Evolution of the NOy-N2O correlation in the Antarctic stratosphere during 1993 and 1995



[1] The sources and sinks of stratospheric reactive nitrogen (NOy) in the Antarctic are known only qualitatively, because of the very few measurements of NOy available in this region. As a result, the effects of stratospheric NOy short- and long-term changes on the stratospheric concentration of ozone, water vapor, and other climate-forcing agents are still uncertain. To better understand the annual cycle of polar stratospheric NOy, we estimate its concentration in the Antarctic stratosphere during part of 1993 and throughout 1995. These estimates are obtained at seven potential temperature levels, extending from ∼18 to 30 km of altitude, and are associated with ground-based measurements of another tracer, N2O, in order to produce NOy-N2O correlation curves that can provide insights on nitrogen sources and sinks. To estimate NOy mixing ratios, we use ground-based and satellite measurements of major NOy constituents, connected by using air parcel trajectories and supplemented by model calculations of minor contributing species for which no suitable measurements exist. All the available NOy-N2O correlation points are averaged over three representative seasonal time periods in 1993 and six periods in 1995. Results show very similar correlation curves during the late summer and the fall of 1995, and again during the early spring 1993 compared with the early and late winter of 1995, although there are large seasonal changes due to transport and to condensation of NOy onto polar stratospheric clouds. We calculate a loss from the latter process of ΔN = (6.3 ± 2.6) × 107 kg of stratospheric nitrogen in the southern polar vortex during 1995. We also compare our correlation curves with those obtained in the Antarctic stratosphere during the Atmospheric Trace Molecule Spectroscopy mission ATMOS/ATLAS-3 in November 1994, finding important similarities but also critical differences that suggest that extra-vortex air is generally not an adequate representation of prewinter inner vortex conditions. Calculations of NOy winter removal in the Antarctic stratosphere which have used extra-vortex measurements as a surrogate for prewinter conditions may thus have underestimated true NOy removal. Our prewinter NOy estimates in the vortex core match values obtained by atmospheric models that incorporate upper atmospheric sources of NOy, supporting the belief that such sources have a significant effect on polar stratospheric NOy.

1. Introduction

[2] The correlation between reactive nitrogen (NOy ≈ HNO3 + NO + NO2 + 2 × (N2O5) + ClONO2 + HO2NO2) and N2O in the lower stratosphere is a commonly used tool to estimate nitrogen loss (denitrification) during polar winters [Fahey et al., 1989; Rinsland et al., 1996; Rex et al., 1999; Popp et al., 2001]. Denitrification takes place when HNO3-enriched Polar Stratospheric Cloud (PSC) particles sediment out of the stratosphere, causing a net loss of nitrogen. The amount of NOy left in the polar stratosphere after winter removal determines the duration of chlorine and bromine activation in spring and therefore has a key role in determining the severity of ozone depletion [e.g., Portmann et al., 1996]. The magnitude of winter denitrification is strongly dependent on stratospheric temperatures and aerosol concentrations, which are also closely linked to climate patterns. The denitrification process is generally limited to altitudes below ∼26 km (θ ≤ 655 K in winter) where temperatures reach the threshold for PSC particles formation.

[3] The NOy lost by the stratosphere during polar winters is produced in the stratosphere mainly at middle and lower latitudes through oxidation of N2O [Brasseur and Solomon, 1984]. The most important sources of NOy in the polar stratosphere indicated by theoretical studies [e.g., Vitt and Jackman, 1996] are transport from lower latitudes, local N2O oxidation, and the local dissociation of N2 by galactic cosmic rays (GCRs) (a permanent source modulated by solar activity), and sporadic solar proton events (SPEs), which are very energetic but intermittent events originating from solar flares. Vitt and Jackman [1996] calculate the yearly averaged contribution made by the sources listed above to the production of reactive nitrogen at polar latitudes. Their results show that transport from lower latitudes is the most important contribution (∼80%), while local N2O oxidation and GCRs share evenly the remaining 20%, except during large SPEs, which can contribute to the stratospheric NOy formation by as much as to 12%. However, Vitt and Jackman [1996] do not consider one more source of polar stratospheric NOy: the mesospheric NOy produced through the dissociation of N2 by GCRs, SPEs, and precipitations of energetic electrons (EEPs, which do not get to the stratosphere). This NOy is transported down to the stratosphere at high latitudes during fall and winter. Although this source of polar stratospheric NOy has been proven to be significant [Solomon et al., 1982; Garcia and Solomon, 1994; Callis and Lambeth, 1998; Callis et al., 2002; Siskind et al., 2000; de Zafra and Smyshlyaev, 2001], it is still neglected in many two dimensional (2-D) and three dimensional (3-D) models (e.g., see model intercomparisons at high latitudes in Park et al. [1999]). The effects of thermospheric and mesospheric production of NOy by energetic electrons precipitation (EEP) on the concentration of stratospheric NOy are evaluated by Callis et al. [2001] using a 2-D transport and chemistry model. They present model results averaged over 18 years, from 1980 through 1997. Their results at high southern latitudes show that NOy production by EEPs in the upper atmosphere, coupled with downward transport during fall and winter, produces a large enhancement of stratospheric NOy when EEPs are considered versus when they are not. Though downward transport occurs in fall and winter, this enhancement is still noticable in the following late summer/fall period with an average increase of 20% in NOy between 20 and 30 km altitude over the Antarctic region. Similar results are obtained by Nevison et al. [1997] using the Garcia-Solomon 2-D model [Garcia and Solomon, 1994], which displays a large downward flux of NOy in the winter hemisphere and an NOy-enriched layer still residing in the 24–32 km range in the summer hemisphere, again a remnant from the descent that occurred during the preceding winter.

[4] Several studies [Fahey et al., 1989; Keim et al., 1997; Rinsland et al., 1996; Rex et al., 1999; Strahan, 1999] estimated the parameters that characterize the NOy-N2O curve when denitrification is absent, mostly at midlatitudes or outside the polar vortices. These reference curves were then compared to NOy-N2O profiles obtained inside the vortices, allowing an estimate of the nitrogen lost for condensation onto and sedimentation of PSC particles. However, the contribution to stratospheric NOy from descent of NOy-enriched thermospheric and mesospheric air over polar regions is not accounted for in reference curves that belong to midlatitude regions. Moreover, in recent years, Michelsen et al. [1998], Kondo et al. [1999], Strahan [1999], Plumb et al. [2000], and Esler and Waugh [2002] have pointed out that mixing of air across the edge of polar vortices can be an important contributor to defining the NOy deficit developed inside the vortices. As a consequence, reference curves obtained at midlatitudes or outside the vortices are not always a reliable representation of prewinter conditions within vortex regions.

[5] Despite the usefulness of NOy-N2O correlations at high latitudes, very few measurements of NOy and N2O concentrations have been carried out in the Antarctic stratosphere, where the relative stability of the southern vortex would allow useful studies of NOy annual cycles and interannual variability. The existing data sets are either very limited in time (such as the Atmospheric Trace Molecule Spectroscopy (ATMOS)/ATLAS-3 measurements that extend for less than 2 weeks [e.g., Rinsland et al., 1996]) or, when lasting for several months [e.g., Keim et al., 1997], cover a small altitude range (in potential temperature, between 360 K and 520 K).

[6] This work aims at improving our knowledge of polar stratospheric NOy by considering an entire annual cycle of NOy behavior, addressing unresolved issues such as the characteristics of the prewinter NOy-N2O correlation in the vortex region, the contribution to stratospheric NOy from mesospheric and thermospheric production, and a quantification of the winter denitrification process in the Antarctic stratosphere. A relatively high degree of isolation is reached in the Antarctic vortex region in winter, and the NOy concentration inside the vortex is usually not strongly perturbed by mixing with extra-vortex air. The above mentioned issues can thus be investigated in the southern polar region with better clarity than in the northern. We will employ a combination of measurements, supplemented by box model chemical calculations, to construct NOy-N2O profiles between ∼18 and 30 km altitude in order to study the NOy annual behavior cycle. The measurements and the box model used are described in the following section.

2. Measurements and Box Model Description

[7] The most abundant NOy constituent from ∼15 km to 27–28 km in altitude is HNO3, while NO2 and NO prevail from ∼28 km upwards, throughout most of the year [Brasseur and Solomon, 1984; Kawa et al., 1992]. The NOy estimates presented here cover the period from October to December (i.e., austral spring-summer), 1993, and the entire year 1995. They are derived from measurements of HNO3, by a ground-based millimeter-wave spectrometer (GBMS) [de Zafra et al., 1997; McDonald et al., 2000; Muscari et al., 2002] operating during 1993 and 1995 from the Amundsen-Scott base located at the South Pole (90°S); NO and NO2 by the Halogen Occultation Experiment (HALOE; data version 19) aboard UARS [Russell et al., 1993]; NO2 by the Polar Ozone and Aerosol Measurement II instrument (POAM II; data version 6) aboard the French SPOT-3 polar-orbiting satellite [Glaccum et al., 1996]; and calculations of minor NOy constituents such as N2O5, ClONO2, and HO2NO2 from a photochemical box model derived from the SUNY-SPb 2-D model [Smyshlyaev et al., 1998].

[8] The overall uncertainty in retrieved GBMS HNO3 measurements (including both systematic and random errors) varies with altitude and time and is bounded approximately between 16% and 22% [de Zafra et al., 1997]. Total relative errors for HALOE NO and NO2 in the altitude range 18–30 km are dominated by systematic uncertainties and decrease monotonically with altitude, ranging 15–50% for NO and 7–45% for NO2 [Gordley et al., 1996]. POAM II NO2 random errors decrease with altitude from 35% at 20 km to 20% at 25 km and drop to 10% between 30 and 40 km [Randall et al., 1998]. We have scaled the POAM II mixing ratio values up by a factor of 1.15 to account for a systematic bias observed by Randall et al. [1998].

[9] HALOE and POAM II data employed in this study complement each other in time, and together provide coverage of NO2 for the time periods covered. HALOE measures both NO and NO2, while POAM II provides only NO2 profiles. Consequently, in those seasons when HALOE data reach latitudes south of 60° (the northern latitude limit for this work) we have given them preference. In the austral spring and summer of 1993 and 1995 HALOE NO and NO2 measurements reach latitudes south of 60°, whereas from April to September 1995 POAM II data are instead used to calculate NOy estimates. In October 1993 and March and October 1995 both data sets were used to evaluate the reliability of model estimates of NO during periods of POAM II data usage. During these periods, HALOE and modeled NO values agree within the uncertainties of the HALOE NO measurements (and always within ∼1.5 ppbv), without showing any systematic difference.

[10] Measurements of N2O to be correlated with the NOy estimates were carried out at the South Pole by the GBMS [Crewell et al., 1995; Cheng et al., 1997]. They are characterized by an overall uncertainty of <40 ppbv below 20 km, reducing to <20 ppbv for altitudes higher than 25 km [Crewell et al., 1995]. All the data sets were interpolated onto 7 potential temperature (θ) levels: 465 K, 520 K, 585 K, 620 K, 655 K, 740 K, and 960 K, (∼18–30 km) using the daily global assimilated pressure and temperature data from the U.K. Meteorological Office [Swinbank and O'Neill, 1994].

[11] The box model used in this study is derived from the SUNY-SPb 2-D model described by Smyshlyaev et al. [1998] and employs the same photochemistry, microphysics, and radiation codes. The 2-D model has passed benchmark tests for chemistry and photolysis given in Stolarski et al. [1995] and has participated in the Models and Measurements Intercomparison II [Park et al., 1999]. With the exception of HNO3, NO2, and NO, initial concentrations of all the chemical species needed to run a full chemistry model were provided for the box model from zonal averages obtained with the 2-D model. Before running the box model, however, the 2-D zonal averages were first brought to a steady state equilibrium with the measured HNO3, NO2, and NO.

[12] Since GBMS took measurements over the South Pole, while both HALOE and POAM II sampled the Antarctic stratosphere at locations mostly north of 80°S, measurements of HNO3 and NOx (NOx = NO + NO2) are never colocated. To compensate for this, we used air parcel trajectories to connect the Pole with locations of HALOE or POAM II measurements. Although the trajectory calculations are themselves subject to uncertainties, several studies have recently employed trajectory-based techniques to carry out data intercomparisons [Morris et al., 2000; Danilin et al., 2002; Muscari et al., 2002], reaching the conclusion that use of backward and forward trajectories is a reliable tool for data validation studies. Morris et al. [2000] and Danilin et al. [2002] have shown that trajectory tracing does not significantly decrease the accuracy of results with respect to traditional approaches that use only colocated data, while greatly increasing the number of data points available for validations, therefore improving the statistical significance of results as well as their spatial coverage. In particular, for HNO3 and NO2 validations, Danilin et al. [2002] found a discrepancy of only <0.5 ppbv and <5%, respectively, when comparing data sets using a trajectory-based technique instead of the tradition approach. On the basis of such results, we attribute a relative contribution of 5% to the NOy uncertainties coming from the use of trajectory tracing (see also section 3).

[13] To minimize uncertainties in trajectory calculations, we trace air parcel movement for no more than 5 days and compute daily averages to find a mean trajectory from a bundle of 8 trajectories originating from a small geographic spread of initial points around the South Pole. We employed the quasi-isentropic Goddard Space Flight Center (GSFC) trajectory model [Schoeberl and Sparling, 1995], which advects air parcels backward or forward in time with respect to an initial position using horizontal wind velocity fields and diabatic heating rates.

[14] Our estimates of NOy mixing ratios over the South Pole at a specific θ level are obtained by running the box model with a 3-min step for photochemistry, while updating pressures, temperatures, and solar zenith angles every 30 min along the air parcel trajectories that link GBMS measurements with those carried out by HALOE or POAM II. When a connecting trajectory starts at the Pole, we match the box model's initial HNO3 mixing ratio with that measured by GBMS over the South Pole, and we change the model's initial NOx (or NO2) until the final NOx (or NO2) mixing ratio at the end of the same trajectory matches the value measured by HALOE (or POAM II). If a trajectory ends at the South Pole, we apply the same method to estimate NOy, but now the role of HNO3 and NOx (or NO2) is reversed: the model's initial NOx (or NO2) is set to match HALOE (or POAM II) data, and the model's initial HNO3 is changed until HNO3 at the end of the run matches the GBMS value measured over the South Pole. Therefore for each air parcel, the most important NOy species are constrained by measurements at one end or the other of a maximum 5-day path, and only mixing ratios of minor constituents rely solely on modeling. Examples of species mixing ratios at both ends of a few air trajectories are listed in Table 1 for a variety of time periods and θ levels. Note that model runs always involve a satellite observation at sunset or sunrise (both HALOE and POAM II employ a solar occultation technique) connected to a GBMS observation with a different solar zenith angle at the Pole. In Table 1, NO and NO2, which have a strong, rapid diurnal cycle in the stratosphere, may hence show relatively large changes between starting and ending values, although NOx (= NO + NO2) remains a quasi-conserved quantity. Out of the several HALOE (or POAM II) profiles intersected by an air parcel trajectory along its path to or from the Pole, we used the profile closest in time (within a 5-day limit) to the day of HNO3 measurements over the South Pole. When NOy estimates for the same day of South Pole data are available on both forward and backward trajectories, they are averaged together to produce the NOy estimate for that day.

Table 1. Examples of Mixing Ratios of Measured Species (HNO3, NO2, and NO), Modeled Species (Grouped Together Under “Others”), and Total NOy for Air Trajectories at Three Different θ Levelsa
 520 K620 K740 K
  • a

    Concentrations are listed at the beginning (under B) and at the end (under E) of each air trajectory. Examples are all from 1995 results. The day of arrival (departure) of the air trajectory to (from) the South Pole is indicated with mm/dd in italics. Next to the date, fw and bk indicate forward and backward trajectories, respectively. Note that mixing ratio values over the South Pole are at the end (under E) of a backward trajectory (bk), but at the beginning (under B) for a forward trajectory (fw). NO* indicates that the NO mixing ratio is modeled, not measured. In these cases POAM II and not HALOE data have been employed, and consequently also different connecting trajectories have been used. The abbreviation “ng” in the mixing ratio boxes stands for “negligible.”

02/01 fw
04/08 bk
04/08 fw
07/23 bk
10/12 bk
10/12 bk
12/01 fw

[15] All the estimates obtained for NOy, as well as N2O measurements, are averaged in 3 periods during 1993 (designated as early spring, late spring, and early summer) and 6 periods during 1995 (late summer, fall, early winter, late winter, early spring, and late spring), with the first day, the last, and the total number of days for each period listed in Table 2. The choice of these periods of time was based on seasonal changes in NOy and N2O mixing ratios and on the available days of data.

Table 2. Definition of the Time Periods (Three in 1993 and Six in 1995) Used to Average NOy and N2O Dataa
 Early SummerLate SummerFallEarly WinterLate WinterEarly SpringLate Spring
  • a

    Each period is described by the first day, the last day, and the total number of days of data averaged together. Depending on the air trajectories that connect the ground-based to the satellite data, different θ levels often have a different number of days of data that can be averaged together within a time period. For this reason the last row of the table indicates a range of available days for each period.

Starts6 December 19931 February 19953 April 199528 June 199515 August 19951 October 19935 October 199521 November 199324 November 1995
Ends1 January 199424 March 199516 May 199531 July 199526 September 199519 October 199325 October 19953 December 19934 December 1995
Number of days of data2–45–95–74–56–72–32–63–42–3

3. NOy-N2O Correlation Curves

[16] The correlations for estimated NOy versus N2O are shown in Figures 1 and 2. In each panel of both figures, we have plotted NOy-N2O data points belonging to different time periods using different symbols (either triangles or squares), while the symbol's color identifies the θ level. In Figure 1, we have plotted all the data obtained for the year 1995, with late summer and fall displayed in Figure 1a, early and late winter in Figure 1b, and early and late spring in Figure 1c. Figure 2 shows results from the 3 available seasons of 1993 compared to data from 1995, with early springs in Figure 2a, late springs in Figure 2b, and early summer 1993 versus late summer 1995 in Figure 2c. Fourth-order polynomial fits to the correlation curves for fall 1995, early and late winter 1995, and the early summer 1993 data are also shown. Their corresponding parameters are listed in Table 3. The curves from these 3 periods were chosen to represent the NOy-N2O correlation in the Antarctic stratosphere before, during, and after winter, respectively. Error bars for the NOy seasonal averages are calculated by adding in quadrature the systematic error of HNO3 measurements, the systematic error of NOx measurements, the 5% uncertainty due to the use of air trajectories to connect data sets (see section 2), and the standard deviation of the mean obtained when binning daily NOy values in seasonal averages. Error bars for the N2O seasonal averages are calculated as δ(N2O)avg = Σiδ(N2O)i/N, where N is the total number of observations averaged together in a season, and the δ(N2O)i are the total (systematic + random) uncertainties of individual measurements.

Figure 1.

NOy-N2O correlation points for the year 1995. (a) Late summer and fall. (b) Early winter and late winter. (c) Early spring and late spring. The color bar at the bottom keys each data point to the θ level that it represents. Solid black lines show fourth-order polynomial fits to the data, with the parameters of the polynomial curves listed in Table 3.

Figure 2.

NOy-N2O correlation points for the year 1993, compared with 1995 data. Correlation curves for early spring are in Figure 2a, for late spring in Figure 2b, and for early summer 1993 and late summer 1995 in Figure 2c. The 1995 data are also shown in Figure 1. The color bar at the bottom keys each data point to the θ level that it represents. The solid thick and the dashed thin black lines in Figure 2a are the fit to the winter 1995 data (also shown in Figure 1b) and the in-vortex ATMOS/ATLAS-3 data from November 1994 (ATMOS-in), respectively. The solid black line in Figure 2c shows a fourth-order polynomial fit to the early summer 1993 data, with its polynomial parameters listed in Table 3.

Table 3. Parameters Corresponding to the Fourth-Order Polynomial Fits to the Correlation Points From Fall 1995, Winter 1995, and Early Summer 1993
 Fall 1995Winter 1995Early Summer 1993
K10.2−1.2−3.2 × 10−2
K2−9.8 × 10−42.3 × 10−2−4.6 × 10−5
K3−2.7 × 10−6−1.8 × 10−4−1.2 × 10−6
K41.3 × 10−84.9 × 10−73.5 × 10−9

[17] In Figures 1 and 2, correlation points representative of each θ level move up or down along the NOy axis, and left or right along the N2O axis, depending on the physical and chemical processes that affect the stratosphere inside the polar vortex. In order to better display the movements of correlation points with time at a constant θ level, in Figure 3 we show the 1995 time series of correlation points at 585 K (solid squares) and 960 K (solid triangles), with different colors indicating different time periods. (See legend and caption for details.)

Figure 3.

The 1995 time series of NOy-N2O correlation points for levels 585 K and 960 K. The color bar specifies the time period that each point represents.

[18] If air descends from higher to lower altitudes, each point will shift along the correlation curve toward the position occupied in the previous season by correlation points representative of higher θ levels. Since higher altitudes have lower N2O mixing ratios, air descent will shift correlation points to the left along the N2O axis. At 960 K, the top θ level used in this study, the movement of the point along the NOy axis will give us an insight on the NOy values at altitudes just above 960 K. Correlation point movements corresponding to air descent are observed from late summer to fall (Figures 1a and 3) and from early winter to late winter (Figures 1b and 3).

[19] When NOy is removed from the gas phase and condenses on aerosol and PSC particles (a process generally confined to θ ≤ 655 K), all the points representative of affected θ levels should move down along the NOy axis, while remaining at a constant N2O mixing ratio. The passage from fall to early winter at level 585 K in Figure 3, for instance, (light blue and green squares, respectively) shows the simultaneous occurrence of nitrogen removal and air descent.

[20] We next consider the effect due to mixing of air inside the vortex with extra-vortex air. During winter and early spring, the lower θ levels inside the vortex have small values of N2O, due to air descent from higher altitudes, accompanied by small values of NOy due to reactive nitrogen condensation onto PSC particles. When this air mixes with extra-vortex air during early and late spring, the corresponding NOy-N2O correlation point moves upward along the NOy axis and to the right along the N2O axis, as seen for level 585 K in Figure 3. At higher altitudes which have never been subject to denitrification, such as levels 960 K and possibly 740 K, NOy mixing ratios inside the vortex depend on the earlier NOy downflux from levels above our altitude range of study. However, the N2O values inside the vortex are still smaller than those present on the same θ level at midlatitudes because of descent [e.g., Manney et al., 1999]. Therefore mixing with extra-vortex air which occurs at these altitudes will still move the correlation point toward the right side of the N2O axis, as shown in Figure 3, but the movement along the NOy axis is less predictable. With these generalities in mind, we next take up analysis of specific behavior exhibited by our correlation curves during various periods.

3.1. 1995 Annual Cycle

[21] During late summer and fall of 1995 (Figure 1a), with temperatures still above the HNO3 condensation threshold, NOy and N2O mixing ratio values are negatively correlated for N2O > 100 ppbv and positively correlated for N2O < 70 ppbv, similar to reference curves observed in previous studies [e.g., Fahey et al., 1989]. Although data points from both late summer and fall lie on the same correlation curve (with the only clear exception being the late summer level at 960 K), all the fall points are shifted along the curve toward significantly lower N2O values with respect to late summer data. This translation along the curve signifies air descent.

[22] We believe that the fall NOy-N2O correlation in Figure 1a is indicative of 1995 conditions throughout the vortex core just before gas-phase nitrogen removal starts. This is supported by two arguments. First, potential vorticity (PV) maps show that the South Pole was always inside the developing vortex during fall 1995, so that GBMS HNO3 and N2O measurements must therefore depict the in-vortex stratosphere. Since PV is a quasi-conserved quantity on 5-day air trajectories, the HALOE NOx (and POAM II NO2) measurements we used also sampled inside the developing vortex. Secondly, the seven daily correlation curves available for the fall period (see Table 2) show only very small and gradual changes during the entire period, suggesting that no abrupt mixing with middle-latitude air took place during our defined fall period. Although there is a gap of about 1.5 months between the end of our fall period and the beginning of our early winter period (see Table 2), the GBMS measurements of HNO3 presented by de Zafra et al. [1997] and Muscari et al. [2002] show that nitrogen removal from the gas phase starts during the first week in June, much earlier than the beginning of our early winter period, and proceeds very rapidly. This indicates that there are only 2–3 weeks between the end of our fall period and the beginning of nitrogen removal, during which mixing with extra-vortex air might alter our fall NOy-N2O correlation. At this time of the year (approximately the end of May), however, the Antarctic vortex is well formed and isolated, and mixing with extra-vortex air is very unlikely.

[23] A sudden change takes place when temperatures in the vortex drop below ∼195 K in the lower stratosphere, and HNO3 gas-phase depletion occurs because of the formation of PSCs. Figure 1b shows early and late winter correlations that are clearly affected by NOy depletion, and by N2O values that have decreased at all levels from fall to early winter because of descent of N2O-depleted air from the upper stratosphere. Both early winter and late winter correlation points depict a similar relationship between NOy and N2O, signifying a vortex core that has remained well isolated from the nondenitrified regions outside. Supporting this observation, a recent study of mixing properties within the Antarctic vortex by Lee et al. [2001] shows that the vortex core is generally strongly separated from midlatitude air by a broad region of weak mixing between equivalent latitudes of 58°S and 68°S (the vortex-edge region). Lee et al. [2001] also point out that the air belonging to the vortex core is, in contrast, very well mixed laterally and therefore stratospheric constituents will be rather uniformly distributed inside this region.

[24] As stated earlier, downward transport inside the vortex is signified by the displacement of data points toward smaller N2O values along the winter correlation curve. At each of the 7 levels, the NOy mixing ratio increases from early to late winter, as N2O decreases. During both early and late winter, the 960 K level (∼30 km) is characterized by N2O ≤ 10 ppbv (see Figures 1b and 3), which is a signature of upper stratospheric and mesospheric air [Rinsland et al., 1999]. The only region where this upper atmospheric air reaches the lower stratosphere is at high latitudes in winter. We would not observe the large amount of NOy (∼18 ppbv) we detect at 960 K coupled with this small amount of N2O (see Figure 1b) if no net production of NOy took place in the upper atmosphere. In fact, ATMOS measurements from middle latitudes, as well as just outside the polar vortex, show NOy ≤ 11 ppbv for N2O ≤ 10 ppbv (see ATMOS-out profile discussed in section 3.3 and shown in Figure 4). The significance of this NOy input is expected to vary with time during winter, as the column of NOy formed in the upper stratosphere and mesosphere descends to levels 960 K and possibly 740 K. This is very nicely illustrated by POAM II and POAM III NO2 data [Randall et al., 1998; Randall et al., 2001] which show the downward motion of the NO2 peak from early July to the end of October. At 740 K, the increase in NOy mixing ratio from early to late winter is what we would expect from air descent. However, given the uncertainties in temperature thresholds for PSC particles formation as well as the uncertainties in temperature measurements, we cannot rule out the possibility that some of the increase of NOy at 740 K in late winter may be due to evaporation of PSC particles.

Figure 4.

Polynomial fits to 1993 and 1995 data (same as solid thick lines in Figures 1 and 2), and ATMOS/ATLAS-3 extra-vortex profile (ATMOS-out) from November 1994 (65–73°S). The shaded area shows the ±1σ uncertainties for ATMOS-out.

[25] We next calculate separate NOy column densities corresponding to our fall and early winter correlation curves, covering the range 10–250 ppbv of N2O in fall (∼18–33 km) and 10–180 ppbv in early winter (∼18–28 km). Although these calculations are performed over a different N2O mixing ratio range in early winter with respect to fall, the NOy at altitudes corresponding to N2O mixing ratios between 180 and 250 ppbv is extremely small in winter (see Figure 1b) and therefore can be neglected. We obtain (2.0 ± 0.4) × 1026 molecules of NOy km−2 (or 4.7 ± 0.9 kg km−2 of nitrogen) for the 1995 fall period, while the column density drops to only (0.28 ± 0.06) × 1026 molecules of NOy km−2 (or 0.6 ± 0.1) kg km−2 of nitrogen for the 1995 early winter period, a change of more than a factor of 7. In order to estimate the total loss of nitrogen in the Antarctic stratosphere, we consider the area inside the Antarctic vortex that is generally characterized by severe HNO3 gas-phase depletion during early winter. In Plate 4a of Santee et al. [1998], the boundary of this region appears to be at 70°(±1°)S equivalent latitude, which encloses approximately (15.4 ± 3.0) × 106 km2, an area slightly larger than the size of Antarctica. Using this area and the fall-to-winter change in column densities estimated above, we find that the nitrogen lost by the stratosphere during the Antarctic winter is ΔN = (6.3 ± 2.6) × 107 kg. This value for ΔN does not take into account some additional NOy removed from the lower stratosphere due to continued descent of air later in winter.

[26] Figure 1c shows how the correlation points shift from early to late spring 1995, when the southern polar vortex starts to weaken, and mixing with the edge region and with external air leads to renitrification of the Antarctic stratosphere. From late winter (squares in Figure 1b) to early spring (triangles in Figure 1c), an increase in both NOy and N2O occurs at the lower θ levels (see also Figure 3). Part of the NOy increase could have also been caused by evaporation of PSC particles as temperatures rise above the evaporation threshold at the end of the late winter period. However, we believe that these evaporating PSC particles are not those that formed three months earlier, at the beginning of winter, in conjunction with the NOy removal we observed at that time (Figures 1a and 1b). The Antarctic lower stratosphere throughout winter is characterized by temperatures consistently well below the PSC formation threshold, insuring that early-forming particles will settle out by late-winter [Hamill and Toon, 1991; Cacciani et al., 1997]. Evaporation of more recently formed PSC particles in late September/early October does not affect our ΔN calculation since the NOy that they release back to the gas phase was never counted in ΔN. This evaporation could explain why, however, the early spring NOy mixing ratio appears larger at 465 K than at 520 K (Figure 1c). It is certainly clear from PV maps of the mid-October lower stratosphere that the GBMS measurements of HNO3 and N2O sampled air inside the vortex, so that these observed spring changes in tracer concentrations were not due to a displacement of the vortex with respect to the South Pole.

[27] In early spring, NOy values at 960 K and 740 K are still probably dominated by air that has descended from higher altitudes rather than by mixing of in-vortex with extra-vortex air, since their corresponding N2O values are still very small. We would expect air at 740 K to show an NOy value in early spring close to the 18 ppbv that was observed at level 960 K in late winter, but Figures 1b and 1c show instead a decrease in early spring by ∼5 ppbv. The reason for this decrease is not clear. Condensation of HNO3 on aerosol or PSC particles can be ruled out, since at this level temperatures in the vortex core and along air trajectories are always too high. In spring, air descent slows down [e.g., Crewell et al., 1995] and the NOy peak that we see in late winter at 960 K cannot have sunk much below level 740 K. It is possible that the data sets we employ for NOy estimates do not have the necessary vertical resolution to detect correctly the sharp peak in NOy-N2O plots that forms in early spring at ∼740 K, and such a peak would therefore be washed out when averaged with lower NOy values above and below 740 K. It is worth noting, however, that NOy-N2O correlations from early spring 1993, early spring 1995, and ATMOS data obtained inside the Antarctic vortex during November 1994 (whose NOy mixing ratio peak is also located at ∼740 K [Rinsland et al., 1999]) all agree very well at level 740 K (see Figure 2a discussed in the next section).

[28] The late spring period shows a partial return of the Antarctic stratosphere to prewinter conditions, with mixing of in-vortex and extra-vortex air producing an increase of both NOy and N2O at most levels, starting first in the middle to upper stratosphere [see also Crewell et al., 1995]. Such a recovery is particularly advanced at 740 and 960 K by late November, when PV maps display no remnants of the vortex. Level 465 K is the only altitude where there is a significant reduction in NOy from early to late spring, possibly because the early spring values can be enhanced relative to extra-vortex air by the evaporation of remaining PSC particles gravitationally settling from higher altitudes (see discussion above).

3.2. Interannual Variations

[29] Results for early and late spring 1993 versus 1995, and early summer 1993 versus late summer 1995 are shown in Figure 2. In Figure 2a, we also show the winter 1995 fit (solid thick line) and the in-vortex ATMOS data from November 1994 (dashed thin line), while in Figure 2c we overlay the fit to the early summer 1993 correlation, which will be compared in the next section to extra-vortex ATMOS data. During all of the three periods, 1993 results are lower in N2O, and often in NOy, with respect to 1995 data. Although PV maps do not indicate that the vortex broke up earlier in 1995 with respect to 1993, this does not preclude mixing events in early spring 1995 which could have caused the differences in tracers concentration that we see in Figures 2a and 2b.

[30] It is interesting to note that results for early spring 1993 lie close to the same correlation curve as early and late winter 1995, except for the value at 960 K (see Figure 2a). This general agreement suggests that the physical conditions, mainly temperature and transport, which determine the shape of the winter NOy-N2O correlation inside the vortex, were very similar during 1993 and 1995 if we assume early spring 1993 is a hold-over proxy for winter 1993. A similarity in the physico-chemical conditions inside the winter vortices of 1993 and 1995 is also supported by vertical descent rate [Kawamoto and Shiotani, 2000], average temperature [Kawamoto and Shiotani, 2000], and NOy downward flux [Siskind et al., 2000] data from 1993 and 1995.

[31] In Figure 2a, we notice a large difference in NOy at 960 K, nearly 4 ppbv, between the early springs of 1993 and 1995. The corresponding small N2O values at 960 K together with NOx downward flux data from 1993 and 1995 [Siskind et al., 2000] render unlikely the possibility that this difference is due to either horizontal mixing of in-vortex with extra-vortex air or different NOy vertical fluxes in 1993 versus 1995. As discussed in section 3.1, during early spring NOy-N2O correlations are so steep at the upper-altitude end of the curve that the vertical resolution of the data sets employed could play a role in apparent values retrieved at those levels.

[32] NOy estimates and N2O measurements for early summer 1993 suggest that during this period (i.e., December) the lower stratosphere was still affected by the dissipating vortex. This is also suggested by the individual NOy-N2O correlations (not shown), which have gone into the early summer average. These evolve rapidly toward larger NOy and N2O values, especially throughout the first half of December. We believe that results found in this work should be generally applicable to other years which show meteorological and dynamical conditions similar to those of 1993 and 1995.

3.3. Comparison With ATMOS/ATLAS-3 Correlation

[33] Figure 4 shows all our polynomial fits to 1993 and 1995 NOy-N2O results (thick solid lines in Figures 1 and 2) and the ATMOS/ATLAS-3 extra-vortex correlation (hereafter referred to as ATMOS-out; thin solid gray line, with 1σ uncertainties outlined by the shaded area). The ATMOS/ATLAS-3 data were obtained using concurrent sunrise measurements of NOy constituents and N2O mixing ratios during 4–11 November 1994. These measurements were carried out at latitudes between 65°S and 73°S, and the corresponding NOy-N2O correlations are representative of conditions both inside the polar vortex (see the ATMOS-in correlation curve shown in Figure 2a) and outside the vortex (ATMOS-out). While the ATMOS N2O mixing ratio is measured with a ±5% uncertainty, NOy concentrations were obtained by Rinsland et al. [1996] by adding together ATMOS measurements of the most important NOy constituents, with a resulting estimated NOy uncertainty of ±15%. The ATMOS/ATLAS-3 measurements belong to a different year with respect to the NOy-N2O correlations presented here, and are very limited in time, but they cover a wide range of altitudes. Since several publications relating to NOy species in the Antarctic stratosphere have used ATMOS-out as a proxy for prewinter conditions within the region of vortex formation [e.g., Rinsland et al., 1996], we believe it is both interesting and useful to compare our results with the ATMOS-out correlation.

[34] The relative difference in NOy between our estimated prewinter conditions (the fit to fall 1995; solid thick black line in Figure 4) and ATMOS-out is ∼20% at the NOy mixing ratio peaks. If we calculate the winter nitrogen loss ΔN (see section 3.1) using the ATMOS-out curve instead of our fall NOy-N2O correlation, we obtain ΔN ∼ 4.8 × 107 kg, which is about 24% smaller than the estimated denitrification obtained using our fall 1995 correlation. The 1994 ATMOS-out curve, measured in mid-spring, shows close agreement with the polynomial fit we obtained for the early summer period of 1993 (dashed thick black line in Figure 4). On the basis of this agreement, we argue that ATMOS-out probably represents a transition period in the polar stratosphere, during which the NOy-N2O correlation evolves continuously from the breakup of the vortex in the spring to its reformation during the following fall.

[35] Another cause for the difference between our fall 1995 predenitrification values and the spring 1994 ATMOS-out measurements can lie in the additional contribution to polar stratospheric NOy from mesospheric and thermospheric sources. This contribution occurs because of the strong downward transport of air inside the vortex during fall and winter and is still noticeable at high latitudes during the following summer in 2-D model simulations [Nevison et al., 1997; Callis et al., 2001]. NOy that has descended from the mesosphere may therefore affect late summer and in-vortex fall NOy-N2O correlations, and possibly even fall correlations just outside the newly forming vortex, but it will be washed out in the region outside the vortex by the following spring, when ATMOS measurements were carried out.

[36] Values we obtained for the NOy peak in late summer and fall of 1995 agree with the pocket of large NOy mixing ratios predicted by the 2-D model of Nevison et al. [1997] for the summer lower stratosphere. We point out, however, that these model results show an even larger mixing ratio peak for NOy (≥25 ppbv) coincident with smaller values of N2O than we observe for our late summer NOy-N2O correlation. This difference may imply that the Garcia-Solomon 2-D model used in Nevison et al. [1997] underestimates meridional transport at high latitudes in summer. The ∼20% difference between our prewinter correlation and ATMOS-out at the NOy peak is also in good agreement with model results by Callis et al. [2001] for the late summer/fall period. Their results, averaged over an 18-year period (1980–1997) show a 20% increase in NOy between 20 and 30 km altitude over the Antarctic region when mesospheric sources (EEPs) of NOy are taken into account versus when they are not.

[37] In recent work by Muscari et al. [2002], the GBMS HNO3 data included in this study were compared to the UARS/Microwave Limb Sounder (MLS) version 5 HNO3 profiles, showing good agreement during most seasons, especially below θ ∼ 655 K. However, during the fall of 1995, which is relevant to our NOy-N2O comparison against ATMOS-out, GBMS values of HNO3 from the South Pole appear larger by ∼3 ppbv in the range between 520 and 960 K than MLS data from the 70–80°S latitude band. Although a decrease of ∼3 ppbv at most levels in our fall NOy estimates (to match MLS HNO3 values) would bring a closer agreement between our prewinter NOy-N2O correlation and ATMOS-out (see Figure 4), Muscari et al. [2002] provide strong evidence that differences between GBMS and MLS during fall are not due to instrumental biases but to lack of colocation between the two data sets during a period of strong poleward gradients in HNO3. Therefore we do not favor the hypothesis of a high bias in GBMS HNO3 as a valid explanation for the difference between our NOy-N2O prewinter correlation and ATMOS-out.

4. Summary

[38] Changes in NOy mixing ratios in the stratosphere have important consequences for seasonal ozone depletion. Moreover, the strength and importance of NOy sources and sinks at polar latitudes is determined by stratospheric temperatures, aerosols concentration, water vapor mixing ratio, solar activity, and other parameters also connected to climate processes. NOy sources and sinks are therefore an additional link between climate forcing and ozone depletion.

[39] To compensate for the scarcity of comprehensive NOy measurements in the Antarctic stratosphere, we have estimated the NOy concentrations in the lower stratosphere during 1993 and 1995, using a combination of ground-based and satellite measurements for the most important NOy components. These measurements, though not colocated, are connected by air trajectory calculations and supplemented with model calculations of minor NOy species. We correlate NOy estimates with N2O measurements, which provide additional information on NOy sources and on downward transport of upper atmospheric air. The Antarctic vortex air is more isolated during winter than air in the Arctic vortex, and large mixing events that would alter the NOy concentration inside the vortex are much rarer, therefore simplifying the characterization of NOy sources and sinks in the southern polar region.

[40] Our results indicate a winter nitrogen removal ΔN = (6.3 ± 2.6) × 107 kg, integrated over the area of intense denitrification. The consistency between early spring 1993 and winter 1995 results, as well as studies by others, suggests that the Antarctic vortex physico-chemical winter conditions may have been very similar during these two years. We believe the results obtained here for 1993 and 1995 should be a good indicator of NOy-N2O correlations for any years when similar meteorological and dynamical conditions exist. Differences between our prewinter vortex region NOy-N2O correlation and extra-vortex ATMOS/ATLAS-3 measurements recommend caution over the use of extra-vortex NOy-N2O curves as a reference for prewinter conditions inside the Antarctic vortex. The agreement between our results and values obtained by atmospheric models that take into account upper atmospheric sources of NOy suggests that such sources might be one of the causes of the extra NOy found in our prewinter estimates with respect to the extra-vortex ATMOS correlation.


[41] This research was supported by the National Science Foundation under grants OPP-9117813 and OPP-9705667, and by NASA under grant NAG 54071. G. Muscari was also supported by the Italian Space Agency (ASI), and S. Smyshlyaev also received support from the Russian Foundation for Basic Research (grant 00-05-64636) and the INTAS Association (project 01-0732). We thank Mark Schoeberl and his collaborators at the Goddard Space Flight Center for access to their trajectory model and Wei-Wu Tan for running it for us. We are grateful to three anonymous reviewers who provided us with useful comments and suggestions.