A detailed analysis of available in situ and remotely sensed N2O and CH4 data measured in the 1999/2000 winter Arctic vortex has been performed in order to quantify the temporal evolution of vortex descent. Differences in potential temperature (θ) among balloon and aircraft vertical profiles (an average of 19–23 K on a given N2O or CH4 isopleth) indicated significant vortex inhomogeneity in late fall as compared with late winter profiles. A composite fall vortex profile was constructed for 26 November 1999, whose error bars encompassed the observed variability. High-latitude extravortex profiles measured in different years and seasons revealed substantial variability in N2O and CH4 on θ surfaces, but all were clearly distinguishable from the first vortex profiles measured in late fall 1999. From these extravortex-vortex differences we inferred descent prior to 26 November: as much as 397 ± 15 K (1σ) at 30 ppbv N2O and 640 ppbv CH4, and falling to 28 ± 13 K above 200 ppbv N2O and 1280 ppbv CH4. Changes in θ were determined on five N2O and CH4 isopleths from 26 November through 12 March, and descent rates were calculated on each N2O isopleth for several time intervals. The maximum descent rates were seen between 26 November and 27 January: 0.82 ± 0.20 K/day averaged over 50–250 ppbv N2O. By late winter (26 February to 12 March), the average rate had decreased to 0.10 ± 0.25 K/day. Descent rates also decreased with increasing N2O; the winter average (26 November to 5 March) descent rate varied from 0.75 ± 0.10 K/day at 50 ppbv to 0.40 ± 0.11 K/day at 250 ppbv. Comparison of these results with observations and models of descent in prior years showed very good overall agreement. Two models of the 1999/2000 vortex descent, SLIMCAT and REPROBUS, despite θ offsets with respect to observed profiles of up to 20 K on most tracer isopleths, produced descent rates that agreed very favorably with the inferred rates from observation.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Each fall, decreasing polar sunlight triggers a series of events resulting in increased circumpolar winds, the formation of an isolated region of stratospheric air known as the polar vortex, and the subsidence or diabatic descent of the vortex air column by several kilometers resulting from radiative cooling [Schoeberl and Hartmann, 1991; McIntyre, 1992]. Quantifying vortex descent is necessary if one is to understand chemical processes, such as ozone loss, occurring within the vortex, as it is otherwise difficult to differentiate between chemical and dynamical changes in ozone.
 Satellite observations are extremely useful for this goal, as satellites offer near-continuous measurement and coverage over significant portions of Earth. However, they also suffer from some major drawbacks, such as low horizontal (>100 km) and vertical (∼2 km) resolution, and for all solar occultation-based instruments, limited coverage at high latitudes during the winter months. While the horizontal resolution is not particularly important for characterizing descent, which affects the vortex as a whole, the vertical resolution can be comparable to the amount of descent for short (less than ∼1 month) time periods. The inability to view polar latitudes during winter also hampers observations inside the vortex, which is typically located poleward of 60°N or south. In the Arctic, stratospheric conditions sometimes displace the vortex away from the pole during portions of the winter, allowing plentiful observations inside the vortex [e.g., Müller et al., 1997; Manney et al., 1999]. However, these conditions are not necessarily frequent in each winter, as was the case during much of 1999/2000.
 Field-based observations (consisting of aircraft- and balloon-based in situ and remote sensing measurements, as well as ground-based remote observations) offer the advantages over satellite observations of high spatial resolution, especially for in situ measurements, and greater vortex penetration in winter months. However, they also suffer from several limitations, namely a lack of temporal coverage (field campaigns are expensive, and weather limits both balloon and aircraft frequency), spatial coverage (there are a limited number of polar locations suitable for airborne research), and altitude (the ER-2 aircraft is limited to ∼20 km, and balloons are limited to ∼30 km). Thus there is no instrument/platform combination that provides ideal vortex coverage.
 In order to study descent within the Arctic vortex during the 1999/2000 winter season, we have combined tracer observations from ten different instruments deployed as part of the Stratospheric Aerosol and Gas Experiment III (SAGE III) Ozone Loss and Validation Experiment (SOLVE) and the Third European Stratospheric Experiment on Ozone 2000 (THESEO 2000) campaigns. These measurements represent a variety of techniques and covered large portions of the vortex from late November 1999 through mid-March 2000. In addition, measurements from three other instruments deployed in prior years were examined to help determine a vortex starting profile. Our inferred descent rates were compared with those from previous studies of Arctic descent. Last, our descent rates were compared with results from two stratospheric chemical transport models that simulated vortex descent for 1999/2000.
2. Data and Methods
 In this paper, we focus on the stratospheric tracers N2O and CH4, which were measured by a large number of instruments during the SOLVE and THESEO 2000 campaigns. These tracers have long photochemical lifetimes (years to decades) in the middle and lower stratosphere and thus were ideal for studies of polar descent within the stratosphere, where transport timescales are on the order of months to years [Brasseur and Solomon, 1995].
 Potential temperature (θ) was chosen as the vertical coordinate because, unlike altitude or pressure, it is conserved under adiabatic displacements. In the absence of mixing, changes in θ are due exclusively to diabatic heating, which is the effect we seek to quantify. To calculate θ, measurements of temperature and pressure are needed. In general, both these quantities were measured locally on the same platform as the tracer instrument; however, for some data products, assimilated pressure and temperature fields were used to generate θ.
 The in situ balloon instruments will be discussed first. The Lightweight Airborne Chromatograph Experiment (LACE) determined the volume mixing ratios (VMRs) of eight gases including N2O by gas chromatography (GC) with a time resolution of 70 s [Moore et al., 2002]. Pressure and temperature associated with the LACE flights were measured by the Jet Propulsion Laboratory (JPL) Ozone instrument [Salawitch et al., 2002] on 19 November 1999, and by the National Oceanic and Atmospheric Administration (NOAA) Frost-Point Hygrometer [Vömel et al., 1995] on 5 March 2000. The J. W. Goethe-Universität cryosampler (BONBON) [Schmidt et al., 1987], which was located at the Forschungszentrum Jülich prior to 1996, obtained up to 16 air samples per flight; VMRs were determined via GC in the laboratory. Pressure and temperature associated with the BONBON flights were measured by sensors located on the balloon payload, and operated by the Centre National d'Études Spatiales (CNES) and J. W. Goethe-Universität/Forschungszentrum Jülich teams, respectively.
 Data from the four in situ ER-2 instruments that measured N2O were combined into a unified N2O (uN2O) data product, following a procedure described by Hurst et al. . This procedure produced a self-consistent, high time resolution (3 s) N2O data set for each SOLVE flight, using an objective method to evaluate and reduce bias among the instruments, followed by weighted averaging. The Argus spectrometer [Jost and Loewenstein, 1999] measured the VMRs of N2O and CH4 using second harmonic detection of an infrared tunable diode laser (TDL) in a multipass absorption cell, with an average time resolution of 3.2 s. The Aircraft Laser Infrared Absorption Spectrometer (ALIAS) [Webster et al., 1994] is a TDL spectrometer that measured the VMRs of four gases, including N2O and CH4, with an average time resolution of 1.4 s. The Airborne Chromatograph for Atmospheric Trace Species IV (ACATS IV) [Elkins et al., 1996; Romashkin et al., 2001] determined the VMRs of eleven gases, including N2O and CH4, by GC with time resolution of 70 s. The Whole Air Sampler (WAS) [Schauffler et al., 1999] filled up to 32 canisters per flight with ambient air, which were analyzed in the laboratory for ∼27 trace gases including N2O and CH4 via GC. Pressure and temperature were measured on the ER-2 by the Meteorological Measurement System (MMS) instrument [Scott et al., 1990].
 Next we will discuss the remote sensing balloon instruments. The MkIV instrument is a Fourier Transform Infrared (FTIR) spectrometer, which measured solar occultation spectra over the wavelength range 1.77–15.4 μm. A large number of atmospheric constituents were measured, including N2O and CH4. Vertical resolution was 2 km, with profiles reported on a 1-km vertical grid [Toon, 1991]. Pressure and temperature profiles, respectively, were derived from spectral fits to temperature-insensitive and temperature-sensitive CO2 lines, assuming the CO2 profile measured by the Harvard CO2 instrument on board the Observations from the Middle Stratosphere (OMS) in situ gondola launched on 19 November 1999.
 The submillimeter-wave limb sounder (SLS) instrument recorded emission spectra of several gases; N2O was measured near 600 GHz. The data presented here were flight averages of one or more limb scan emission spectra. Profiles were obtained from limb scan spectra using an iterative least squares fitting procedure, with a nominal vertical resolution of 2.5 km. Temperature and pressure profiles were obtained from a balloon sonde launched on the day before the flight [Stachnik et al., 1992].
 On board the DC-8 aircraft was the airborne submillimeter radiometer (ASUR), which retrieved complete vertical profiles of several gases, including N2O, every ∼2000–6000 s from pressure-broadened emission lines at 652.833 GHz. The vertical resolution was 5–10 km, with profiles calculated on a 2-km vertical grid. Flight track data containing temperature and pressure for calculating θ were provided by the NASA Goddard Data Assimilation Office (DAO) [von König et al., 2000].
 Data from two spaceborne instruments were used in this paper. The Halogen Occultation Experiment (HALOE) instrument is located on the Upper Atmosphere Research Satellite (UARS), and uses solar occultation to measure vertical profiles of several gases, including CH4. The data used here were level 2, version 19. The instrument vertical resolution was 1.6 km at the Earth's limb, and the altitude range extended from 15 km to more than 60 km. Latitudinal coverage was from 80°S to 80°N over the course of one year. Because of the HALOE orbit and the Earth-Sun geometry during the Arctic winter months, the maximum northern coverage was limited to ∼50°N from November 1999 through January 2000, and up to ∼65°N in October 1999 and February–March 2000. As a result, data throughout most of the winter were limited to extravortex regions. Temperature and pressure were determined by spectral fitting of the 2.8 μm CO2 band [Russell et al., 1993b] above 35 km (10 mbar); below this altitude, they were obtained via interpolation from the National Meteorological Center (NMC) assimilated data product.
 The Atmospheric Trace Molecule Spectroscopy (ATMOS) instrument is an FTIR spectrometer similar to MkIV, which has flown on several Space Shuttle missions. ATMOS measured the abundances of ∼30 gases, including N2O and CH4, from 12–80 km altitude with a vertical resolution of 2–3 km. Version 3 data were used here. Northern latitude coverage ranged up to 69°N during the second Atmospheric Laboratory for Applications and Science (ATLAS-2) mission (8–16 April 1993), and up to 49°N during the ATLAS-3 mission (3–12 November 1994) [Gunson et al., 1996]. Pressure and temperature were determined by spectral fitting of the CO2 absorption bands [Stiller et al., 1995].
2.2. Data Selection
 Data were screened prior to use to determine their quality, based on the following criteria: calibration against known standards such as the NOAA Climate Monitoring and Diagnostics Laboratory (CMDL) reference gases, the existence of intercomparison with other instruments [e.g., Chang et al., 1996; Herman et al., 1998; Michelsen et al., 1999; Toon et al., 1999; Moore et al., 2002], consistency of the N2O:CH4 correlation with other instruments when both measurements were available for a single flight, and consistency of the profiles themselves with those of other instruments. Because data from dissimilar instruments were being compared, we have used total errors, consisting of precision and accuracy estimates, as well as uncertainties of calibration gases and/or spectroscopic parameters, in our analyses.
 Each data point of each flight was classified as vortex, extravortex, or mixed (and therefore excluded), using one of several categorization methods. For the balloon flights, each of which did not extend over large portions of the Arctic, hemispheric maps of modified Ertel's potential vorticity (PV) [Lait, 1994], provided by the NASA Goddard DAO, were examined on several θ surfaces to determine vortex classification. To define the vortex edge, the Nash criterion [Nash et al., 1996] was used for flights from 1999/2000 and the Argus OMS in situ flight of 30 June 1997; for the BONBON flight of 30 November 1991, the vortex edge was defined by the maximum PV gradient [Bauer et al., 1994].
 For the unified N2O data, flights were either classified as entirely extravortex from PV maps, or they were filtered to retain only vortex segments using the technique recently described by Greenblatt et al. . This technique used N2O versus θ profiles measured inside the vortex to accurately determine the inner edge of the vortex boundary region; the inner edge was used to remove both large-scale flight segments of extravortex air, as well as small filaments of mixed composition.
 For the ASUR data, PV was converted to equivalent latitude (EqL) [Butchart and Remsberg, 1986] and points between 70°N and 90°N EqL were assumed to be inside the vortex. A comparison of the different months during 1999/2000 when the ASUR instrument was deployed showed that the vortex edge was located between 50° N and 70° N EqL. Each flight used consisted of one or more profiles measured inside the vortex, from which a single weighted average profile was constructed.
 HALOE data were filtered using a rotation/strain diagnostic called Q, which was integrated along rotational streamlines, with the vortex edge being defined as the rotational streamline with integrated Q of zero [Fairlie et al., 1999; Pierce et al., 2002]. The data were further constrained to the high midlatitudes (>45°N). Due to the small amount of data available at these latitudes, monthly average extravortex profiles were used for this paper. PV and Q were obtained from the United Kingdom Meteorological Office (UKMO) UARS assimilated wind and temperature fields [Swinbank and O'Neill, 1994].
 The ATMOS tracer data were sorted according to isentropic maps of PV and VMRs of CH4, HNO3, O3, and N2O. Averages included profiles with similar tracer distributions throughout the altitude range considered within selected geographic regions. Profiles displaying inversions (e.g., regions where θ was not single-valued with respect to tracer VMR) were excluded from the averages. Classifications (vortex and midlatitude extravortex) were confirmed by comparison with tracer:tracer correlations [Michelsen et al., 1998a, 1998b].
2.3. Binning and Inversion
 Filtered data were binned into 5 K θ intervals in order to facilitate comparison among data sets. The tracer VMR reported for each θ bin was calculated from the weighted mean of all data within the bin. Gaps in the data were filled in using linear interpolation.
 For the purpose of comparing changes in θ on N2O or CH4 surfaces of constant VMR (isopleths), the data were “inverted” so that θ could be expressed as a function of tracer VMR. Inverted data were binned in intervals of 10 parts per billion by volume (ppbv) N2O and 40 ppbv CH4. The resulting tracer-binned data were calculated as follows: all data points whose VMR value ±1 standard deviation (σ) fell within a given tracer bin were considered in the calculation for that bin, and the minimum and maximum θ of all qualifying points were determined. The θ for the tracer-binned data point was defined as the midpoint of this range, and the 1σ uncertainty in θ was defined as half the range. Tracer-binned data were constructed from θ-binned data, rather than raw data, primarily to overcome data sparseness for some balloon flights. Gaps in the resulting data were filled in using linear interpolation.
Figure 1 illustrates this idea, using a portion of the LACE N2O profile of 19 November 1999 as an example. The region from 150–210 ppbv illustrates the need to consider the N2O uncertainties in the calculation of θ for a given tracer bin, because N2O error bars in this region were much larger than the 10-ppbv bin size, and therefore single data points contributed to the calculation of θ in several bins. Within the 230 and 240 ppbv N2O bins, where the θ-binned data display an inversion with respect to N2O, and at 300 ppbv and above, where the θ-binned data became essentially vertical, the midpoint of the range of θ values within a bin served as the most useful definition of θ. The ±1σ envelopes of the N2O-binned data (dashed line) generally encompassed those of the θ-binned data (dotted line), which indicates that this approach provides a conservative estimate of the uncertainty relative to θ. The overestimate of the N2O-binned envelope with respect to the θ-binned envelope between 250 and 260 ppbv N2O was due to the lack of any θ-binned data points in these two N2O bins; the interpolation scheme smoothly filled in this region from surrounding data.
 SLIMCAT is a three-dimensional chemical transport model using horizontal winds and temperatures from the UKMO-UARS assimilation [Swinbank and O'Neill, 1994] to drive transport on an isentropic (θ-based) vertical coordinate. Vertical motion is diagnosed by calculating heating rates from a radiation model. The abundances of chemical species in the stratosphere are calculated using a detailed chemistry scheme [Chipperfield, 1999]. The model simulates the distribution of all species involved in stratospheric ozone depletion, as well as some trace gases, such as N2O. The SLIMCAT model run started on 21 October 1991 and was integrated at low resolution (7.5° × 7.5° × 18 θ levels) until November 1999. Output from this run was then used to initialize a higher resolution simulation from 3 November 1999 to 31 March 2000, on 24 θ levels ranging from 335 to 2726 K. It was assumed that an EqL greater than 70°N was inside the vortex for the entire winter, and data between 70° and 80°N EqL were used to produce vortex averages over this time period. The N2O data were recorded daily, but for the current analysis the data were averaged over 10 day periods, and interpolated as described above in 5 K θ intervals. Data were also inverted into 10 ppbv N2O bins. No error bars were calculated for the original data, so all data points were treated equally, and uncertainties in N2O or θ represented the observed variation for a given θ or N2O bin, respectively.
 REPROBUS is a three-dimensional chemical transport model whose advection is driven by the 6-hourly European Centre for Medium-Range Weather Forecasting (ECMWF) meteorological analysis [Lefèvre et al., 1994, 1998]. The model includes a comprehensive treatment of gas-phase and heterogeneous stratospheric chemistry, as well as a semi-Lagrangian transport scheme integrated with 2° × 2° horizontal resolution, using all three components of the wind field. Calculations were performed on 42 vertical levels from the ground up to 0.1 hPa. N2O and CH4 tracer fields were initialized from zonal mean climatologies on 1 November 1999, and the model was run until 30 March 2000, recording N2O and CH4 once every 10 days. As with the SLIMCAT model, data between 70° and 80°N EqL were used to produce vortex averages, except for the two dates following the vortex break-up (20 and 30 March) when data were averaged between 80° and 90°N EqL, in order to obtain enough points within the vortex for good averaging statistics. Average tracer VMR values were calculated on 5 K θ intervals over the range 300–650 K. Data were inverted into 10 ppbv N2O and 40 ppbv CH4 intervals, the same as for the observational data. No error bars were originally calculated, so all data were treated equally, and uncertainties in θ represented the observed variation for a given N2O or CH4 bin.
3. Results and Discussion
 The 1999/2000 winter was characterized by unusually low stratospheric temperatures [Manney and Sabutis, 2000], which led to extensive PSC formation and subsequent ozone loss when sunlight returned to the vortex in mid-January [e.g., Rex et al., 2002]. By early November, the vortex was fairly well established with a strong polar jet above 800 K, but a weak jet at lower levels; by early December, the vortex was established down to 465 K [Manney and Sabutis, 2000]. This vortex persisted until mid-April [Waugh and Rong, 2002]. Planetary wave activity occurred throughout the winter season, with three pulses occurring during the measurement period: late November, mid-January through early February, and early March [Newman et al., 2002].
3.1. Data From SOLVE/THESEO 2000
Figure 2 shows the balloon-borne data from 1999/2000. The data can be roughly divided into two groups: late fall, consisting of the 19 November LACE and 3 December MkIV and SLS profiles, and mid to late winter, consisting of the 27 January and 1 March BONBON profiles and the 5 March LACE profile. The late fall profiles lie at higher θ levels than the winter profiles over almost the entire range of N2O and CH4 values; these differences are a result of descent. The late fall profiles also show a larger degree of variability than the later profiles, a point that will be addressed below. The March BONBON and LACE profiles, measured four days apart and launched from the same location (Esrange, Sweden), are in very good agreement with each other for both N2O and CH4, with the exception of a difference seen in N2O, which falls significantly outside the 1σ error bars beginning at 478 K. The cause of this discrepancy is not understood, but may be linked to a difference in the N2O:CH4 correlations of these profiles below 40 ppbv N2O (see below). The 27 January BONBON profile lies 9 ± 6 K (1σ) above the two March profiles over 40–280 ppbv N2O and 680–1600 ppbv CH4; the difference is attributed to late winter descent.
Figure 3 shows the unified N2O data measured on the ER-2 in two panels, with a balloon profile shown for reference on each panel. Data are shown after binning into 5 K θ intervals for clarity. The profiles are very tight over the entire time period, with the ±1σ of N2O within a given θ bin averaging 11.4 ppbv over the range 370–470 K, the approximate vortex altitude region sampled by the ER-2. A decrease in θ on a given N2O level can clearly be seen over the 52-day period, and often from flight to flight.
Figure 4 shows ASUR profiles in three panels. A reference in situ balloon profile is also shown for each panel to indicate how closely the ASUR data agree with in situ measurements made at comparable times. A systematic overestimation of N2O at low VMR is apparent, which is due to the inherent limited vertical resolution of ASUR (5–10 km) and the high curvature of the profile in that region. At higher N2O values (e.g., above ∼180 ppbv), where the change in slope is more gradual, the agreement with in situ measurements is much better. We first applied the ASUR averaging kernel to the balloon and unified N2O data, producing profiles with the same vertical resolution as ASUR. The resulting smoothed data showed much better agreement with ASUR at all N2O levels. Although using such an approach would allow comparison below 150 ppbv, it was desirable to retain the higher vertical resolution of the other instruments. Therefore, in the analyses which follow, no such averaging was applied to the data, and only the ASUR data on the 200 and 250 ppbv N2O isopleths were used.
3.2. N2O:CH4 Correlation
 The N2O:CH4 correlations for balloon data from 1999/2000 are presented in Figure 5. The five data sets show very good agreement with each other, except in the region between ∼10 and 40 ppbv N2O, where the LACE flight of 5 March differs significantly from the other data, including the two BONBON flights. The difference in the LACE correlation may indicate real differences in the N2O:CH4 correlation curve, because the MkIV profile of 15 March (not shown) closely follows the LACE correlation in this region. However, N2O values below 50 ppbv were not utilized in the work presented in this paper, so the disagreement does not impact our conclusions.
 Also shown in the figure are the ATMOS vortex and midlatitude correlations [Michelsen et al., 1998a], constructed from measurements made from 1992 to 1994. The data from 1999/2000 fall in between the two ATMOS correlations between ∼30 and 150 ppbv N2O, and fall at higher CH4 values above ∼170 ppbv N2O. The secular trends in N2O and CH4 [Intergovernmental Panel on Climate Change (IPCC), 1996] are responsible for most of the differences at high N2O (i.e., lower altitudes), but at lower N2O values the influence of tropospheric trends is expected to be smaller because of the shorter photochemical lifetimes N2O and CH4. The lack of sensitivity to such trends at these altitudes is supported by the excellent agreement between ATMOS correlations from 1994 and MkIV and ALIAS II correlations from 1997 [Herman et al., 1998]. The differences shown in Figure 5 are consistent with differences between the ATMOS early fall developing vortex (or protovortex) and extravortex correlations reported by Manney et al. . Such a difference can be attributed to some descent followed by mixing of extravortex air into the vortex, which is expected during this time of year [Manney et al., 2000; Plumb et al., 2000]. The differences, however, are not as substantial as those between the extravortex correlation and the correlation observed in the spring vortex in 1993 (the ATMOS vortex curve in Figure 5). These differences have similarly been attributed to significant unmixed descent followed by mixing with extravortex air [Michelsen et al., 1998a, 1998b, 1999].
 For the purposes of this paper, it was not necessary to understand why the N2O:CH4 correlation for the 1999/2000 winter differed from the ATMOS correlations, because the correlation was only needed to compare descent on N2O and CH4 isopleths. Thus a piecewise, weighted least squares fit was generated, based on the five balloon flights, between 700 and 1800 ppbv CH4 (44–323 ppbv N2O). This fit is shown in Figure 5, with its functional form listed in the caption. The fit was used to generate the CH4 isopleths corresponding to the N2O isopleths displayed in Figure 6.
3.3. Late Fall Vortex Inhomogeneity
 Hemispheric PV maps at 450 K show low gradients through early December 1999 [see, e.g., Newman et al., 2002, Figure 11], suggesting some mixing of vortex and midlatitude air in the lower stratosphere. It is clear from examination of the three fall balloon profiles (Figure 2) and the fall ASUR data (Figure 4a) that the vortex tracer profiles were quite variable in late fall. For instance, the mean MkIV-LACE difference was 23 ± 3 K between 110 and 200 ppbv N2O, and 20 ± 4 K between 960 and 1200 ppbv CH4. The mean SLS-LACE difference was about half as large, 11 ± 1 K between 110 and 150 ppbv N2O. As MkIV and SLS were flown on the same balloon payload, the differences in their measurements were likely due to the fact that the instruments observed air masses in different directions, corresponding to a horizontal separation of ∼800 km.
 The ASUR instrument sampled large regions of the vortex between 30 November and 16 December. For this portion of the year, the data reveal a variability, indicated by ±1σ range in θ, which was similar to the MkIV-LACE differences: 19 K at 200 ppbv N2O and 25 K at 250 ppbv. For late winter ASUR profiles (27 February to 15 March), the variability dropped to 17 K at 200 ppbv and 11 K at 250 ppbv (excluding the flight of 8 March, which exhibited unusually high N2O deviations; see magenta curve in Figure 4c). By comparison, the unified N2O variability in late winter (26 February to 12 March) was 7.4 K at 200 ppbv and 9.8 K at 250 ppbv. Thus, though the noise in the ASUR data is certainly greater than for unified N2O, particularly at 200 ppbv N2O, the large decrease in the ASUR variability at 250 ppbv is indicative of a more homogenous vortex in late winter than late fall.
 Further evidence for an inhomogenous vortex is given in a study by Ray et al. . They found that they could explain the differences in tracer-tracer relationships between the November and March LACE profiles with a simple differential diabatic descent model. In this model, air parcels underwent different amounts of descent prior to the 19 November measurement, resulting in air with significantly different tracer VMRs on the same θ surfaces, which subsequently mixed [see, e.g., Plumb et al., 2000]. Their model simultaneously fit six measured stratospheric tracers, including N2O and CH4, producing a self-consistent result with differential diabatic descent ranging from ∼2 km at 15 km to ∼6 km above 30 km. Thus their analysis indicated that the vortex was inhomogenous in the fall but became well mixed later in the winter.
Manney et al.  cited evidence for vortex inhomogeneity even earlier in the winter of 1994/1995. From 3–12 November 1994, the ATMOS instrument observed several laminae within the developing polar vortex whose N2O:CH4 correlations indicated both entrainment of tropical air, as well as midlatitude air that had undergone significant descent. The tracer versus θ profiles indicated large deviations from the bulk correlation in these regions, resulting in a vortex with substantial inhomogeneity.
 All of the above indicate that the vortex tends to be fairly inhomogenous in late fall. By early November, the vortex is typically well established only above 600 K [Waugh and Randel, 1999], and in 1999 had only recently formed in the lower stratosphere [Manney and Sabutis, 2000]. Thus it is a reasonable conclusion that complete mixing had not yet occurred. The smaller variability in θ for a given tracer isopleth seen later in the winter supports the idea of a more uniformly mixed vortex.
3.4. Late Fall Average Profile
 For the purpose of estimating an average late fall vortex profile with realistic uncertainties, we have combined the LACE and MkIV balloon profiles in the following manner. (The SLS profile was not included because of its large error bars at low and high N2O levels. However, the SLS measurements were bounded by the other two profiles, so their omission did not detract from the goal of creating a profile that spanned the range of observations. The ASUR data were not included due to their low vertical resolution). It was not appropriate to calculate error-weighted averages because the profiles did not look at the same air parcels. Thus, for each θ bin, the minimum and maximum tracer VMR ±1σ errors were determined; the “average” VMR was defined as the midpoint of that range, and the 1σ uncertainty was defined as half the range. The procedure was applied independently to the N2O and CH4 data. The date assigned to each profile was the average date of the two data sets, 26 November 1999. The profiles are shown in Figure 7.
3.5. Time Series
 The balloon, unified N2O and ASUR data have been combined and presented as a time series along several N2O and CH4 isopleths in Figure 6. Data shown here have been inverted as described in the previous section to generate θ as a function of N2O or CH4. The levels of 50, 100, 150, 200, and 250 ppbv N2O were chosen to span the useful range of the measurements: at levels below 50 ppbv, there is almost no unified N2O data, and the balloon profiles become quite steep in θ, making comparisons along isopleths difficult; much above 250 ppbv, the curves cluster on top of one another at all time periods, indicating marginal descent and probably significant mixing with midlatitude air on these isopleths [e.g., Abrams et al., 1996a]. CH4 levels of 720, 920, 1080, 1280, and 1480 ppbv were chosen to correspond approximately (to within one 40-ppbv CH4 bin) to these N2O levels, based on the average N2O:CH4 correlation derived from the balloon measurements, shown in Figure 5. Simple fits through the data (linear least squares fits in dense data regions (November–December, January–February, and February–March deployments), and connected points in between these regions) have been applied to the N2O data to indicate the general progress of the descent throughout the winter. These same fits have been reproduced on the CH4 panel for reference. The more limited CH4 data are very consistent with these fits throughout the winter.
 The progress of descent is evident in detail in Figure 6. The rates of descent generally increased with decreasing tracer VMR, as indicated by the slope of the fits, and the greatest rates of descent occurred before 1 January. Between the midwinter (January–February) and late winter (February–March) data segments, the descent rates slowed considerably. Data from all instruments were in excellent agreement with each other throughout the winter, with the exception of the LACE and MkIV late fall data, which lay outside each other's error bars on the 150 ppbv N2O and 1080 ppbv CH4 isopleths because of the vortex inhomogeneity discussed earlier.
3.6. Vortex Starting Profile
 Many studies of vortex descent have calculated the total descent over the winter season starting from profiles measured in late fall [e.g., Bauer et al., 1994; Strahan et al., 1994; Müller et al., 1996; Hartmann et al., 1997]. However, it is known that descent can begin quite early in the winter season. For instance, Michelsen et al. [1998b] and Manney et al.  showed that profiles measured inside the Arctic vortex during 3–12 November 1994 had already descended several km relative to extravortex profiles. Abrams et al. [1996a] calculated from analysis of PV contours that descent had begun in the fall of 1992 in the Arctic between 1 October and 12 November, with the earliest dates occurring at the highest altitudes (>42 km). Rosenfield et al.  chose a starting date of 1 November for model simulations of Arctic vortex descent in 1988/1989 and 1991/1992, with significant descent rates calculated from the first day of the simulations.
 By 19 November 1999, the date of the first LACE profile, significant descent had occurred at all altitudes relative to extravortex profiles, as will be demonstrated shortly. Thus using this profile (or the 26 November 1999 average fall profile) as the starting point for calculating descent would miss a significant amount of descent that occurred earlier in the fall. In order to determine the total vortex descent in 1999/2000, a reasonable starting profile had to be established. Prior to vortex formation in late fall, air in polar regions is constantly mixed over horizontal distances of several thousand km [Schoeberl and Hartmann, 1991]. Thus the tracer versus θ profile of polar air before vortex formation strongly resembles air of midlatitude composition. This has been confirmed observationally by a number of studies [Bauer et al., 1994; Abrams et al., 1996a; Herman et al., 1998; Jost et al., 1998; Michelsen et al., 1998a]. It has been recognized that the tracer versus θ profiles exhibit small seasonal changes [Abrams et al., 1996a], and that tropical profiles appear significantly different due to their partial isolation from the midlatitudes [e.g., Jost et al., 1998; Herman et al., 1998; Michelsen et al., 1998a]. However, by limiting our examination to high-latitude, extravortex regions, we have assembled a selection of candidate vortex starting profiles.
 These profiles are shown in Figure 7, along with the 26 November 1999 average vortex profile for reference (solid black line). The extravortex profile from the BONBON instrument, shown as a solid red line, was launched from Kiruna, Sweden on 30 November 1991 [Bauer et al., 1994]. Averaged extravortex profiles from the April 1993 and November 1994 ATMOS missions are shown as orange and green solid lines, respectively (H. A. Michelsen, private communication, 2001). The Argus profile, shown as a solid cyan line (for N2O only), was measured in Fairbanks, Alaska on 30 June 1997 [Jost et al., 1998]. It exhibited several large excursions to low N2O that have been interpreted as vortex remnant filaments, and were therefore omitted from Figure 7. The limited extravortex data from 1999/2000 included a unified N2O profile from 16 March (dark blue solid line, N2O only), and monthly averaged HALOE profiles from December, January, and March poleward of 45°N (magenta, black, and red, respectively, with dotted lines; CH4 only).
 While there is variation among profiles of up to ∼70 ppbv N2O and ∼400 ppbv CH4 on certain θ levels, the overall pattern is consistent through multiple years and seasons, and is, most importantly, quite distinct from the 26 November 1999 average vortex profile. Therefore, we assert that any profile from the set presented here can be used as a proxy for the vortex starting condition for 1999/2000. The November 1994 ATMOS average was chosen, which when both N2O and CH4 were considered, occupied a roughly median value at all θ values. We denoted this profile as the “START” profile, and used it in all subsequent calculations. Note that the uncertainty associated with this profile does not represent the full range of year-to-year variations in starting profiles shown in Figure 7. Extending the uncertainties in θ to include these variations would render the starting profile useless, however, because the error bars would be larger than the differences in θ on most N2O isopleths. Thus we have chosen to retain the original uncertainties, with the caveat that the actual change in θ for 1999/2000 may lie outside the stated error bars.
3.7. Descent Calculations
 In order to determine the amount of descent between measurements made at different times, differences in θ between profiles have been calculated, with the results shown in Figure 8. Vertical bars indicate ±1σ uncertainties, which were calculated as the quadrature sum of individual profile uncertainties. The diamonds/solid lines represent differences (Δθ, expressed in K) between the START profile and the 26 November 1999 average; squares/dotted lines, between the 26 November average and the 27 January 2000 BONBON profile; and triangles/dashed lines, between the BONBON profile and the 5 March 2000 LACE profile. Differences are shown for both N2O and CH4 data in each case.
 For the START to 26 November difference, Δθ ranged from 397 ± 15 K (1σ) at the lowest VMR values (30 ppbv N2O and 640 ppbv CH4), to 28 ± 13 K above 200 ppbv N2O and 1280 ppbv CH4. These differences were larger than those observed in any other profile pair, and indicated that a great amount of descent had occurred by late fall 1999. Such large descent was also inferred for early November 1994, when the ATMOS instrument examined air both inside and outside the forming polar vortex [Michelsen et al., 1998b]. While that study expressed descent in terms of changes in altitude only, differences in terms of θ have been calculated and found to be comparable to our results: 220–240 K at 30 ppbv N2O and 640 ppbv CH4, and 380 K or more below 10 ppbv N2O and 390 ppbv CH4 (H. A. Michelsen, private communication, 2001).
 The 26 November to 27 January differences were much smaller than the START 26 November differences below 170 ppbv N2O and 1160 ppbv CH4, but were fairly comparable above these isopleths. From 40–280 ppbv N2O and 680–1600 ppbv CH4, which covered most of the tracers' dynamic ranges, Δθ was fairly constant at 54 ± 10 K. Between 27 January and 5 March, Δθ was smaller still, only 10 ± 6 K over the same ranges. The differences between the START profile and the 5 March LACE profile (not shown), as an indication of the total descent during 1999/2000, ranged from 501 ± 21 K at 30 ppbv N2O and 640 ppbv CH4 to 49 ± 23 K above 200 ppbv N2O and 1280 ppbv CH4.
 It is important to emphasize that these results probably represent underestimates of the amount of descent, because the effects of mixing have not been accounted for. “Mixing” here indicates the process of vortex homogenization, incorporating air that has either become entrained from midlatitudes, or undergone differential diabatic descent within the vortex. In either situation, adjacent air masses will have undergone different amounts of descent, and tend to mix along θ surfaces. Both of these processes can be active during vortex formation in the fall, and can substantially alter tracer:tracer as well as tracer:θ relationships [Plumb et al., 2000]. The general effect is to reduce the apparent descent [Abrams et al., 1996b; Müller et al., 1996]. However, once the vortex has become isolated from the rest of the stratosphere, which is typically complete by late fall, the assumption of no mixing is a fairly good one. For instance, in the study by Ray et al. , which explicitly modeled mixing as a result of differential diabatic descent in the early fall, only small changes (<20 K) in the calculated descent between the 19 November 1999 and 5 March 2000 LACE measurements were seen relative to the simple case where mixing was ignored.
 Descent rates are shown in Figures 9a–9e, as well as in Table 2. Rates are shown for all profile pairs shown in Figure 8, with the exception of the START to 26 November difference, since the effective 1999 date of the START profile was not known. Rates for several unified N2O profile pairs are also shown. Although the unified N2O-based rates display generally higher error bars than those based on the balloon data, the latter were included because their higher time density enabled a more detailed temporal reconstruction of the descent rates. There is a general decrease in the descent rate with time for all N2O isopleths shown, ranging from between 0.56 ± 0.18 and 1.05 ± 0.21 K/day in the earliest time interval (26 November to 27 January), to between −0.17 ± 0.37 and 0.33 ± 0.47 K/day in the latest interval (26 February to 12 March). This is entirely expected, and is a reflection of increasing solar radiation slowly reducing the net radiative cooling as winter passes into spring. There is also a general decrease in the rate with increasing N2O isopleth: e.g., for the time interval 27 January to 5 March, the rate decreases from 0.39 ± 0.15 K/day at 50 ppbv to 0.13 ± 0.19 K/day at 250 ppbv. This N2O trend reflects the greater radiative cooling at higher altitudes, while the much lower descent by 250 ppbv is indicative of the increased mixing near the tropopause. However, a notable exception is the 200 ppbv isopleth, which has the highest descent rate in the earliest interval.
Table 2. Calculated Descent Rates (K/day) for the 1999/2000 Wintera
Range of Dates
N2O Isopleth, ppbv
Uncertainties are 1σ.
26 Nov. to 27 Jan.
0.97 ± 0.18
0.81 ± 0.21
0.69 ± 0.25
1.05 ± 0.21
0.56 ± 0.18
0.82 ± 0.20
27 Jan. to 5 March
0.39 ± 0.15
0.33 ± 0.09
0.33 ± 0.15
0.07 ± 0.09
0.13 ± 0.19
0.24 ± 0.14
Unified N2O Measurements
23 Jan. to 2 Feb.
0.75 ± 0.35
0.00 ± 0.35
0.50 ± 0.35
0.25 ± 0.79
0.41 ± 0.34
2 to 26 Feb.
0.31 ± 0.15
0.52 ± 0.23
0.31 ± 0.23
0.21 ± 0.44
0.35 ± 0.16
26 Feb. to 12 March
0.17 ± 0.24
0.33 ± 0.47
−0.17 ± 0.37
0.00 ± 0.53
0.10 ± 0.25
26 Nov. to 5 March
0.75 ± 0.10
0.63 ± 0.13
0.55 ± 0.16
0.68 ± 0.13
0.40 ± 0.11
0.61 ± 0.13
 Weighted average descent rates across N2O isopleths for each time interval are also shown in Figure 9f, as well as in Table 2, decreasing from 0.82 ± 0.20 K/day between 26 November and 27 January to 0.10 ± 0.25 26 February to 12 March. Finally, a winter average descent rate (between 26 November and 5 March; results for 26 November to 12 March are very similar) is shown in Table 2 for each N2O isopleth, as well as a grand average. A decrease from 0.75 ± 0.10 K/day at 50 ppbv to 0.40 ± 0.11 K/day at 250 ppbv is observed, with the average winter 1999/2000 descent rate across 50–250 ppbv N2O at 0.61 ± 0.13 K/day.
Rosenfield et al. , who simulated both the 1988/1989 and 1991/1992 winters, reported average rates for 22 January to 21 March (0.60–0.95 K/day) which were significantly higher than the similar time period 27 January to 5 March in our study (0.07 ± 0.09 to 0.39 ± 0.15 K/day). While the Rosenfield et al.  results were similar in magnitude to the 1991/1992 study of Lucic et al.  (0.55–0.9 K/day), that study began earlier in the year (7 December), where larger descent rates would be expected. The only other 1988/1989 study included here which calculated descent rates after 1 January was that of Schoeberl et al. , and those descent rates were much lower (0.09 ± 0.16 to 0.36 ± 0.05 K/day).
Bauer et al.  reported observation-based descent rates from 31 October 1991 to 18 January 1992 of 2.7–3.4 K/day at 50–100 ppbv N2O, far higher than those calculated for the similar period 26 November to 27 January in our study (0.81 ± 0.21 to 0.97 ± 0.18 K/day). The difference may be partly due to the earlier starting time of the calculation; Strahan et al.  reported a model descent rate for October 1991 of 1.2 K/day at 250 ppbv N2O; although not reported in the study, the rate at 50–100 ppbv would probably be much larger, consistent with the pattern of larger rates on lower N2O isopleths. However, Rosenfield et al. , who began their simulation on 1 November, almost coincident with the first Bauer et al.  profile, reported substantially smaller rates though 22 January (1.02–1.20 K/day at 50–100 ppbv N2O), consistent with our results. It is possible that the Bauer et al.  study, which utilized data from one launch location, sampled anomalously large local descent, which was not observable in the vortex average study of Rosenfield et al. .
 It was hoped that by examining studies that reported descent rates in terms of both θ and z, a simple correspondence could be derived between the two sets of rates. However, the data showed little or no relationship between the two coordinate systems; for instance, Rosenfield et al.  reported a substantial increase in the z-based descent rates after 22 January, whereas the θ-based rates in the same study decreased markedly. Another example is Bauer et al. , whose z-based rates, while on the high end of the range reported by other studies, did not stand out nearly as much as the θ-based rates discussed earlier (see Table 3). Thus the relationship between θ and z appears complex, and we did not attempt to make any comparison between studies that reported descent rates in terms of z and the θ-based 1999/2000 descent rates.
3.9. Model Comparisons
Figure 10 shows the comparison between the REPROBUS model and fits of the observational data. The agreement differed considerably from N2O to CH4, with each tracer exhibiting better agreement on some isopleths and worse agreement on others. In terms of absolute differences, the N2O model data agreed most favorably with the observational fits from 150–250 ppbv, while the CH4 data agreed quite well for the 720 ppbv isopleth (equivalent to 50 ppbv N2O), but had worse agreement for all other isopleths. Most of the differences observed may be due to poor model tracer initializations and/or inadequate model tracer photochemistry. However, while the differences between model and observation may be considerable for several isopleths, they did not change significantly with time (with the exception of the 50 ppbv N2O and 1480 ppbv CH4 isopleths, whose model observation differences increased markedly with time). This relative agreement indicates that the model-derived descent rates were, in fact, fairly accurate throughout the winter. The disagreement at 50 ppbv N2O may indicate that descent was not strong enough in the model at these altitude levels, while the disagreement at 1480 ppbv CH4 was probably due to the close proximity to the bottom of the vortex, which is typically located near 400 K [e.g., Manney et al., 1994].
 A comparison of the SLIMCAT N2O results and fits of the observational data is shown in Figure 11. Offsets between model and observation were of the same magnitude (∼20 K) as the REPROBUS differences for the 100–250 ppbv isopleths, and the 50 ppbv isopleth agreed much better with the observed data than did the 50 ppbv N2O REPROBUS isopleth. There was a slight change in the offsets between 100 and 200 ppbv around day 27, resulting in better agreement between model and observation, while the 250 ppbv isopleth offset increased slightly over time. The 50 ppbv isopleth, despite its larger error bars, appeared to track the observed data best, with the exception of days 37 and 87, where error bars were much higher; these increases were associated with a more distorted vortex, resulting in a significantly wider range of VMRs sampled in the 70°–80°N EqL region. Despite these θ offsets, the good agreement in slopes between the SLIMCAT results and the observations indicated that the descent rates calculated were very similar to those derived from our study.
 The descent of the 1999/2000 Arctic vortex has been quantified using N2O and CH4 tracer measurements from a number of instruments. Coverage spanned the last stages of vortex formation in late November 1999 through vortex breakup in mid-March 2000, with two large gaps (17 December to 13 January and 4–25 February). Changes in θ were determined on five N2O and CH4 isopleths, and descent rates were calculated on each N2O isopleth for several time intervals throughout the fall and winter. From 26 November to 5 March the vortex descended 65 ± 12 K when averaged over 40–280 ppbv N2O and 680–1600 ppbv CH4. The maximum descent rates occurred in the late fall/early winter phase: between 26 November and 27 January, the calculated rate was 0.82 ± 0.20 K/day averaged over 50–250 ppbv N2O. By late winter (26 February to 12 March), the average rate had decreased to 0.10 ± 0.25 K/day. Descent rates also decreased with increasing N2O isopleth; for instance, the winter average (26 November to 5 March) descent rate varied from 0.75 ± 0.10 K/day at 50 ppbv to 0.40 ± 0.11 K/day at 250 ppbv. The winter average rate over 50–250 ppbv N2O was 0.61 ± 0.13 K/day.
 Larger variability in θ on tracer isopleths in late fall versus winter indicated that the vortex was much less homogenous in the late fall. The degree of inhomogeneity, derived from several instruments (LACE, MkIV, SLS, and ASUR), was 19–23 K over the range 110–200 ppbv N2O and 960–1200 ppbv CH4. This inhomogeneity decreased significantly in the March data sets; the unified N2O variability averaged only 7.4 K at 200 ppbv N2O. Because of the vortex inhomogeneity in late fall, an average fall vortex profile was constructed from the 19 November LACE and 3 December MkIV measurements, which encompassed the observed variability of all the fall measurements. This profile, referred to as the 26 November 1999 vortex average, was used in subsequent calculations of vortex descent.
 A comparison of high-latitude, extravortex profiles measured in different years and seasons, while revealing significant VMR differences on a given θ level, also showed that extravortex profiles were more similar to each other than to the late fall profiles. Therefore, the median profile, an average extravortex profile measured by ATMOS in November 1994, was chosen as the vortex starting profile for 1999/2000 (denoted “START”) and was used to estimate the amount of descent prior to the first vortex measurements in November and December. The difference between the START profile and the 26 November average profile showed significant descent prior to the first profiles measured in fall 1999/2000, up to 397 ± 15 K at the lowest N2O and CH4 levels (30 ppbv N2O and 640 ppbv CH4), and falling to 28 ± 13 K for N2O larger than 200 ppbv and CH4 larger than 1280 ppbv. Such early descent has been inferred for prior years in the Arctic, e.g., 1988/1989 and 1991/1992 [Rosenfield et al., 1994], 1993/1994 [Abrams et al., 1996a] and 1994/1995 [Michelsen et al., 1998b]. While these differences did not include the effects of mixing, as first-order approximations they indicated that a great deal of descent had occurred prior to the late fall vortex measurements.
 Descent rates from previous θ-based studies corroborate the results of our study very well across a wide range of time periods and N2O isopleths, with the exception of one observational study [Bauer et al., 1994] and the late winter time period of one model study [Rosenfield et al., 1994]. Because it has been shown, both with our own results and those of prior studies, that descent rates increase substantially with decreasing N2O or increasing altitude, it is more meaningful and useful to calculate rates for several tracer isopleths (as we have done), θ levels, or z levels, than to report a single number. Likewise, because of the large change in descent rates observed throughout the fall and winter, a single, season average descent rate is not as informative as rates reported for several selected times.
 Both the REPROBUS and SLIMCAT models calculated descent rates that agreed quite favorably with observations. Although both models showed disagreements with observations of up to ∼20 K (except for the 50 ppbv N2O isopleth of REPROBUS), the causes were somewhat different for each model. For REPROBUS, the quality of the tracer initializations directly affected the agreement, as there was no “spin-up” period in the model. For SLIMCAT, where tracer distributions were initiated in 1991, the differences with observation were due almost exclusively to model chemistry and transport, which over the eight year run time of the model should be regarded as very good. Given the different approaches employed by the two models for deriving these rates (direct use of the vertical wind field component for REPROBUS versus a radiative model for SLIMCAT), the success of both models in reproducing the observed descent via trace gas measurements should be highlighted.
 We thank James J. Margitan (Jet Propulsion Laboratory) for providing potential temperature data from his ozone instrument on the 19 November 1999 balloon flight, Paul A. Newman and Leslie R. Lait (NASA Goddard) for providing Nash algorithm analysis of PV data for 1999/2000, Ingeborg Levin (Institut für Umweltphysik, Heidelberg, Germany) for performing the CH4 analysis for the BONBON flight of 30 November 1991, F. William Irion and Michael R. Gunson (Jet Propulsion Laboratory) for producing the ATMOS version 3 data sets used in this paper, and Hope A. Michelsen (Sandia National Laboratories) for providing averaged extravortex ATMOS data from 1993 and 1994. We are indebted to Susan E. Strahan (NASA Goddard) and Rolf Müller (Forschungszentrum Jülich) for helpful discussions and to Gloria L. Manney (Jet Propulsion Laboratory/New Mexico Highlands University), Hope A. Michelsen, Leonhard Pfister (NASA Ames), and Henry (Rennie) Selkirk (NASA Ames) for commentary on the manuscript. Thanks go to Jeff Grose (NASA Ames) for his expertise in all aspects of the Argus instrument operations, the crews and pilots of the ER-2 and DC-8 aircraft (NASA Dryden), the Centre National d'Études Spatiales (CNES), and the National Scientific Balloon Facility (NSBF) teams for flight operations during the difficult conditions of the Arctic winter. We thank Michael Kurylo and Phil DeCola (NASA Headquarters) for funding SOLVE investigators and the European Commission for funding THESEO 2000 investigators. We thank the Forschungszentrum Jülich for funding the BONBON balloon flight on 1 March 2000. M. Chipperfield and B.-M. Sinnhuber are grateful to the UK Natural Environment Research Council's Upper Troposphere/Lower Stratosphere (UTLS) Programme, which provided funding for their work. J. Greenblatt is grateful for funds provided by the National Research Council, which made his research at NASA Ames possible.