Middle atmospheric O 3, CO, N 2 O, HNO 3, and temperature profiles during the warm Arctic winter

Transcription

1 Click Here for Full Article JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112,, doi: /2006jd007849, 2007 Middle atmospheric O 3, CO, N 2 O, HNO 3, and temperature profiles during the warm Arctic winter Giovanni Muscari, 1,2 Alcide G. di Sarra, 1,3 Robert L. de Zafra, 4 Francesco Lucci, 1 Fabrizio Baordo, 1 Federico Angelini, 1,5 and Giorgio Fiocco 1 Received 28 July 2006; revised 27 March 2007; accepted 7 May 2007; published 19 July [1] Ground-based measurements of stratospheric constituents were carried out from Thule Air Base, Greenland (76.5 N, 68.7 W), during the winters of and , involving operation of a millimeter-wave spectrometer (GBMS) and a lidar system. This work focuses on the GBMS retrievals of stratospheric O 3, CO, N 2 O, and HNO 3, and on lidar stratospheric temperature data obtained during the first of the two winter campaigns, from mid-january to early March For the Arctic lower stratosphere, the winter is one of the warmest winters on record. During a large fraction of the winter, the vortex was weakened by the influence of the Aleutian high, with low ozone concentrations and high temperatures observed by GBMS and lidar above 27 km during the second half of February and in early March. At 900 K (32 km altitude), the low ozone concentrations observed by GBMS in the Aleutian high are shown to be well correlated to low solar exposure. Throughout the winter, PSCs were rarely observed by POAM III, and the last detection was recorded on 17 January. During the lidar and GBMS observing period that followed, stratospheric temperatures remained above the threshold for PSCs formation throughout the vortex. Nonetheless, using correlations between GBMS O 3 and N 2 O mixing ratios, in early February a large ozone deficiency owing to local ozone loss is noted inside the vortex. GBMS O 3 -N 2 O correlations suggest that isentropic transport brought a O 3 deficit also to regions near the vortex edge, where transport most likely mimicked local ozone loss. Citation: Muscari, G., A. G. di Sarra, R. L. de Zafra, F. Lucci, F. Baordo, F. Angelini, and G. Fiocco (2007), Middle atmospheric O 3, CO, N 2 O, HNO 3, and temperature profiles during the warm Arctic winter , J. Geophys. Res., 112,, doi: /2006jd Dipartimento di Fisica, Università degli studi di Roma La Sapienza, Rome, Italy. 2 Now at Istituto Nazionale di Geofisica e Vulcanologia, Rome, Italy. 3 Also at Divisione Ambiente Globale e Clima, Ente per le Nuove Tecnologie, l Energia e l Ambiente, Santa Maria di Galeria, Italy. 4 Department of Physics and Astronomy, and Institute for Terrestrial and Planetary Atmospheres, Stony Brook University, State University of New York, Stony Brook, New York, USA. 5 Now at Istituto di Scienze dell Atmosfera e del Clima, Consiglio Nazionale delle Ricerche, Rome, Italy. Copyright 2007 by the American Geophysical Union /07/2006JD007849$ Introduction [2] During the period January March of 2002 and January February of 2003, a joint effort of the University of Rome La Sapienza and the State University of New York at Stony Brook, with the logistic support of the Danish Meteorological Institute and the U.S. National Science Foundation, carried out ground-based measurements of stratospheric constituents from the Network for the Detection of Atmospheric Composition Change (NDACC, previously known as NDSC) site at Thule Air Base, Greenland (76.5 N, 68.7 W). This joint research work involved the operation of a millimeter-wave spectrometer (GBMS) [de Zafra, 1995], providing stratospheric profiles of HNO 3,N 2 O, and HCN and strato-mesospheric profiles of O 3 and CO [de Zafra and Muscari, 2004, and references therein], and a lidar system [Fiocco et al., 1996] for detecting Polar Stratospheric Clouds (PSCs) and measuring temperature and pressure profiles in the stratosphere [di Sarra et al., 2002, and references therein]. This work focuses on the GBMS retrievals of O 3, CO, N 2 O, and HNO 3, and on lidar temperature data obtained during the winter campaign of [3] For the Arctic lower stratosphere, the winter of was one of the warmest winters on record for the past 27 years (e.g., see Manney et al. [2005] and NOAA winter bulletins available online at gov/products/stratosphere/winter_bulletins/index.html). Temperatures observed in the Arctic lower stratosphere were below PSC formation thresholds only for very few days and in very limited areas, well below the average duration in time and area of coverage observed during the period 1978/ /2004 [Manney et al., 2005]. [4] This makes winter a rare opportunity for addressing important unresolved issues concerning the winter Arctic stratosphere. One of the challenges is to separate the impact that dynamical versus chemical pro- 1of14

2 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER cesses have on ozone concentrations inside the polar vortex [e.g., Manney et al., 2006], and moreover, not just in the lower stratosphere where ozone is most subject to chlorine photochemical depletion. In particular, warm winters can help test our ability to quantify ozone loss due to heterogeneous activation versus changes in ozone concentration owing to dynamical processes or homogeneous chemistry, factors affecting the whole O 3 column but more difficult to evaluate during years of a strong, cold vortex accompanied by large-scale PSC activity. A thorough understanding of both strong and weak vortex phenomena is critical when it comes to identifying the impact of the predicted long-term cooling of the Arctic lower stratosphere [World Meteorological Organization, 2006, and references therein]. [5] The main dynamical processes affecting ozone concentrations inside the Arctic vortex are descent of air from the upper stratosphere and intrusion of midlatitude air into the vortex. Variations in concentration due instead to chemistry can originate from the heterogeneous activation of chlorine and bromine compounds, which deplete ozone as the sun rises over polar regions (or as air moves into sunlit regions within a displaced vortex), and/or the shift in the photochemical equilibrium driven by the confinement of air in the polar winter darkness [e.g., Solomon, 1999]. [6] In order to evaluate the impact of these different processes on O 3 concentrations inside the polar vortex, correlations of ozone with long-lived compounds ( tracers ) have often been employed in the literature [e.g., Proffitt et al., 1990; Michelsen et al., 1998]. However, there are different views on how reliable these correlations can be for quantifying ozone loss inside the polar vortex, and how mixing across the vortex edge can be taken into account (for contrary conclusions concerning the effects of crossboundary transport see, e.g., Plumb et al. [2000] and Müller et al. [2005]). In particular, some of the critical issues involve the compactness of correlation curves, whether the vortex barrier produces an O 3 /tracer correlation inside the vortex that is different from the one outside, regardless of ozone loss, what the appropriate timing is to establish a valid intravortex reference curve, and what the impact of continuous or abrupt air mixing across the vortex edge is on the shape of the inner vortex correlation. [7] In particular, once mixing started to be recognized as a factor influencing chemical ozone loss calculated by using O 3 -tracer correlations [e.g., Michelsen et al., 1998], various authors have differed over whether mixing could cause an overestimation [e.g., Michelsen et al., 1998; Plumb et al., 2000] or exclusively an underestimation of ozone loss [Müller et al., 2005]. Usually, the different conclusions reached by the various authors originated from different initial premises. For example, earlier studies did not always realize that the shape of the in-vortex correlation curve, before the beginning of ozone loss, was different from the shape of the extravortex correlation. [8] Recently, Manney et al. [2006] discussed Earth Observing System Microwave Limb Sounder (EOS MLS) O 3 and N 2 O measurements during the particularly cold Arctic winter , characterized also by several episodic mixing events. Using N 2 O as a tracer for quasihorizontal mixing and vertical descent, they evaluated the averaged cumulative ozone loss in the vortex and in the outer vortex regions from 23 January to 10 March. In their thorough-going discussion, it is emphasized that during extended periods of time quasi-horizontal mixing affected N 2 O mixing ratio values in both of the regions considered, prevailing over vertical transport and hence altering their estimates of air descent, ultimately making uncertain their calculations of ozone loss. Discussing results displayed by Manney et al. [2006, Figures 5 and 6], they make it clear that the mixing processes which took place during such a dynamically active winter either mimicked or masked chemical loss, depending on the vortex region and time. In particular, they also examined O 3 changes using O 3 -N 2 O correlation curves and found local O 3 decreases up to ppmv at 100 to 120 ppmv of N 2 O, but argued that this deficiency could be at least partially caused by mixing processes. [9] To date, most Arctic ozone/tracer correlation studies have emphasized results from cold winters with strong vortex isolation, in order to understand intravortex versus extravortex contrasts and signatures. As a result, relatively little work has been done on ozone/tracer correlations in weak vortex years marked by frequent cross-boundary advection. It is one of the purposes of this paper to investigate a year with these conditions. [10] A topic of significant interest is also that of correlation and feedback effects between variations of ozone concentration and temperature in the stratosphere. A complete description of the thermal structure and photochemistry of the middle atmosphere is not possible unless the ozone abundance is well characterized. Low ozone is almost always correlated to cold temperatures, owing to several processes: (1) PSCs form at low temperatures and cause the activation of ozone depleting halogens; (2) a well isolated vortex causes low temperatures inside the vortex as well as less ozone transported from midlatitudes to polar regions; and (3) radiatively, less ozone in the stratosphere leads to a decrease in temperature. However, somewhat complicating the question of photochemical depletion by chlorine, it has also been shown that air parcels observed in the middle polar stratosphere outside the polar vortex can be characterized by low ozone pockets, which are accompanied by relatively high temperatures. These extravortex regions of low ozone form in the middle stratosphere when air from the midlatitudes is trapped in the polar winter darkness of the Aleutian high for at least a few days, causing a shift toward less ozone in the photochemical balance between odd oxygen species and molecular oxygen [Manney et al., 1995; Morris et al., 1998; Nair et al., 1998]. Even though ozone concentrations above the lower stratosphere are less critical for determining the total ozone column and the solar UV-B radiation reaching the surface, they are very important for explaining the altitude structure of the stratospheric cooling observed over the past decades along with a decrease of the NH ozone levels during spring and a strengthening of the polar vortex. 2. Observing Techniques and Measurements Description [11] Lidar temperature profiles from Thule have been carried out by the University of Rome La Sapienza at various times since 1990 and represent a unique long-term data set on Arctic stratospheric temperatures. Temperature 2of14

3 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER profiles have been retrieved from lidar data following the method described by Marenco et al. [1997]. [12] The system transmitter is composed of a frequencydoubled, pulsed, Nd:YAG laser operating at 4 Hz and able to provide 200 mj per pulse at 532 nm. The beam is collimated by a beam expander so that its divergence is reduced to 0.15 mrad. The receiving optics consist of an 800 mm F/8 Cassegrain telescope and a photon counting apparatus. The acquisition of the backscattered signal is performed in photon counting mode by a multichannel analyzer. The signal from lower levels is cut off by a rotating chopper in order to avoid saturation of the photomultipliers due to the large signal intensity. In order to reduce the signal-to-noise ratio, the backscattered signal is integrated for 3 5 hours allowing the retrieval of temperature profiles between approximately 25 and 60 km altitude. The signal sampling frequency would set the vertical resolution to 150 m, but a smoothing is performed and the lidar signal is averaged over 4.5 km. Uncertainties on temperature values increase with altitude from 1 K at 25 km to 10 K at the maximum altitude probed. In January 2003, the laser was replaced by a new Nd:YAG source and temperature profiles since then are retrieved up to km. Temperature and pressure profiles up to 25 km used in this work are radiosonde data provided by the Weather Station at Thule. The sondes used are VAISALA RS80 and their temperature uncertainties are estimated to be 0.2 K from the ground to 50 hpa, 0.3 K from 50 to 15 hpa, and 0.4 K below 15 hpa. The uncertainty on the pressure measurements is 0.5 hpa. [13] O 3,N 2 O, HNO 3, and CO spectra were recorded by using a ground-based millimeter-wave spectrometer (GBMS) comprising a high sensitivity cryogenically cooled heterodyne receiver in conjunction with two Acousto-Optical Spectrometers (AOSs). One AOS (hereinafter referred to as wide band ) has a total spectral range of 600 MHz and a resolution of 1.2 MHz/channel, while a second AOS (hereinafter referred to as narrow band ) has a spectral window of 50 MHz tunable within the range of the wideband spectrometer, and a resolution of 65 KHz/channel. For a hypothetical spectral line with a pressure broadening coefficient averaging around 3 MHz/mbar, the combination of these two AOSs allows the retrieval of vertical profiles of atmospheric constituents from approximately 15 to 80 km altitude. The vertical resolution of the GBMS is limited by the inversion (deconvolution) of pressure-broadened spectra, and averages one pressure scale height, or 7 km in the Arctic lower and midstratosphere. Nonetheless, careful testing has proved that the GBMS is able to pinpoint the mixing ratio maximum of the observed chemical species (when a peak in mixing ratio is present) to less than 2 km altitude in the mid to lower stratosphere. [14] We observed the pure rotational transition lines of O 3 at GHz, N 2 O at GHz, CO at GHz, and a compact cluster of HNO 3 lines centered at GHz. Ozone and CO spectra are measured with a 1.5-hour integration. About 3 hours of integration are needed for HNO 3 and N 2 O lines, which are weaker than those of O 3 and CO. Given the altitude distribution of these species, we use only wide band data for O 3, HNO 3 and N 2 O, and only narrow band data for CO in this study. [15] The relevant measure of horizontal resolution for our data is directly determined by its temporal resolution, since stratospheric air is always moving past our observation point. From the observing durations given above for various molecules (1.5 hours for O 3, 3 hours for N 2 O, usually taken in immediate succession) and a typical stratospheric wind speed range of km/hr at 10 hpa altitude [e.g., Manney et al., 2005, Figure 2], we derive a typical swath of km along the wind direction, within which our O 3 /N 2 O correlations discussed in section 5 are established. However, in the direction across the wind stream (therefore in the direction across the boundary of the polar vortex when the GBMS samples air from the vortex edge region) the GBMS horizontal resolution is 20 km at the 520 K level. [16] Within the 600 MHz window sampled for O 3 measurements, we also observed the wing of a second ozone line centered at GHz, which is superimposed on the GHz line. For wide band data, the deconvolution of these two pressure-broadened spectral lines was carried out with an iterative matrix inversion method (hereinafter referred to as IMI) employing a vertical smoothing constraint [Twomey, 1977]. [17] In the past, all the GBMS O 3 wide band spectra [e.g., Cheng et al., 1996] were processed using the Chahine- Twomey (C-T) deconvolution technique [Twomey et al., 1977]. The key difference between these two techniques consists in the ability of the IMI algorithm (extensively used for our HNO 3 data processing [e.g., de Zafra et al., 1997]) to deconvolve spectra with multiple lines, while the C-T scheme is only useful to process single line spectra. In previous studies using GBMS O 3 data extending well above balloon altitude limits, we could independently calculate the expected emission intensity from the contaminating GHz line wing using ozonesonde data available for the location where the GBMS was operating (e.g., the South Pole [Cheng et al., 1996]) and subtract this background line from O 3 spectra. This was possible since the pressure-broadened line wing arises almost entirely from O 3 at altitudes accessible to balloonsondes. During January March 2002, however, no ozonesonde was launched from Thule. As a result, subtraction of the background emission line at GHz from spectra added too much uncertainty to the retrieval of GBMS O 3 profiles, and we decided to adopt the IMI instead. The IMI is essentially immune to the choice of the initial O 3 vertical profile used to start the iterative process, and this is one more advantage of IMI with respect to the C-T method. A drawback when using the IMI instead of the C-T is the larger uncertainties in the retrieved O 3 mixing ratios at the lowest and highest limits of the vertical extent of the deconvolution. From a detailed error analysis (which also involved running numerous tests to evaluate the accuracy of the deconvolution process applied to lines synthesized from known O 3 vertical profiles), and from comparisons with concurrent balloon and UARS/MLS O 3 data (unpublished results), we estimate a maximum 1s relative error on the retrieved O 3 mixing ratio vertical profiles of 75% up to 17 km altitude, 33% from 18 to 20 km, 15% from 21 to 27 km, and less than 13% from 29 to 65 km, with a minimum absolute error of 0.2 ppmv. In fact, GBMS O 3 retrievals obtained using the IMI technique agree within 10% accuracy with concurrent balloon-borne 3of14

4 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER Figure 1. (a) Colored contour maps of lidar/balloon temperature; (b) GBMS O 3 ; (c) CO; and (d) N 2 O, with the relative color scale to the right of each panel. In Figures 1b 1d, lidar/balloon temperature contours of Figure 1a are superimposed on the colored maps with solid black lines. White gaps in GBMS data indicate when no measurement is available for more than 48 hours, because of either poor weather conditions, instrumental malfunctioning, or unreliable spectral data. 4of14

5 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER Figure 2. Contour map of GBMS HNO 3. measurements in the lower stratosphere above 18 km when the latter are available (e.g., during the several South Pole campaigns or during the Thule campaign of winter ). [18] N 2 O retrievals are obtained using the C-T deconvolution technique and have an overall 1s uncertainty of 15% or 5 ppbv, whichever is larger [Crewell et al., 1995]. Early comparisons of the GBMS N 2 O retrievals with other data sets are given by Emmons et al. [1994]. Furthermore, we checked the 2002 GBMS N 2 O values observed inside the Arctic vortex against other N 2 O data sets available in the literature for the Arctic vortex region [Bauer et al., 1994; Emmons et al., 1994; Urban et al., 2004; Huret et al., 2006]. We found that the in-vortex 2002 GBMS N 2 O mixing ratio values are consistent with values found in the literature for altitudes above 18 km (460 K), but at the lowest isentropic level considered in this study (435 K or 17 km), GBMS in-vortex N 2 O data are systematically lower than values in the literature by about ppbv. Although such potential bias does not affect the general results and conclusions of this paper, we will not speculate on results obtained with the GBMS data at the 435 K isentropic level. [19] Nitric acid profiles are obtained by inverting a cluster of HNO 3 lines with the IMI method. The overall 1s uncertainty in the retrieved HNO 3 mixing ratio profiles varies with altitude and time and is bounded between 16% and 22% or 0.5 ppbv, whichever is larger [de Zafra et al., 1997; Muscari et al., 2002]. GBMS CO mixing ratio vertical profiles from the two Arctic campaigns of and have already been published by de Zafra and Muscari [2004], where a detailed description of data reduction and retrieval accuracy are presented. In the present work, we have highlighted the CO profiles in the altitude range between 30 and 45 km (see Figure 1), where the overall 1s uncertainty is below 0.15 ppmv. 3. Arctic Meteorological Conditions During Winter [20] A detailed account of the meteorological conditions in the Arctic stratosphere during winter (which we make extensive use of in this section) is provided by the European Ozone Research Coordinating Unit (EORCU) based at the Department of Chemistry at the University of Cambridge, and can be obtained on the World Wide Web at A complementary discussion giving year-by-year comparisons of larger-scale structure and long-term trends in the Arctic stratosphere over the period is given by Manney et al. [2005]. [21] The Arctic vortex developed during October and November 2001, and at the end of November, for a couple of days, the temperature fell below thresholds of formation for PSC particles over the altitude range km. A minor warming began on 30 November and lasted until 8 December, followed by a return to conditions suitable for formation of PSCs in a restricted region within the vortex, in the altitude range km on 9, 10, and 11 December. A strong warming then took place, lasting until early January. PSC conditions were again established briefly in the lower stratosphere on 6 9 January, and this was the last occasion in the 2001/2002 winter for PSC formation on a synoptic scale. These theoretical time limits on PSC formation agree well with observations by the Polar Ozone and Aerosol Measurement (POAM) III instrument [e.g., Lucke et al., 1999] aboard the SPOT 4 satellite, which detected PSCs from 26 to 30 November and during mid-december, when PSCs were observed at altitudes as high as 27 km. PSCs were again observed by POAM III from 7 to 17 January, after which no further PSCs were detected (EORCU, winter report 2001/2002). The lack of sunlight during the period of activation limited the amount of chemical loss that took place to perhaps about 10% in the total ozone column averaged over the Arctic vortex [Goutail et al., 2005]. The ozone loss that took place during this winter is at the low end of the range observed over the previous ten years. [22] As previously stated, during winter the polar vortex was strongly perturbed by the Aleutian high (AH) which gives rise to so-called early warming events in the polar winter, as well as the low ozone pockets described in section 1. For a detailed morphology of the AH and its 5of14

6 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER ratio for O 3, CO, and N 2 O from January to March 2002, with balloon/lidar stratospheric temperatures also superimposed on each molecular contour map. A contour map of GBMS HNO 3 mixing ratio values during the same period of time is shown in Figure 2. Balloon temperatures provided by the Weather Station at Thule are used up to 24 km, between 24 and 27 km we use a weighted average of balloon and lidar data, and from 27 to 45 km altitude the data are purely from lidar. [24] During the first few days of data (19 23 January), several lines of evidence point to a slanted or perhaps layered vortex boundary over Thule. CO and temperature measurements are consistent with the presence of the vortex above Thule at altitudes between 30 and 45 km (Figures 1a and 1c). Temperatures are lower and CO mixing ratios are larger with respect to days immediately preceding and following this first period of observations. In the middle atmosphere enhanced concentrations of CO are a valuable mark for vortex air [de Zafra and Muscari, 2004] since CO is formed mainly in the mesosphere and thermosphere and is transported downward by the descent of air inside the polar vortex. In Figure 3, the time series of Ertel s Potential Vorticity (PV) based on NCEP reanalysis at potential temperatures of 1000 K (34 36 km altitude) and 1300 K (42 45 km) show large values of PV between days 19 and 23, providing additional independent evidence that the vortex was over Thule at those higher altitudes. During the same days, however, altitudes near 27 km over Thule are outside the vortex, as suggested by N 2 O mixing ratio (mr) values above 100 ppbv (Figure 1d) [e.g., Manney et al., 1999] and temperatures which remain constant between 214 and 218 K at 27 km altitude. The PV time series at 750 K (Figure 3) shows relatively small values and further confirms the presence above Thule of extravortex air in the altitude range around 27 km prior to 25 January. A slanted vortex above Thule is a feature observed also later in the winter by means of GBMS N 2 O measurements. PV values at 475 K (19 km) increase consistently only Figure 3. From bottom to top, time series of Ertel s Potential Vorticity (PV) over Thule based on NCEP reanalysis at potential temperature levels of 475 K, 550 K, 750 K, 1000 K, 1300 K (17 18, 21 22, 26 27, 34 36, and km), respectively. Values on the vertical scales are in PV units (1 PVU = 10 6 Km 2 kg 1 s 1 ). Dashed horizontal lines are indications of the inner vortex edge determined by the maximum value in PV gradient. interaction with the polar vortex see, e.g., Harvey and Hitchman [1996]. 4. Evolution of Stratospheric Chemistry and Dynamics Over Thule [23] In Figure 1, four colored contour maps show stratospheric values of balloon/lidar temperatures and mixing Figure 4. Time series of ozone columnar content from 17 to 45 km. 6of14

7 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER Figure 5. Ertel s Potential Vorticity (PV) maps based on NCEP reanalysis at 475 K for (left) 4 February and (right) 7 February. The location of Thule is indicated with a solid circle. Contour lines indicate PV values from 2.4 to 4.0 K m 2 kg 1 s 1, at 0.4 intervals. after 24 January, when the temperature gradient also steepens toward low values. With very few GBMS measurements of N 2 O available during this time window (only days 21 and 22 prior to day 31), we have chosen temperature and PV data to provide a better indication of the timing of the vortex passage over Thule. [25] From 26 January to 1 February (days 26 to 32), the full observed column of stratospheric air above Thule moves inside the vortex within at most 2 days, working from lower to higher altitudes. This can be inferred in Figure 1 by observing the steep decrease in lower stratospheric temperature occurring from days 24 to 29, the time lag in decreasing temperatures in the km range, the small N 2 O mixing ratios at km (not available during this interval until day 31), as well as the PV values shown in Figure 3. Also CO measurements are consistent with the presence of vortex air above Thule between days 26 and 32. This is illustrated in Figure 1c by both the large amount of CO in the km altitude range and by the small peak in CO mixing ratio on day 28 at 32 km altitude, and can also be seen clearly at higher altitudes (above 45 km) as shown by de Zafra and Muscari [2004, Figure 6]. [26] During this in-vortex observing period, the GBMS measured stratospheric O 3 mixing ratio values lower than those observed outside the vortex (see Figure 1) and the GBMS O 3 columnar content between 17 and 45 km (displayed in Figure 4) reaches its minimum of (140 ± 21) DU on 31 January (O 3 and N 2 O data points from this day are marked in Figure 9 in section 5). During the following 10 days, the position of the polar vortex with respect to Thule is not easily determined, and depends on the isentropic level considered (as later discussed). In Figure 5, two PV maps based on NCEP reanalysis at 475 K show Thule in the proximity of the vortex edge, more toward the outside (day 35) or the inside (day 38) of the vortex at this potential temperature level. This rapid shift in the vortex position with respect to Thule caused the O 3 column contents (Figure 4) to vary considerably from day to day, remaining below 190 DU most days, but jumping up to 234 ± 35 DU on day 33. Of course, the steep decrease to 140 DU surrounding days is not to be ascribed solely to chemical ozone loss, since vertical descent inside the vortex (which brings large amounts of ozone below 17 km altitude) and the relative height of the tropopause can have a large impact on the observed difference. We specifically note in Figure 1b a general diminishment of O 3 at altitudes between approximately 20 and 35 km during this in-vortex period. Since air from outside the vortex at this time of the year cannot be considered a valid representation of conditions inside the vortex at the time of its formation [e.g., Plumb et al., 2000], we will not attempt to calculate the vortex air descent rate comparing in-vortex and extravortex GBMS N 2 O measurements. In the discussion of O 3 -N 2 O correlations in the following section, we suggest that chemical loss at lower altitudes played a role in the small values of columnar ozone measured around day 31. [27] The trough in O 3 mixing ratio at 32 km (see Figure 1b, days and 46 55) is a feature of midstratospheric O 3 mixing ratios observed both in the Antarctic [Cheng et al., 1996] and in the Arctic [Manney et al., 1995] and is characteristic of regions isolated in the dark for several days [Morris et al., 1998; Nair et al., 1998]. The ozone photochemical lifetime below 26 km is long (approximately months) in the winter hemisphere unless aerosols or PSCs are present, whereas above 40 km photochemistry prevails over transport and the ozone lifetime is approximately a day or less. The ozone concentration in the atmospheric layer between 26 and 40 km is 7of14

8 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER Figure 6. Illumination fraction along 10-day back trajectories ending over Thule during late January to early February 2002 (solid circles connected by a solid line) and GBMS ozone mixing ratio (open triangles connected by a dotted line) at the potential temperature level of 900 K. The illumination fraction is defined as the ratio between the time interval during which the air parcel received solar illumination and the total duration of the trajectory. therefore driven by the transition from dynamical to photochemical timescales, and if transport is cut off in polar winter regions, the ozone will readjust to the lower concentrations implied by its photochemical equilibrium in dark regions. The typical offset from the Pole for the center of vortex circulation in the Arctic, particularly during years of a recurrent strong Aleutian high, allows a significant fraction of air within the vortex to experience some length of exposure to sunlight. Using 10-day air parcel trajectories at 900 K (32 km), we find however, as shown in Figure 6, that air masses which arrived over Thule during this period around 29 January to 1 February experienced very little solar exposure compared to air masses arriving shortly before and after these dates. As illustrated in Figure 6, the minimum of solar exposure is reached during 29 and 30 January, matching very well the minimum in O 3 mr at 900 K. Figure 7 shows that a good correlation exists throughout the observing period between O 3 mixing ratio values measured at 900 K and the solar exposure experienced by the air parcel during the 10 days preceding the measurement, indicating that at this altitude O 3 photochemical lifetime prevails over transport. [28] As mentioned earlier, in the period between 2 and 12 February (days from 33 to 43), Thule is in the vicinity of the vortex edge (see Figure 5) and the N 2 O m.r. (Figure 1d) as well as PV time series (Figure 3) show rapid changes in air characteristics from day to day and from one altitude level to the next (see also Table 1). The stratosphere over Thule below km and above about 40 km altitude shows vortex characteristics, evident from larger than usual CO mixing ratio values above 40 km (Figure 1c), low N 2 O concentrations at around 20 km (Figure 1d), and large PV values at 475, 1000, and 1300 K (Figure 3), while at the same time the stratosphere at around km ( K) is often characterized by large N 2 O concentrations and low PV values, which are both typical of extravortex air. [29] On February (days 43 44), the low altitude vortex moved away from Thule. This is seen in Figure 1d, which shows N 2 O mixing ratio and temperature sharply increasing at altitudes up to km. This is shortly followed by a sharp drop in O 3 mixing ratios in the km range, clearly visible in Figure 1b by day 46. This feature is similar in appearance to the midstratospheric O 3 depletion seen through days when the vortex was over Thule, but the temperature field of Figure 1a, as well as CO, N 2 O (Figures 1c and 1d), and PV (Figure 3) all have the signature of extravortex air. Air parcel trajectories at 900 K (33 km) displayed in Figure 8 and obtained from the automailer system of the GSFC [Schoeberl and Sparling, 1994] show the passage from a cyclonic (vortex) to an anticyclonic circulation over Thule on 15 February (day 46). All the trajectories at 900 K arriving over Thule from 15 February to 5 March (not shown) confirm the continuous presence of the anticyclonic circulation. During the first 10 days of this period, the stratosphere is characterized by particularly low values of the ozone mixing ratio above 30 km (see discussion above on the photochemical equilibrium at these altitudes) and by high temperature values measured by the lidar above 25 km. Starting at around day 55, the air sampled by the GBMS still belongs to the anticyclone but experiences a longer solar exposure during the previous 10 days and the photochemical equilibrium shifts to larger Figure 7. Correlation scatterplot between all the winter GBMS ozone mixing ratio values at 900 K and illumination fraction values along 10-day trajectories at 900 K (the illumination fraction is defined as in Figure 6). The correlation coefficient r is of14

9 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER Table 1. Results of the Selection Process That Associated Each O 3 -N 2 O Correlation Data Point Shown in Figure 9 to the in-vortex (In), to the Vortex Edge (Edge), or to the Extravortex (Out) Regions a D\q edge edge edge edge out out out edge in in in 31.9 in in in in in in in in in in in 32.9 in in in in in in in in in in in 34.9 in in in edge edge out out out out out out 39.9 in in in in edge out out out edge in in 40.9 in in in in edge out out out in in in 41.9 in in in in edge edge out out out in in 43.0 edge edge edge edge edge edge edge edge edge edge edge 45.0 edge edge edge edge edge out out out out edge edge 48.9 out out out edge edge edge edge out out out out 50.7 out out out out out out out out out out out 52.8 out out out out out out edge edge out edge edge 54.8 edge edge edge edge out out out out out out out 56.0 out out out out out edge edge edge out out out 57.7 out out out out out edge edge edge out out out 57.9 out out out out out edge edge edge out out out 59.9 out out out out out edge edge edge out out out 62.1 out out out out out out edge edge out out out 64.2 out out out out out out edge edge out out out a Days of data (D) are expressed in decimal day of the year Theta levels (q) are in Kelvin. Note that not all the days of measurements are listed here, but only those included in the O 3 -N 2 O correlation study. amounts of O 3. Ten-day solar exposure and O 3 mixing ratio values at 900 K are well correlated also throughout this period and the corresponding data points are displayed in the correlation graph of Figure 7. This feature in air parcels processed through the anticyclonic Aleutian high was noted (and termed ozone pockets ) by Manney et al. [1995] and later attributed to an altitude-dependent shift in photochemical equilibrium between O, O 2 and O 3 by Morris et al. [1998] and Nair et al. [1998]. The presence, strength, and duration of the Aleutian high is directly related to major stratospheric warmings (see for example Manney et al. [2005] for a definition of such events) and to the shift of a weak polar vortex off the North Pole [Manney et al., 1995]. The analysis of Manney et al. [2005] also records that during the winter of a second major warming took place in the Arctic stratosphere during mid- February 2002, in agreement with the timing of lidar and GBMS observations of air processed by the Aleutian high over Thule. Starting on day 45, mesospheric CO is confined to higher altitudes with respect to previous days [see de Zafra and Muscari, 2004, Figure 6a] suggesting that the strong anticyclonic circulation from the Aleutian high affects also vertical motion in the mesosphere [see also de Zafra and Muscari, 2004]. [30] For the entire winter period discussed so far, the amount of HNO 3 recorded by the GBMS inside the polar vortex (starting on 22 January, see Figure 2) does not show any clear signs of depletion due to uptake onto PSC particles. This is in agreement with the lidar soundings that show no PSC particles over Thule and with POAM III measurements that did not detect PSCs anywhere in the vortex after 17 January (see section 3). However, it should be pointed out that the uptake of small quantities of HNO 3 onto PSCs particles might not be detected by the GBMS, Figure 8. Ten-day back trajectories at 900 K ending over Thule on 12 February, 22 February, and 4 March (days 43, 53, and 63). Five trajectories are plotted in each case. One ends exactly at Thule, and the other four end at locations surrounding Thule at four cardinal points separated 90. Numbers along each trajectory indicate the distance in days to the arrival of the air parcel at Thule. Trajectories arriving in Thule on 12 February display a cyclonic circulation, whereas trajectories drawn in the center and right plots show an anticyclonic circulation. 9of14

10 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER given the 16 22% of overall 1s uncertainty on the measured HNO 3 mixing ratios (section 2). 5. Impact of Mixing Across the Vortex Edge [31] As discussed above, correlation relationships between O 3 and long-lived tracers have been employed in a number of studies aimed at quantifying O 3 loss. Correlations between GBMS O 3 and N 2 O data points, as well as the N 2 O and O 3 vertical profiles used to produce the correlations, are displayed in Figure 9. In order to construct these correlations, our derived O 3 and N 2 O mixing ratio profiles were first interpolated onto 11 potential temperature levels (from 435 K to 1005 K, see Table 1, at vertical intervals of approximately 2 km altitude) and then values belonging to each level were paired together. Correlation curves were constructed using O 3 and N 2 O observations carried out typically in immediate succession, therefore within a total of 4 to 5 hours (19 pairs out of a total of 33 measurements of both N 2 O and O 3 ), and the dates of these selected observations are listed in Table 1. Out of the 19 correlation profiles obtained, we have displayed correlation points designated as belonging to in-vortex, extravortex, and edge of the vortex air masses (see below for the selection criteria), using solid stars, solid circles, and open triangles, respectively. Symbols are displayed in color to distinguish data points relative to the different isentropic levels, and the uncertainties in N 2 O and O 3 measurements at each isentropic level are plotted at the bottom of Figures 9b and 9c, respectively. However, the uncertainties displayed are only indicative since they are computed by applying the known percentage error (see section 2) to average profiles of N 2 O and O 3 (the latter displayed in Figure 9c with a solid black line, see caption of Figure 9 for more details). As already pointed out in section 2, given the poor agreement between the GBMS N 2 O mixing ratios and other N 2 O measurements from inside the Arctic vortex at the lowest level considered (435 K or 17 km), we will not base our conclusions on results obtained at this isentropic level. However, we note that results at this level are consistent with the behavior of O 3 and N 2 O mixing ratios that is observed at the other lower stratospheric isentropic levels. [32] In order to associate data points with a specific region (inside, outside, or at the edge of the vortex) we used GBMS N 2 O measurements from the lowest 8 levels (435 K to 775 K), and NCEP reanalysis PV data for the upper 3 levels 840 K, 915 K, and 1005 K. In order to assign GBMS O 3 -N 2 O data points to one of the above mentioned regions, in the lower stratosphere we trust the GBMS N 2 O observations (O 3 and N 2 O measurements were carried out within a total of 4 to 5 hours) more than the temporally and spatially coarser Potential Vorticity data analysis [e.g., Greenblatt et al., 2002]. Although for most days of GBMS observations the two different tracers (PV and N 2 O) are consistent with one another in indicating the intravortex versus extravortex origin of air masses sampled with the GBMS, the two tools do not perfectly agree in the complex period between days 31 and 41, when the vortex edge region remains close to Thule (see Figures 1d, 3, and 5). As altitude increases the N 2 O gradient across the vortex edge is not as sharp as it is in the lower stratosphere, and it is too small compared to the GBMS N 2 O mixing ratio uncertainties for relying on N 2 O measurements when separating invortex from extravortex air. For this reason, we switched to PV time series to distinguish between vortex and extravortex O 3 and N 2 O data points at the upper 3 isentropic levels considered in Figure 9. These upper levels and therefore the PV values we chose as boundaries of the vortex edge at each level are not critical for the discussion that follows. [33] We emphasize that what in the present work is defined as the vortex edge region (a region that, as discussed below and shown in Figure 9b, at 500 K is constrained between 65 and 140 ppbv of N 2 O mixing ratio at the end of January) is at times considered in the literature to be inside the vortex. This is the case when, for example, only the center of the vortex edge is taken as the boundary separating in-vortex from extravortex air. In Figure 9b, the dotted and dashed black lines depict the N 2 O profiles used as the outer limit of the vortex core and the center of the vortex edge, respectively, as deduced from the literature [Greenblatt et al., 2002; Manney et al., 2003; Urban et al., 2004; Manney et al., 2006]. Over time, descent of air inside the vortex changes the N 2 O profiles at the vortex edge boundaries, and the N 2 O profiles we have chosen and displayed in Figure 9b are relative to the end of January/ beginning of February, when most of the GBMS observations inside the polar vortex were carried out. In Figure 9a we also overlay the O 3 -N 2 O correlation curves indicated by Müller et al. [2005] and Tilmes et al. [2004, hereinafter referred to as TM_0405] as representative of the early winter vortex air (deduced, in fact, combining Müller Figure 9. (a) O 3 -N 2 O correlations, (b) N 2 O mixing ratio vertical profiles, and (c) O 3 mixing ratio profiles. In all panels, solid stars indicate data points relative to intravortex air, solid circles indicate data points relative to extravortex air, and open triangles depict air associated to the vortex edge region. Large open stars around solid stars indicate GBMS data from 31 January, while reference to Table 1 shows that the three groups of six closely bunched solid stars at levels 435 K, 480 K, and 520 K come from days (late January to mid-february). Colors indicate the isentropic level to which each point refers, from 435 K to 1005 K (see Table 1). In Figure 9a, O 3 -N 2 O correlation curves deduced from Müller et al. [2005] and Tilmes et al. [2004] and representative of the early winter vortex air and of the typical chemically disturbed vortex air (see discussion in text) are depicted with black and red dotted lines, respectively. The two dotted lines for each color represent the upper and lower boundaries of these reference curves. The extravortex air reference curve, also from Müller et al. [2005], is shown with a solid black line. In Figure 9b, the dotted and dashed black lines represent the outer limit of the vortex core and the center of the vortex edge, respectively. At the bottom of Figure 9b, 1s uncertainties for a typical N 2 O mixing ratio profile are shown, with the typical profile located between the dotted and dashed black lines (not shown for clarity). At the bottom of Figure 9c, 1s uncertainties for a typical O 3 mixing ratio profile are shown, with the profile indicated with a solid black line. 10 of 14

11 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER Figure 9 11 of 14

12 MUSCARI ET AL.: ARCTIC STRATOSPHERE IN WINTER et al. [2005, Figure 7] and Tilmes et al. [2004, Figure 4], with black dotted lines indicating the upper and lower boundaries), of the typical chemically disturbed vortex air (red dotted lines indicating the upper and lower boundaries), and of extravortex air (solid black line). These curves were derived by TM_0405 from a large number of O 3 /N 2 O correlation curves obtained using observations from the Halogen Occultation Experiment (HALOE) and the Improved Limb Atmospheric Spectrometer (ILAS), with early winter reference curves displaying a significant variability from year to year [e.g., Tilmes et al., 2004, Figures 4 and 5]. [34] There are several features worth pointing out in Figure 9a: [35] 1. The extravortex O 3 -N 2 O relationship obtained using GBMS data agrees well with the averaged extravortex reference curve of Müller et al. [2005], though our solidcolor dots generally show larger values of O 3 at midaltitudes (green through dark blue); [36] 2. At most isentropic levels, data points identified as belonging to the edge of the vortex region (open triangles) lie in between intravortex and extravortex data points, suggesting they lie on lines of isentropic mixing. [37] 3. Data points at 480 K (gray solid stars, open triangles, and solid circles) run very close to the correlation curve (i.e., within the red dotted lines) that Müller et al. [2005] consider a typical correlation that develops when an established ozone loss has taken place inside the vortex. They are characterized by an O 3 deficiency of 1.5 ppmv at 40 ppbv of N 2 O (in-vortex data points) and of 1.2 ppmv at 120 ppbv of N 2 O (vortex edge data points) with respect to the early vortex relation (bounded by black dotted lines) deduced from TM_0405. [38] 4. The in-vortex GBMS O 3 -N 2 O correlation points at 435 K, 480 K, and 515 K (black, gray, and blue solid stars) fit the established ozone loss curve very well. Reference to Table 1 shows that all these low altitude intravortex points (the three groups of six closely bunched stars) come from days (late January to mid-february). Had we chosen to use PV values (Figure 3) for these low altitude points, only days 31 and 32 would have been classified as intravortex. [39] The ozone deficiencies observed in Figure 9a (and described above) are quite substantial and could be attributed to chemical ozone loss, although such a significant loss is generally observed in the Arctic only during very cold winters and intense PSCs activity, which is not the case for winter (see section 3). However, the low O 3 mixing ratios measured at 480 K (and at 435 K, which, however, we do not rely upon for our conclusions, as mentioned above) inside the polar vortex (gray stars in Figure 9a) seem difficult to explain in any other way but through chemical ozone loss, and only in part could be accounted for by the large relative uncertainties in the GBMS O 3 data at these lower levels (see Figure 9c). [40] If ozone loss has in fact taken place inside the vortex, and was observed by the GBMS at the end of January/ beginning of February (see Table 1 and Figure 1), the activation of chlorine and bromine compounds must have occurred by early January, when temperatures conducive to PSC particles formation were last registered inside the Arctic vortex. A preliminary analysis of the Odin/SMR measurements indicates that elevated ClO amounts were present at 21 km altitude over northern Europe and Greenland on 7 8 January 2002 (EORCU, winter report 2001/ 2002), confirming that a short-lived activation of chlorine compounds happened in early January. Although PSCs were detected by POAM III as late as 17 January, no enhancement of Odin/SMR ClO after 7 8 January was reported (EORCU, winter report 2001/2002). Ozone loss could therefore have happened in the first half of January and, because of the long lifetime of ozone in the lower stratosphere, the deficiency could have persisted at least until the beginning of February, when the GBMS observed it inside the polar vortex. This loss, however, must have taken place in a limited region of the vortex, perhaps because of limited transport into sunlight in January. Still further confirmation of this can be found in the estimates of 10% for the total ozone loss for winter 2001/2002 obtained by Goutail et al. [2005] using SAOZ data and a 3-D chemistry transport model. [41] If air parcels corresponding to the grey open triangles, representing in Figure 9a the vortex edge region at 480 K, were not influenced by chemical loss, we would expect larger O 3 values at N 2 O mixing ratios between 100 and 150 ppbv. In particular, these data points (the gray open triangles) lack about 1 ppmv of O 3 to fit well with the early vortex reference curve deduced from TM_0405. It is unlikely that these air parcels, with N 2 O mixing ratios characteristic of the vortex edge, have been affected by local ozone loss. It is instead plausible that air masses lying at the edge of the vortex have been affected through mixing by the ozone loss occurred in the vortex core. These data points (together with the 435 K correlation points indicated in Figure 9a by black open triangles) could be misunderstood as a sign of local ozone loss when compared to the early vortex relation from TM_0405. This implies a local overestimation of ozone loss which affects, above all, regions located in the proximity of the vortex edge. 6. Summary [42] The Arctic lower stratosphere was unusually warm during the winter. Observed temperatures were below PSC formation thresholds only for a very few days and in very limited areas, and never after mid-january. Consistent with this picture, during the winter activation of ozone-depleting halogens in the lower stratosphere was last observed on 7 8 January 2002 (EORCU, winter report 2001/2002). The winter is therefore a valuable opportunity for quantifying changes in overall ozone mixing ratio profiles in the Arctic vortex due primarily to dynamical processes or homogeneous (gas phase) chemistry. [43] We present measurements of stratospheric constituents obtained by a millimeter-wave spectrometer and a lidar system working concurrently at the NDACC site of Thule Air Base from mid-january to early March Using the GBMS stratospheric O 3, CO, N 2 O, and HNO 3 measurements and lidar temperature observations we characterize the polar stratosphere over Thule in the altitude range between 17 and 45 km, discussing dynamical and chemical processes that took place in the stratosphere inside and outside the Arctic vortex. This was made possible by the 12 of 14