primarily because temperatures between 350 and 380 K (∼13.5–15.5 km) were generally high in January (Figures 3 and 4), and almost all of the PSC particles had evaporated at altitudes above 12.5 km, as described in sections 4.2 and 4.3. In February, the Arctic minimum temperatures were continuously lower than TNAT at altitudes as low as 350 K (∼13.5 km), enabling more particles to fall efficiently to DC-8 altitudes. Consequently, the redistribution process, which occurred in February, should be partly responsible for the observed increase in the NOy level in late February/March.
 Diabatic descent of nitrified air masses can also partly account for the delay of the first observation of the NOy redistribution at DC-8 altitudes. Because temperatures at 475 K (∼21 km), where intensive denitrification was observed (Figure 9b), were close to or below TICE only until the end of January and were well above TICE in February, denitrification must have been less intense in February. In fact, although large HNO3-containing particles with diameters of 10 to 20 μm were detected from the ER-2 between 20 January and 7 March in a total of seven flights, the number and spatial extent of these particles were significantly less after 20 January [Fahey et al. , 2001]. Consequently, if nitrification occurred just above the DC-8 altitude in January, a slight diabatic descent could lead to the NOy increase in late February/March.
 The N2O mixing ratios observed from the DC-8 are shown versus potential temperature in Figure 12. In this figure, median values of potential temperature for various N2O ranges are shown for the three DC-8 deployments. The change in the potential temperature in air masses with N2O = 280–290 ppbv was about 8 K between the January and March deployments (44 days apart). Assuming this change was solely due to diabatic cooling, it corresponded to an altitude displacement of 0.5–0.8 km. A similar change in the potential temperature was also detected from the unified N2O measurements from the ER-2 in air masses with N2O = 250 ppbv for the same time period [Greenblatt et al., 2002]. A similar change was also detected between mid-January and mid-February 1992 from the DC-8 N2O measurements made during AASE 2 [Collins et al., 1993]. Figure 12 also shows that the altitude displacement was greater between the December and January deployments (1.0–1.7 km for air masses with N2O = 294–300 ppbv). As described above, denitrification quite likely started at the end of December. Air masses that were influenced by nitrification earlier in the winter could descend farther by the time of the late February/March observations. On the other hand, the altitude regions where PSC particles evaporated were generally high in early winter, as seen in Figures 3 and 4, indicating that nitrified air masses needed to descend relatively long vertical distances to be observed from the DC-8. Influence of the diabatic descent of nitrified air masses was therefore critically determined by the vertical structure of the temperature field and its temporal change through the winter.
 The compactness of the NOy-N2O relationship observed from the DC-8 in late February/March is now examined. From the discussion above, PSC particles likely fell to above DC-8 altitudes in December and January and as low as DC-8 altitudes in February. When PSC particles evaporated within a particular altitude layer, the effects were initially confined to a limited horizontal area, and the NOy-N2O correlation became inhomogeneous in that layer. Consequently, the observed compactness indicates that air masses influenced by nitrification were generally well mixed with surrounding air masses so that local enhancements in the NOy-N2O correlation diminished before the DC-8 observations in late February/March.
 In Figure 13, the ΔNOy values obtained in the March deployment are shown as a function of potential temperature and modified potential vorticity (MPV, defined by Lait  and θ0 = 420 K was assumed). As anticipated from the correlations of ΔNOy and N2O (Figure 6) and of N2O and potential temperature (Figure 12), higher ΔNOy values were generally found at higher potential temperatures. Figure 13 also shows that the ΔNOy values on each of the isentropic surfaces were relatively uniform, especially in air masses with MPV = 12–23 PV units (10−6 K m2 kg−1s−1) at potential temperatures of 320–350 K. The average ΔNOy value in these air masses was 400 ± 196 pptv. This result further confirms that air-mixing processes tended to homogenize the influence from nitrification in space by the time of March deployment.
 Model calculations provide some insights into the redistribution and mixing processes. In Figure 14, the results from the two model cases are shown for the 27 February flight track, i.e., the case in which the process leading to denitrification is activated only between December 1999 and January 2000 and the case in which the process is activated only in February 2000. As seen in this figure, the denitrification process, both in December–January and February, likely contributed to the observed increase in NOy. The process in December–January resulted in a systematic increase in NOy in air masses with N2O < 305 ppbv, and the resulting NOy-N2O relationship is compact, presumably due to mixing processes. On the other hand, the process in February resulted in a sporadic increase in NOy in air masses with N2O < 295 ppbv and the resulting NOy-N2O relationship has a relatively large standard deviation, presumably due to a mixing time insufficient to achieve a compact relationship. Considering the compactness of the N2O-NOy relationship observed in late February/March, the process in December–January could have contributed more to the observed increase in NOy.
4.5. Comparison with Earlier Measurements
 In Figures 15a and 15b, the N2O-NOy relationships obtained in earlier measurements in the Arctic or northern midlatitudes are shown with the median SOLVE/DC-8 NOy-N2O values. For these plots, N2O values in earlier measurements were scaled due to an interannually increasing trend of N2O in the troposphere by 0.5 ppbv/year [Ishijima et al., 2001].
Figure 15a shows that the high NOy values, as well as the steep dNOy/dN2O gradient, observed in March 2000 have not been observed before in the Arctic and northern midlatitudes from either the DC-8 or the ER-2. During AASE 1, conducted over the Arctic in January and early February 1989, episodic events of unusually high NOy values with mixing ratios between 3 and 12 ppbv were observed from the DC-8 at altitudes of 10–12.5 km during several flights late in that mission [Hübler et al., 1990]. In one of the significant cases, NOy levels were elevated for about 30 min (390 km along the flight track). Substantial denitrification inside the Arctic vortex was observed by the ER-2 during that winter, and the observed NOy increase was attributed to a redistribution of NOy. Although O3 and N2O data were limited, the observed local enhancements in the NOy/O3 and NOy/NOy* ratios indicate that these air masses had recently been influenced by the evaporation of PSC particles and had not yet been diluted with surrounding air. From its nature, individual nitrification events occur within a limited time and space. The accumulation of these events and mixing processes will result in the systematic increase in NOy and compact correlation between NOy and N2O, as observed during SOLVE.
 AASE 2 was conducted over the Arctic between January and March 1992. Very similar N2O-NOy relationships were observed from the DC-8 in January and March, although systematically lower NOy values were observed in February [Weinheimer et al., 1993]. They showed that the very small change in the NOy-N2O relationship was consistent with the fact that denitrification was less extensive in that year. In fact, the AASE 2 N2O-NOy relationship is similar to the SOLVE MkIV/ER-2 relationship (Figure 15a).
 During the Polar Stratosphere Aerosol Experiment 1 (POLSTAR 1) campaign (January and February 1997), NOy values systematically higher than the earlier measurements were observed in the Arctic at potential temperatures between about 340 and 360 K from the Deutsches Zentrum fur Luft- und Raumfahrt (DLR) Falcon aircraft [Ziereis et al., 2000]. The NOy values were systematically higher than the SOLVE Mk-IV/ER-2 values by about 1 ppbv (Figure 15b). In this case, the dNOy/dN2O gradient was similar to those from the earlier DC-8 and ER-2 measurements, and the increase in NOy appeared with a positive offset. Because temperatures at 50 hPa during the 1996/1997 Arctic winter did not fall below TNAT before mid-January, which was just before the Falcon measurements, aircraft emissions were proposed as a source of reactive nitrogen [Ziereis et al., 2000].
 Nitrification was clearly detected from satellite measurements in the 1996/1997 Arctic winter at 12–16 km between mid-February and the beginning of March, soon after denitrification was observed at altitudes between 17 and 22 km [Kondo et al., 2000; Irie et al., 2001]. The area where the temperature was lower than TNAT extended to as low as 13 km in February and increases in HNO3 of about 1 ppbv were observed at the 370-K isentropic surface (∼13 km) at equivalent latitudes higher than 65°N. As in the 1999/2000 winter, the vertical temperature profile largely controlled the vertical extent of the nitrification influences.
5. Summary and Conclusions
 During the SOLVE campaign, measurements of NOy, N2O, and O3 were made in the Arctic lower stratosphere at altitudes between 6 and 12.5 km from the NASA DC-8 aircraft in December 1999 and January and late February/March 2000. In January and late February/March, measurements were also made at altitudes up to 21 km from the ER-2 aircraft.
 The N2O mixing ratio measured from the DC-8 ranged between 280 and 320 ppbv. The NOy-N2O correlations obtained in the December and January deployments were comparable and were also similar to the reference correlation established using the MkIV balloon measurements made during SOLVE prior to the onset of denitrification, which was in good agreement with the ER-2 extra-vortex measurements. During the March deployment, NOy values obtained from the DC-8 were slightly higher systematically than those observed during the previous deployments, although a compact correlation between NOy and N2O was maintained. The increase in NOy (ΔNOy = NOy – NOy* was especially evident (0.5 to 1 ppbv) in air masses with N2O values of 280–290 ppbv (altitudes of 11.0–12.5 km and potential temperature of 340–350 K).
 The daily minimum temperatures at 450–500 K (∼20–22 km) in the Arctic fell below TICE between late December and mid-January, and intensive denitrification was observed from the ER-2 at altitudes between 17 and 21 km in January and March [Popp et al., 2001]. In fact, large HNO3-containing particles with diameters of 10 to 20 μm were detected from the ER-2 between 20 January and 7 March [Fahey et al., 2001]. The number and spatial extent of these particles were significantly less after 20 January, in accordance with the increase in temperatures. Although measurements at altitudes below 16 km were limited to over northern Scandinavia, they show increases in NOy due to nitrification at 15–18 km and 14–16.5 km in January and March, respectively [Popp et al., 2001]. Considering that the vortex edge was often located near Kiruna, more intensive nitrification than that observed by the ER-2 was possible in the center of the vortex, where atmospheric temperatures were lower. The observed results presented in this paper indicate that nitrification extended to altitudes as low as 10–12.5 km over the wide area in the 1999/2000 Arctic winter.
 The significant effect of nitrification was observed only in late February/March at DC-8 altitudes, primarily because temperatures between 350 and 380 K (∼13.5–15.5 km) were generally high in January, and almost all of the PSC particles had evaporated at altitudes above 12.5 km. Because of the severe denitrification in January, nitrification, which occurred just above the DC-8 altitudes, followed by the diabatic descent of air, was considered to partly account for the observed increase in NOy in late February/March. In February, the Arctic minimum temperatures were continuously lower than TNAT at altitudes as low as 350 K (∼13.5 km), and more particles could efficiently fall to DC-8 altitudes, although ongoing denitrification was likely to be less intense, according to the ER-2 PSC particle measurements. Air masses influenced by nitrification were well mixed with surrounding air masses and the compact correlation between NOy and N2O was established before the DC-8 March deployment. The average increase in NOy (ΔNOy) was 400 pptv in air masses with MPV of 12–23 PV units (10−6 K m2 kg−1s−1) at potential temperatures of 320–350 K.
 The observed features are generally reproduced well by the REPROBUS 3-D CTM only when the denitrification process was included. In particular, the good agreement of the model-calculated NOy-N2O correlations with the observations in both the January and March deployments confirmed our interpretation that the increase in NOy observed from the DC-8 in late February/March was due to the redistribution of NOy.
 The results presented in this paper show the redistribution processes in the 1999/2000 Arctic winter affected the NOy levels at altitudes as low as 10–12 km over a wide area. The wide spread influence at these low altitudes was never observed during previous NASA/DC-8 and ER-2 observations in accordance with unusually cold temperatures in the 1999/2000 Arctic stratosphere. The results presented in this paper also show that the vertical structure of the temperature field is a critical factor determining the vertical extent of the NOy redistribution.
 We are indebted to all SOLVE participants for their cooperation and support. Special thanks are due to the flight and ground crews of the NASA DC-8 for helping make this effort a success. We thank T. Wada for his assistance with the data analyses and N. Toriyama, M. Kanada, and H. Jindo for their technical assistance with NOy measurements. We also thank D. W. Fahey and R. S. Gao at the NOAA Aeronomy Laboratory for providing ER-2 NOy data, D. Hurst and J. Elkins at NOAA CMDL, S. Schauffler and E. Atlas at NCAR, H. Jost at NASA ARC, C. Webster and R. Herman at JPL, and J. Greenblatt at Princeton University for providing ER-2 unified N2O data. We also thank P. A. Newman at NASA GSFC for calculating the modified potential vorticity values along the DC-8 flight track. We also thank R. Selkirk at NASA ARC for providing GEOS-3 meteorological analysis prepared for the SOLVE mission by the Data Assimilation Office (DAO) at NASA GSFC. The meteorological data were supplied by the European Center for Medium-Range Weather Forecasts (ECMWF) and the Norwegian Institute for Air Research (NILU). The modeling work presented in this study was supported by the European Union through the THESEO-2000/Eurosolve project. This work was also supported in part by the Ministry of Education, Science, Sports, and Culture (MESSC).