Chemical ozone loss in the Arctic vortex in the winter 1995–96: HALOE measurements in conjunction with other observations

. Severe chemical ozone loss has been detected in the Arctic in the winter and spring of 1995–96 by a variety of methods. Extreme reductions in column ozone due to halogen catalysed chemistry were derived from measurements of the Halogen Occultation Experiment (HALOE) on board the Upper Atmosphere Research Satellite in the Arctic vortex. Here, we discuss further aspects of the HALOE observations in the Arctic over this period. Potential problems, both in the data themselves and in the methodology of the data analysis are considered and the reason for the di(cid:128)erences between the Arctic ozone losses deduced from HALOE data version 17 and 18 is analysed. Moreover, it is shown that HALOE measurements in the Arctic in winter and spring 1995–96 compare well with observations by other ground-based and satellite instruments.


Introduction
Since 1989, chemical ozone loss in the Arctic vortex has been inferred from ozonesonde and aircraft measurements (e.g. Hofmann et al., 1989;Prott et al., 1990Prott et al., , 1993Hofmann and Deshler, 1991;Koike et al., 1991;KyroÈ et al., 1992;Browell et al., 1993;von der Gathen et al., 1995;Rex et al., 1998a), albeit to a lesser extent than over Antarctica. More recent observations that also included satellite data (Larsen et al., 1994;Manney et al., 1994aManney et al., , 1996aDonovan et al., 1995;Bojkov et al., 1995;Wirth and Renger, 1996;Goutail et al., 1998a;Rex et al., 1998b) have indicated particularly strong chemical ozone loss in the Arctic vortex for early 1993 and 1995. Chemical ozone change in the Arctic is dicult to quantify, since dynamical processes cause substantial ozone variations (see e.g. Manney et al., 1994a;von der Gathen et al., 1995;MuÈ ller et al., 1996, and references therein). Total ozone is a quantity that is especially problematical in this respect, as it is both subject to signi®cant short term reductions due to high-pressure systems in the troposphere and characterised by a strong seasonal cycle in high latitudes. Climatological values for Northern Hemisphere high-latitude stations show an increase of total ozone between November and March by more than 100 DU (e.g., Hansen et al., 1997).
In March 1996 and, record low values of total ozone were measured in the Arctic vortex ; an observation which triggered further detailed investigations (e.g. GRL, 1997). While for both winters substantial chemical ozone depletion is inferred (e.g. Donovan et al., 1996Donovan et al., , 1997Manney et al., 1996bManney et al., , 1997MuÈ ller et al., 1997a, b) there is consensus that direct dynamical eects contribute signi®cantly to the unusually low ozone columns observed in March 1997 (Donovan et al., 1997;Manney et al., 1997). Consistent with these ®ndings, the proxy ozone derived from HALOE tracer observations, which re¯ects the dynamical situation in the absence of chemistry, indicates much less descent (and thus lower total ozone) in the vortex in March 1997than in March 1996(MuÈ ller et al., 1997a. Here, we focus on the situation in the polar vortex in the Arctic winter and early spring of 1995±96. During this winter, temperatures were both extremely low and more persistently low than usual, if compared to long term records Naujokat and Pawson, 1996). This caused frequent polar stratospheric cloud (PSC) formation and subsequent chlorine activation of large vertical extent and uncharacteristically long duration (Hansen and Hoppe, 1996;Santee et al., 1996a;MuÈ ller et al., 1997a). Indeed, formation and sedimentation of ice particles as well as dehydration of stratospheric air were observed for the ®rst time in the Arctic vortex in January 1996 (VoÈ emel et al., l997;Hintsa et al., 1998). Consistent with the continuing large-scale chlorine activation, anomalously low ozone levels were observed in the vortex throughout the lower stratosphere (Donovan et al., 1996;MuÈ ller et al., 1997a), large chemical ozone loss rates were derived from ozonesonde observations (Rex et al., 1997), and unusually low ozone columns were observed by ground-based (Hansen et al., 1997;Bojkov et al., 1998;Goutail et al., 1998b) and satellite (MuÈ ller et al., 1997a) instruments in the Arctic. Further, total ozone column measurements of the GOME instrument (see later), indicate that the whole stratospheric vortex in March 1996 is characterised by extremely low ozone levels. Moreover, the expected enhancement of ground level UV-B radiation was observed (under clear sky conditions) at two midlatitude stations: Glasgow, 56 N (Moseley and MacKie, 1997) and Garmisch-Partenkirchen 47.5 N (Seckmeyer et al., 1997); UV-B levels signi®cantly higher than normal for the time of the year were measured during the ®rst days of March 1996.
We concentrate here on the analysis of observations by the Halogen Occultation Experiment (HALOE) (Russell et al., 1993) on the Upper Atmosphere Research Satellite (UARS) inside the Arctic vortex in winter and spring 1995±96. The relationship between an eectively inert trace substance (CH 4 ) and chemically more active compounds (HCl and O 3 ) in this data set have been used previously to deduce chlorine activation and chemical ozone loss (MuÈ ller et al., , 1997a. The purpose of this contribution is to provide further details of the observations in 1995±96 and to investigate questions that could not be fully discussed in MuÈ ller et al. (1997a) in particular the following questions: belts (Russell et al., 1993). Because of the UARS orbit, the location of these latitude belts varies with season. In Fig. 1 (Nash et al., 1996), which occurred in early November 1995(Coy et al., 1997. Compared to mid-latitude air, vortex air is characterised by enhanced potential vorticity (PV) and, due to diabatic descent, substantially lower CH 4 mixing ratios. Especially towards the vortex edge, in a region of strong gradients of both PV and tracer mixing ratios, however, the resolution of PV computed from meteorological analyses is not sucient to yield a good correlation with tracer observations, which show more ®ne scale structure (Fairlie et al., 1997;Tuck and Prott, 1998). Here, we use the HALOE tracer (CH 4 ) observations to discriminate between mid-latitude and vortex air and discuss PV as a corroborating evidence. Four examples of occasions when HALOE sampled Arctic vortex air in the winter 1995±96 are shown for CH 4 and O 3 in Fig. 2 for November 22, 1995, at 47 N, in Fig. 3 for January 30, at 49 N, in Fig. 4 for March 16, at 64 N, andin Fig. 5 for April 1, 1996, at 70 N. CH 4 and O 3 mixing ratios are shown versus longitude and potential temperature (as the vertical coordinate). Owing to downward transport through diabatic descent in the vortex, CH 4 mixing ratios inside the vortex are lower on a speci®c isentropic surface than outside. On all four days, vortex air characterised by lower CH 4  16, 1996 (uarsday 1648), at 64 N. Vortex air is discernible over the whole altitude range at 294 E mixing ratios is discernible (though not always over the whole altitude range), much as observed in previous years . To use the relationship between an eectively inert trace substance and chemically more active compounds has been established as a method to discriminate chemical change from large variations due to dynamical processes (e.g., Prott et al., 1993;. Compact relationships are expected for species with suciently long photochemical lifetimes (Plumb and Ko, 1992). In particular, neglecting mixing across the vortex boundary, an unchanging relationship between ozone and chemically inert tracers (such as N 2 O, CH 4 or HF) is predicted for the air mass inside the polar vortex, if no chemical ozone loss would occur (Plumb and Ko, 1992). This is actually observed in HALOE measurements of the O 3 / CH 4 relation between November 1995 and late January 1996 (MuÈ ller et al., 1997a), i.e. over a time period when no substantial chemical ozone loss is expected due to the lack of sunlight. Although signi®cant ozone loss rates (in ppb per hour of sunlight) have been observed for January 1996, the accumulated ozone loss then is still small (Rex et al., 1997). Furthermore, the relationship of the inert compounds HF and CH 4 in the Arctic vortex in 1995±96 is conserved between November 1995 and April 1996 (Fig. 6). Any deviation from the initial O 3 /CH 4 relation in the vortex over winter and spring is therefore an indication of chemical ozone change.
From this analysis, MuÈ ller et al. (1997a) concluded that local chemical O 3 destruction has led to reductions of about 60% over the height range of 400±480 K (with peak losses of about 70%) between late January and early April. This is consistent with the results of Rex et al. (1997), who infer an ozone loss rate of 27±34 ppb per day in the lower stratosphere over this period from ozonesonde measurements that corresponds to an accumulated O 3 loss at about 470 K of % 64%. Somewhat lower ozone loss rates (22 ppb per day for February 1996) are derived from MLS observations , the temporal behaviour of O 3 , however is very similar. The local chemical ozone loss in the vortex in the lower stratosphere that is re¯ected in a change in the relation of O 3 and CH 4 mixing ratios over winter and early spring, may be integrated to calculate the chemical change in column ozone: We use the ÔÔearly vortex'' O 3 / CH 4 relationship (Table 3) to compute a proxy for the ÔÔunperturbed'' ozone (O 3 ), expected for March and April in the absence of chemical change. The separation between the observed and the proxy ozone pro®les is a measure of the accumulated chemical ozone loss over winter and early spring (MuÈ ller et al., , 1997a. Therefore, vertically integrating over the dierence between measured and proxy ozone, yields a diagnostic measure of the de®cit in column ozone in March/April in the lower stratosphere (between about 150 and 25 hPa). Note that this calculation is not aected by complications of the seasonal change in total ozone (see earlier) as we focus here solely on the situation in March/April and use CH 4 as a reference frame, thereby taking the eect of diabatic descent in the polar winter latitudes into account. In Table 1, the ozone column loss for each individual HALOE pro®le in the vortex during March and April 1996 is listed. All vortex measurements show severe chemical ozone loss exceeding 100 DU, where typical values range between 120±160 DU. In Table 1  also noted are the column ozone (above 100 hPa) observed by HALOE as a reference and the PV at the 550 K isentropic level (calculated from UKMO meteorological analyses) as an approximate indication of the location of the observation relative to the vortex boundary. Moreover, column ozone and the PV at 550 K is listed in Table 2 for the HALOE vortex observations in November 1995 and in January 1996 for completeness.

GOME observations
The Global Ozone Monitoring Experiment (GOME) is a new passive remote sensing instrument launched by ESA aboard the second European Research Satellite (ERS-2) on 21. April 1995 , and references therein). The total column amount is retrieved from GOME measurements in the wavelength range between 325 and 335 nm of the upwelling radiance from the atmosphere and the extra-terrestrial irradiance.
The technique known as dierential optical absorption spectroscopy (DOAS) is used (Eisinger et al., 1997, and references therein). GOME measurements of total ozone in high northern latitudes in late winter and early spring 1996 show very low values in March 1996 (Fig. 7) in accordance with total ozone derived from SBUV-2 (Newman et al.,  . The low ozone regions in Fig. 7 are related to the location of the polar vortex, which was generally centred o the pole towards northern Europe (Naujokat and Pawson, 1996;Manney et al., 1996b) in early 1996. Indeed, the region of low total ozone in early 1996 is in general colocated with the position of the polar vortex. In Fig. 7 the ozone distribution on March 16, 1996 (see also Fig. 4) is shown; de®ning the edge of the vortex at % 35 PVU (Rummukainen et al., 1994;Rex et al., 1997), which corresponds to the maximum gradient in potential vorticity, clearly, the low total ozone values occur within the con®nes of the polar vortex. This ®nding is consistent with MLS observations which show that the region of very low ozone mixing ratios in the lower stratosphere (at % 465 K) in early 1996 is situated inside the polar vortex  and with HALOE measurements of the ozone vertical pro®le that show very low mixing ratios inside the polar vortex in March and early April 1996 (MuÈ ller et al., 1997a). These observations are in accordance with the notion that substantial chemical ozone loss occurred inside the vortex in early 1996.

Discussion of uncertainties in the analysis of the HALOE observations
The O 3 /CH 4 relation in the incipient vortex in 1995 The reliability of a method employing ozone-tracer relationships to deduce chemical ozone loss, depends on the quality of the initial ozone-tracer relation, which is used as a reference state for the conditions before the onset of chemical ozone destruction. MuÈ ller et al. (1997a) have used an empirical O 3 versus CH 4 relation O 3 2X98CH 4 3 À 11X20CH 4 2 9X52CH 4 2X14 1 (for O 3 and CH 4 in ppmv and valid for 0.5 ppmv`CH 4 1.6 ppmv), derived from HALOE vortex observations in late November 1995 as a reference. One needs to demonstrate that this relation in the vortex, shortly after its formation phase in early November (Coy et al., 1997), is accurately known and is not in¯uenced substantially by out of vortex air. The joint occurrence of low methane mixing ratios, indicating diabatic descent, and high PV, both characteristics of vortex air, is used as a criterion to discriminate vortex from out    However, in MuÈ ller et al. (1996), where this issue has been discussed in some detail, only examples of an application of the methodology for the winters 1991±92 and 1992±93 are shown. Therefore, we present here a typical example of a HALOE observation inside the ÔÔearly vortex '' in 1995±96, namely on November 22, 1995'' in 1995±96, namely on November 22, (uarsday 1533. In Fig. 8, the HALOE observations of CH 4 mixing ratios on November 22, 1995 at 47 N are shown against pressure and longitude. Note that the longitude-height pattern is very similar to a presentation where potential temperature is used as a vertical coordinate (Fig. 2). The air inside the vortex at 118 E clearly stands out as a region of low CH 4 mixing ratios, which are caused by diabatic descent. Indeed, the potential vorticity at this location is enhanced throughout the lower stratosphere ( Fig. 9; Table 4); an observation which corroborates the classi®cation of this air mass as vortex air.
As observed in earlier winters , the O 3 /CH 4 relationship inside the vortex in mid and late November 1995 is clearly distinct from that observed outside (Fig. 10). Further, the outside vortex relations show much more variability than those inside the vortex and the ozone mixing ratios for a given CH 4 mixing ratio are larger outside than inside the vortex (Fig. 10). Moreover, the ozone vertical pro®les measured inside the vortex in November 1995 are distinct from the pro®les measured outside (Fig. 11); on constant potential temperature (H) surfaces, they generally show higher ozone mixing ratios inside the vortex than outside. This observation is consistent with the notion of an ozone increase on H-surfaces caused by the subsi-  Table 2 is shown. Measurements inside the vortex are shown as black circles, outside measurements are shown as blue plus signs. Overplotted in red is the early vortex relation (Eq. 1) Fig. 11. As Fig. 10, but showing vertical pro®les of O 3 mixing ratio (in ppmv) against potential temperature as the vertical coordinate. Measurements inside the vortex are shown as black circles, outside measurements are shown as blue plus signs dence of air inside the vortex (as long as air is transported downwards from below the ozone mixing ratio maximum). Thus, the ÔÔearly vortex'' CH 4 /O 3 relation (Eq. 1) may be considered as characteristic of the incipient vortex.

Dierences in the ozone losses deduced from HALOE data versions 17 and 18
An examination of the O 3 /CH 4 relationships between November/December and March/April for the ®rst ®ve winters of HALOE observations in the Arctic revealed reductions in ozone inside the vortex region, clearly manifested by signi®cantly lower O 3 volume mixing ratios for equal values of the CH 4 mixing ratio . This study was based on HALOE version 17 (V17) data. A reanalysis of the derived ozone loss for those winters based on version 18 (V18) data (MuÈ ller et al., 1997a) yielded larger calculated ozone losses. The reason for the observed increase in the calculated ozone losses when V18 data are used instead of V17, is due to a combination of several smaller changes in the HALOE data, which sum up to the observed eect. Firstly, O 3 mixing ratios have increased from V17 to V18. Therefore, the empirical O 3 /CH 4 relations derived from V18 data show larger ozone mixing ratios (Table 3). However, the increase is stronger at higher altitudes and for larger O 3 mixing ratios, so that the estimated ozone loss increases. (We have also veri®ed this through sensitivity calculations.) Further, CH 4 mixing ratios at lower altitudes (below about 50±70 hPa) have decreased from V17 to V18, which also leads to an increase in the calculated ozone loss. Both eects together are responsible for the fact that the ozone loss calculated previously from V17 data was underestimated.
There is a remaining uncertainty in V18 data, however, regarding CH 4 measurements at sunrise (SR) at lower altitudes, below % 50 hPa, namely that CH 4 mixing ratios are possibly underestimated below this  1993 (uarsday 574); sunset observations (at 65.4 N, 11.2 E and 62.9 N, 105.8 E)  altitude by`10±15%. Since the observations in the incipient vortex in November 1995 (and in all other winters as well), are taken at sunset (SS), this implies a possible overestimate of the calculated loss in the ozone column for SR observations. Sensitivity calculations, performed to derive an estimate of the consequences of this potential problem indicate that 35 DU is an upper limit for the possible overestimate. The SS observations in the vortex in early 1996 (26 March to 15 April), however, are not aected by this uncertainty in V18 data. Moreover, this question is related to an uncertainty in the accuracy of the HALOE sun tracking, a problem which is to a large extent caused by the stratospheric aerosol layer. In the polar vortex, however, the stratospheric aerosol is strongly reduced due to diabatic descent (Browell et al., 1993;Wirth et al., 1994). Indeed, the O 3 /CH 4 relations inside the vortex derived on two occasions where SS and SR measurements overlap inside the Arctic vortex (Figs. 12,13) show good agreement. Furthermore, preliminary V19 data available for SS/SR overlap, for which the problem with the HALOE sun tracking is expected to be solved, indicate that the CH 4 /O 3 relation does not change substantially between V18 and V19 for SR vortex observations.

Comparison of HALOE measurements with other observations in the winter 1995±96
Although for the winter 1995-96 no other direct measurements of the CH 4 /O 3 relation are available, the comparison of the HALOE derived CH 4 /O 3 relation for the early winter 1991, where such measurements are available, showed good agreement . Moreover, the comparison of HALOE ozone pro®les within the Arctic vortex in 1995±96 with observations by Fig. 13. As Fig. 12 (Fig. 14) and ground-based lidar observations (Donovan et al., 1996) (Figs. 15, 16) both in March 1996 and November and December 1995 shows good agreement. Both this agreement as well as the relative homogeneity of the HALOE (MuÈ ller et al., 1997a) and of the lidar ozone pro®les (Donovan et al., 1996) inside the vortex in March and April 1996, if potential temperature is used as the vertical coordinate, indicate that the ozone mixing ratios inside the Arctic polar vortex are fairly homogeneous at this time. The variation among individual ozone pro®les observed by HALOE throughout the lower stratosphere in March and early April 1996 is typically % 0.5 ppm. The clearly stronger variability seen in the total ozone observations (see also Fig. 7) are strongly correlated with lower stratospheric temperatures (Manney et al., 1996b, Fig. 8). This is typically the case for short-term¯uctuations in total ozone, which are caused dynamically by tropospheric disturbances (e.g. Dobson et al., 1929;McKenna et al., 1989).
A further important quantity which has an impact on the conclusions regarding chemical ozone loss, is the descent rate implied by the HALOE CH 4 observations. Unfortunately, to our knowledge, there are no studies of the descent rates in the Arctic vortex for the winter 1995±96. There are problems in comparing descent rates derived for dierent Arctic winters, which show a certain year to year variability  and, even more, descent rates for dierent hemispheres (Manney et al., 1994b;Rosen®eld et al., 1994;Lahoz et al., 1996). Further, estimates of the descent rate in the vortex from the motion of tracer isopleths across isentropes constitute only a lower limit of the actual descent rate (Tuck and Prott, 1998). Figure 17 shows the methane vertical pro®les measured by HALOE on April 7, 1993 at 64 N, which may be directly compared to Fig. 3 of Abrams et al. (1996) and which show very good agreement. Consequently, the implied lower limit of the descent rates (% 0X8 km/month at 20 km) are in accordance as well. Further, we show in Fig. 18 that the morphology of the CH 4 vertical pro®les in the high latitudes in early April 1996 is not too distinct from what is observed by both ATMOS (Abrams et al., 1996) and HALOE (Fig. 17) in 1993.
Moreover, HALOE HCl observations provide information on the extent of chlorine activation. Extremely low HCl mixing ratios were observed in early March 1996 (MuÈ ller et al., 1997a), demonstrating the occurrence of chlorine activation in late winter 1996. This ®nding is consistent with the observation of a The MLS data were selected to ful®l the following requirements: (1) latitude between 45 and 60 N: (2) longitude between 340 and 360 or 0 and 60 E: (3) UKMO PV at 465 K greater than 2.5 10 À5 K m 2 (kg s) À1 , so that they are comparable with the HALOE measurements. The two dashed lines show the range of the MLS observations, their mean value is indicated by the dotted line. The comparable values for the column ozone above 100 hPa are 197 DU and 195 DU for HALOE and MLS respectively  (Donovan et al., 1996) between November 29, and December 29, 1995. The grey area indicates the variation of the lidar pro®les by AE one standard deviation around the mean pro®le  (Donovan et al., 1996) between March 10 andMarch 22, 1996 large scale, severe decrease in gas-phase HNO 3 by MLS  and the simultaneous observation of very large aerosol extinction by HAL-OE (Fig. 19) on March 3, 1996 (in high northern latitudes around 0 W), both indicative of extensive PSC formation. Because the deactivation of chlorine in the Arctic requires more than 2±3 weeks (MuÈ ller et al., 1994;Douglass et al., 1995;Santee et al., 1996b) ozone loss must have continued during March 1996; a conclusion which is in accordance with the ozone loss rates deduced from the Match ozonesonde-campaign in 1995±96 (Rex et al., 1997).

Conclusions
We have examined simultaneous HALOE observations of O 3 with a chemically inert tracer (CH 4 ), inside the Arctic vortex in the winter and spring of 1995±96 to detect chemical ozone depletion in the lower strato- Fig. 19. HALOE measurements of aerosol extinction (5.26 lm) in km À1 at 52 N on March 3, 1996 sphere, despite strong variations caused by dynamical processes. Severe chemical ozone column loss of 120± 160 Dobson units is derived for this period (see also, MuÈ ller et al., 1997a). Here, we have considered further aspects of the HALOE observations in the Arctic over this period. It was demonstrated that the November 1995 HALOE observations may be used as a reference for the chemically unperturbed conditions inside the incipient vortex. Further, it was shown that the reason for the observed increase in the calculated column ozone losses when V18 data are used instead of V17 (MuÈ ller et al., , 1997a, could be understood as the eect of a combination of several smaller changes. Finally, it was concluded that there is good agreement between the relevant HALOE measurements in the Arctic in winter and spring 1995±96 with observations by other groundbased and satellite instruments.