Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Calculation of chemical ozone loss in the Arctic winter 1996–1997 using ozone-tracer correlations: Comparison of Improved Limb Atmospheric Spectrometer (ILAS) and Halogen Occultation Experiment (HALOE) results
Forschungszentrum Jülich, Institute of Stratospheric Research (ICG-I), Jülich, Germany
 The ozone-tracer correlation method is used to deduce the stratospheric ozone loss in the Arctic winter 1996–1997. Improvements of the technique are applied, such as a new calculation of the vortex edge [Nash et al., 1996] and an improved early vortex reference function. Winter 1996–1997 is characterized by a late formation and an unusually long lifetime of the polar vortex. Remnants of vortex air were found until May. Chemical ozone losses deduced from two satellite data sets, namely Improved Limb Atmospheric Spectrometer (ILAS) and Halogen Occultation Experiment (HALOE), are discussed. The ILAS observations allow a detailed analysis of the temporal evolution of the ozone-tracer correlation inside the polar vortex and, in particular, of the development of the early vortex. For November and December 1996, it is shown that horizontal mixing still influences the ozone-tracer relation. Significant PSC related chemical ozone loss occurred beginning at mid-February, and the averaged column ozone loss is increasing toward the middle of May. From April onwards, ozone profiles in the vortex became more uniform. The decrease of ozone in the vortex remnants in April and May occurred due to chemistry. HALOE observations are available for March to May 1997. In the period 4–16 March 1997, the calculated ozone loss deduced from HALOE and ILAS is in good agreement. The average of the result from the two instruments is 15 ± 7 Dobson units (DU) inside the vortex core, in the altitude range of 450–550 K. At the end of March, a discrepancy between HALOE and ILAS ozone loss arises due to a significant difference (0.6 ppmv) between the two data sets in the relatively low ozone minimum measured at 475 K. Nonetheless, both data sets consistently show an inhomogeneity in ozone loss inside the vortex core at the end of March. The vortex is separated in two parts, one with a large ozone loss (HALOE 40–45 DU, ILAS 30–35 DU) and one with a moderate ozone loss (HALOE 15–30 DU, ILAS 5–25 DU) for 450–550 K. The ozone loss from HALOE in 380–550 K at that time was calculated to be 90–110 DU for the large ozone loss and 20–80 DU for the moderate ozone loss. The vortex average of column ozone loss from HALOE inside the vortex core at the end of March is 61 ± 20 DU, which is an increase of about 20% compared to the earlier study by Müller et al. [1997b] brought about by the improvement of the technique.
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.
 Substantial loss of stratospheric ozone in high latitudes has been observed in many Arctic winters since the late 1980s. At that time experimental activities in the Arctic were strongly expanded with the aim of assessing the extent of ozone loss during the boreal spring and to investigate whether this loss can reach Antarctic values. In the Antarctic polar vortex a dramatic ozone loss has occurred during spring since the late 1970s [e.g., WMO, 1998; Jones and Shanklin, 1995].
 Results are presented on the basis of observations obtained by two solar occultation satellite instruments: ILAS (Improved Limb Atmospheric Spectrometer) aboard the ADEOS satellite (Advanced Earth Observing Satellite) and HALOE (Halogen Occultation Experiment) aboard the UARS, the Upper Atmosphere Research satellite [Russell et al., 1993]. An 8-month record of data for trace species from November 1996 to June 1997 was obtained with ILAS [Suzuki et al., 1995; Sasano et al., 1999]. The instrument took measurements about fourteen times per day in the high latitude region of both hemispheres. Vertical profiles of, e.g., ozone, nitrous oxide, and methane were measured in the range from ≈10 km or cloud top up to 70 km. Owing to the good spatial and temporal coverage of ILAS data inside the polar vortex it is possible for the first time to obtain a complete winter season record over the entire Arctic vortex using the tracer correlation approach (Figure 1, green symbols).
 Previous studies suffer from incomplete temporal and spatial data coverage [e.g., Woyke et al., 1999; Müller et al., 1996, 1997a, 1997b, 2001]. The HALOE instrument makes measurements fifteen times per day at sunrise and sunset along two latitude belts that move between 80°N and 80°S in about 45 days (Figure 1, blue symbols). In winter 1996–1997, HALOE observations are available inside the Arctic early vortex in November and December 1996 and inside the vortex in spring (March, April and May 1997). For spring a comparison of the HALOE and ILAS observations is possible because the measurements were made at very nearby locations.
 In winter 1996–1997, the polar vortex formed relatively late. Before it was fully established at the end of December, horizontal mixing occurred from outside the vortex and the minimum temperature remained above 195 K. Afterward the vortex grew quickly and remained strong and symmetrically centered around the pole until the end of April [Coy et al., 1997]. Record low temperatures over a 17-year period were reached in March and April [Coy et al., 1997]. In this work, the ozone-tracer relation for chemically unperturbed conditions was derived from ILAS observations for the period 1–15 January. We show that a significant portion of ozone was destroyed during February and March in the largely sunlit vortex enhanced by the significant denitrification that occurred in mid-February [Kondo et al., 2000]. Until the end of May remnants of vortex air with very low ozone values were observed. The calculated losses from ILAS and HALOE observations inside the vortex core are in agreement considering profiles with moderate ozone mixing ratios. At the end of March, the ILAS ozone loss is significantly smaller than calculated from HALOE observations, considering profiles with a relatively low ozone minimum at 475 K. Nonetheless, HALOE and ILAS observations consistently indicate an inhomogeneity of the ozone loss inside the vortex in March. The vortex air is divided into two parts, one with moderate ozone loss and one with strong ozone loss. This picture is consistent with previous model studies [McKenna et al., 2002].
2. Ozone-Tracer Correlation Technique
 The ozone-tracer correlation technique (described in detail elsewhere [e.g., Proffitt et al., 1993; Müller et al., 2001]) is used to obtain a quantitative estimate of the chemical ozone loss inside the polar vortex. The technique is based on the condition that the relation between long-lived tracers remains constant inside the isolated polar vortex [Plumb and Ko, 1992; Müller et al., 2001]; it is shown later in this paper that this is the case in the winter 1996–1997. The lifetime of O3 is greater than ≈100 days in winter at high latitudes [e.g., Proffitt et al., 1992; WMO, 1990]. Thus, an empirical relation between ozone and other long-lived tracers like CH4, HF, and N2O can be derived for the “early vortex”. This serves as a reference for chemically unperturbed conditions. Deviations from this early winter reference function occur due to ozone destruction by halogen-catalyzed reactions [Proffitt et al., 1993; Müller et al., 1996, 2001; Salawitch et al., 2002]. In the course of the spring, ozone depletion increases due to increasing solar exposure of the polar vortex (in the presence of sufficiently large mixing ratios of active chlorine) and is sometimes enhanced by significant denitrification.
 The chemical ozone loss ΔO3 is quantified as the difference of the actually measured ozone and the corresponding ozone proxy Ô3. Ô3 is the expected ozone mixing ratio in the absence of chemical processing. The observed profile of the long-lived tracer is converted to Ô3 with the (“early vortex”) ozone-tracer reference function. The vertically integrated ΔO3 is an estimate of the column ozone loss at these locations. The column ozone loss is converted into Dobson units (DU): one Dobson unit is 2.69*1016 molecules/cm2. The uncertainty of ΔO3 and thus of the column ozone loss is deduced from the uncertainty (σ) of the “early vortex” reference relation.
 In the present study, the ozone loss in the Arctic winter 1996–1997 was calculated from satellite observations obtained by the ILAS and HALOE instruments. Both instruments measured ozone profiles and different long-lived tracers (CH4 and N2O by ILAS, and CH4 and HF by HALOE). A criterion derived by Nash et al.  was used to decide whether profiles are inside or outside the entire vortex and the vortex core. In the present case, the edge of the vortex was defined where the gradient of the potential vorticity (PV) reaches its maximum constrained by the maximum velocity of the wind jet. Also a vortex boundary region was defined regarding the second derivative of a certain PV profile following Nash et al. . The criterion was applied to the UKMO meteorological analyses at 1200 UTC each day and at three different altitudes (potential temperature Θ: 475, 550, 650 K). We used trajectory calculations to determine the location of HALOE and ILAS profiles at 1200 UTC (Figure 3). The profiles of HALOE and ILAS were considered to be within the entire vortex if the PV value at their position at 1200 UTC is greater than the PV value of the defined vortex edge at the three theta levels. If the PV values for the profiles were in addition greater than the PV values of the defined (inner) vortex boundary region, they were considered to be inside the vortex core. We define the region between the vortex edge and the vortex core as the “outer vortex”.
 Ozone loss is calculated for a maximum altitude range of 380–550 K from HALOE observations because within this range the empirical ozone-tracer reference relations are valid. At lower altitudes toward the bottom of the vortex, the vortex is less strong and there is mixing with air from outside the vortex. Tracer values (here N2O) from ILAS observations are greatly scattered at altitudes lower than 450 K, thus the ILAS ozone loss is calculated for an altitude range of 450–550 K. Altogether, for the altitude range of 450–500 K the strongest ozone loss was observed for this winter.
3. Comparison of ILAS and HALOE Ozone Observations
 A comparison of ozone measured by ILAS (Version 5.2), HALOE (Version 19) and ozone sondes is performed to assess the differences in the amount of the ozone loss calculated from ILAS and HALOE observations. HALOE measurements inside the outer vortex and in the vortex core are available from 4 March to 4 April, so that ILAS and HALOE are directly comparable only in this time interval. With respect to ILAS and HALOE mean ozone profiles over this time period (Figure 2) the ozone values measured by ILAS are systematically greater than those measured by HALOE.
 The mean deviation between HALOE and ILAS ozone profiles in 380–550 K is 0.25 ppmv for profiles inside the outer vortex and 0.16 ppmv for profiles inside the vortex core. This result is in agreement with the validation analysis of Sugita et al. , who found that ILAS ozone mixing ratios are on average about 8% greater than correlative measurements obtained by HALOE for the altitude range of interest from 24 March to 2 April. Although there are no HALOE vortex profiles over the whole winter period we assume that the mean deviation between HALOE and ILAS does not change significantly inside the outer vortex and the vortex core in the course of the winter and spring.
 For the period 25 to 31 March, the ILAS and HALOE measurements are obtained at adjacent positions and times inside the polar vortex (Figures 1 and 3). Thus the corresponding ILAS and HALOE profiles most likely observed the same air masses for this time period.
 For example on 26 and 28 March (Figure 4) some HALOE profiles (solid lines) are measured at nearly the same geographical longitude and latitude as some of the profiles observed by ILAS (dashed lines). Both HALOE and ILAS ozone profiles were observed with two different characteristics in the height range of 450–550 K. On the one hand, there were profiles showing relatively low ozone values (minimum values, HALOE: 0.9–1.1 ppmv, ILAS: 1.6–1.8 ppmv) and, on the other hand, there were profiles with moderate ozone values (above 2 ppmv at ≈475 K). This tendency can also be observed for the ozone values at the 475 K level (Figure 3). Ozone values were relatively low if the position of the observed profile was inside a region with high PV values, i.e., inside the vortex core. Profiles outside that region show much greater ozone values as observed both by ILAS and HALOE.
 At the end of March, HALOE measurements show much lower minimum ozone values than found by ILAS. Considering the minimum of the low ozone values, the difference between HALOE and ILAS measurements (about 0.6 ppmv) is much greater than the systematic deviation between HALOE and ILAS ozone profiles in general. Comparing HALOE and ILAS ozone loss, we have to take into account the fact that there will be substantial differences in profiles with high ozone loss, while the moderate ozone loss is comparable.
 A comparison between the satellite data and ozone sonde measurements is performed to assess the observed differences between the satellite data sets. The moderate ozone values measured by HALOE and ILAS instruments from 26 March to 2 April are in the same range. As described above, ILAS ozone mixing ratios are slightly greater (Figure 5). A great amount of data is available from ozone sondes inside the polar vortex in March and April, displaying similar ozone mixing ratios (not shown here).
 We specifically investigated the ozone profiles obtained by ILAS and HALOE, where the minimum ozone mixing ratios in 450–500 K are relatively low. At the end of March and the beginning of April there are about 6 ozone sonde measurements with minimum ozone values below 1.5 ppmv. Three of these profiles (Figure 5) show a relatively low minimum (about 1 ppmv) comparable to the HALOE minimum values.
 The sonde observations are made at ≈10° north of the position of HALOE and ILAS observations and are therefore not directly comparable. Nonetheless, the Ny-Ålesund measurements are made at the same equivalent latitude (corresponding to 50 to 55 PVU at 475 K) as the HALOE and ILAS observations at 69°N and 110–140°E (Figures 4 and 5). However, the low ozone values are measured by the sondes over a rather narrow layer. Such an ozone profile will be smoothed in the observations of a satellite instrument with the characteristics of HALOE or ILAS. The resulting profile depends on the specifics of the instrument characteristics and on the smoothing algorithm. When the regions of substantially depleted and strongly layered ozone in March are observed by HALOE, the HALOE ozone is biased low below the ozone minimum compared to the ozone sonde measurements, while the minimum of the ozone profile is in agreement. On the other hand, ILAS tends to be biased high around the minimum of the ozone sonde profiles.
 The tracers CH4 and N2O measured by ILAS are in principle suitable for the ozone-tracer correlation method, but currently only N2O provides sufficient data quality. CH4 was not used because the ILAS Version 5.2 CH4 data still differ from the validation data with the magnitude of the discrepancy depending on season [Kanzawa et al., 2002].
4. Evolution of Ozone-Tracer Correlation in the Early Vortex
 For the application of the ozone-tracer technique it is necessary to derive an early winter reference function from which the ozone loss is calculated. This reference function represents an ozone-tracer relation under chemically unperturbed conditions inside the isolated vortex. Deviations from this relation are presumed to be caused only by chemical reactions. It is necessary to ensure that the ozone-tracer relation will not change significantly due to horizontal isentropic mixing processes so that the vortex must be fully established when the early winter reference function is derived. Utilizing the large number of profiles measured by ILAS in November and December 1996 in high latitudes, it is possible to describe the evolution of the ozone-tracer relation inside and at the edge of the early vortex with N2O as the long-lived tracer (Figure 6). The different colors of the profiles of the ILAS observations denote different parts of the vortex sorted by the equivalent latitude at three theta levels: 475, 550 and 675 K [e.g., Lary et al., 1995]. Profiles with an equivalent latitude lower than 60°N are assumed to be outside the vortex (black lines). Red lines stand for profiles located at the edge (equivalent latitude <70°N), blue profiles are measured inside the vortex toward the edge (equivalent latitude <80°N), and green profiles are measured inside the vortex toward the core (equivalent latitude >80°N). All other profiles outside the vortex with an equivalent latitude lower than 50°N are indicated as black symbols.
 From November to January there is a separation of profiles inside (colored profiles) and outside the vortex (black profiles) in Figure 6 (all panels). The relation of profiles inside and at the edge of the early vortex changed over the course of this period. At the beginning of November when the vortex began to form, there is a large scatter of the ozone mixing ratio from 2 to 4.5 ppmv ozone at higher altitudes (N2O < 130 ppbv) see Figure 6, panels a and b. Mixing across the vortex edge was strong enough to change the relation in the vortex core in November and December, so the very low ozone values of 2 ppmv vanished by January. During the first half of December the vortex was strongly disturbed by air from outside due to mixing processes. Blue profiles inside the vortex show a stronger scatter in December than in November (Figure 6, panels c and d). At the end of December, the vortex became stronger and more isolated. In January and probably during the whole springtime, there was still some mixing of air from outside the vortex to the edge of the vortex (Figure 6, panel d, red lines). However, inside the vortex toward the core (Figure 6, panel d, green lines) the ozone-tracer relation became more uniform and was no longer influenced by these mixing processes (such a result was likewise obtained for winter 1999–2000 by Ray et al. ). The relation did not significantly change during January (Figure 7, panels c–d). Thus the correlation function for profiles inside the vortex (green lines in Figure 6, bottom panel) cannot be calculated from observations obtained before the beginning of January. Should a small amount of isentropic mixing of outside air occur, changes of the vortex ozone-tracer relation in spring can only lead to increased ozone mixing ratios relative to a constant tracer mixing ratio. This is true although the ozone mixing ratios on some potential temperature surfaces (e.g., 450 K) outside the vortex are significantly lower than inside. Thus, isentropic mixing in of outside air affects the ozone-tracer relationship differently than the ozone-potential temperature relationship. Following isentropic lines from inside to outside the vortex (not shown), mixing occurs from low ozone and low tracer mixing ratios to high ozone and high tracer mixing ratios. For example, considering ozone-tracer relations in Figure 6, ozone mixing ratios from profiles obtained at high equivalent latitudes are less than ozone values from profiles obtained at lower equivalent latitudes over the whole tracer range. Therefore, if the horizontal influence of mixing from the vortex boundary is neglected, the calculated ozone loss can only be underestimated [see also Müller et al., 2001].
5. ILAS Results
 The temporal evolution of the destruction of ozone inside the polar vortex can be followed in a chronological way by analyzing the ILAS O3/N2O relation (Figure 7). An empirical early winter reference function for winter 1996–1997 was fitted to the ILAS profiles inside the early polar vortex for the period 1–8 January (see section 4).
 Using N2O as a long-lived tracer (mixing ratios in ppbv) and O3 (mixing ratios in ppmv) this early winter reference function (valid for range 20 ppbv < N2O < 250 ppbv) is:
with an uncertainty of σ = 0.28 ppmv (Figure 7). The function is shown in all panels of Figure 7 with the corresponding uncertainty indicated by the area shaded in gray.
 The changes in the ozone-tracer relation that are due to isentropic mixing processes caused by an incomplete isolation of the early vortex (panels a and b) have been discussed above. In January the O3/N2O relation inside the vortex is within the range of uncertainty of the early winter reference function (equation (1); Figure 7, panels c, d, and e). Deviations from the reference function due to the chemical destruction of ozone by active chlorine compounds are expected to increase with increasing sunlight duration. This can be observed in the ILAS data. At the beginning of February, profiles are scattered a little below the reference function above ≈120 ppbv N2O. This deviation from the reference function above 450 K potential temperature indicates the beginning of chemical ozone loss (Figure 7, panel e). From the middle of February onwards, deviations of the early winter reference function become clearly noticeable and become more and more pronounced in the course of the spring (Figure 7, panels f to m).
Table 1 summarizes the average of the derived column ozone loss between 450 and 550 K for different time intervals and for different parts of the vortex. The column ozone loss is calculated from the end of February to the middle of May. The technique is not well suited to quantify very small amounts of ozone loss at the beginning of February, because the range of uncertainty is larger than the calculated ozone loss. The standard deviation of the average ozone loss (standard deviations in brackets in Table 1) is shown for each part of the vortex. The error was estimated from the uncertainty of the reference function as described above.
Table 1. Ozone Loss in Dobson Units in 450–550 K Calculated From ILAS Data
Mean Vortex (Standard Deviation)
Mean Core (Standard Deviation)
Mean Outer Vortex (Standard Deviation)
7.3 ± 5.6 (6.9)
9.4 ± 5.3 (4.9)
5.8 ± 5.8 (7.9)
17.4 ± 6.3
7.4 ± 5.7 (11.5)
13.9 ± 5.5 (8.6)
0.7 ± 6.0 (10.3)
26.8 ± 6.1
7.6 ± 5.9 (12.5)
15.5 ± 5.6 (8.9)
0.7 ± 6.1 (11.0)
35.4 ± 6.0
11.9 ± 5.9 (11.4)
19.9 ± 5.5 (8.3)
4.6 ± 6.2 (8.5)
35.3 ± 4.6
13.3 ± 6.0 (9.7)
18.9 ± 5.8 (8.8)
8.8 ± 6.2 (8.0)
39.1 ± 6.2
24.4 ± 6.5 (12.3)
32.4 ± 6.4 (5.8)
16.7 ± 6.6 (12.0)
41.4 ± 6.4
32.6 ± 6.6 (7.0)
33.8 ± 6.5 (6.7)
29.9 ± 6.9 (7.4)
45.2 ± 5.4
 The most rapid change in the N2O/O3 relation occurs between 22 February and 20 March, when the temperatures inside the vortex were still very low (Figure 7, panels i to k) and sufficiently large mixing ratios of active chlorine were present. At the end of March HCl begins to recover from the extremely low activated levels [Müller et al., 1997b]. Therefore, any further chemical ozone loss is likely due to NOx chemistry later in the season [Grooß et al., 1998; Hansen and Chipperfield, 1999]. From 22 February to 20 March, the maximum ozone loss increased by about 18 DU (from 17 ± 6 DU up to 35 ± 6 DU) for 450–550 K (Table 1, fifth column). Both the mean ozone loss over the entire vortex and the vortex core increased over this period. Throughout March and the beginning of April, the standard deviation of average ozone loss inside the vortex core (above 8 DU) is greater than for February and for the end of April and May. In this time period the deviation of ozone-tracer profiles from the reference function are very inhomogeneous, especially between 21 March and 15 April 1997 (Figure 7, panels l and m). A separation between some profiles with relatively low ozone mixing ratios and many more profiles with moderate ozone values are becomes obvious.
 From the middle of April to May, the deviation from the reference function becomes more and more uniform, (Figure 7, panels n and o). In May only low ozone values are measured inside the vortex core. The mean ozone loss inside the entire vortex and the vortex core increases further to 32 ± 6 DU in April and 34 ± 7 DU in May (inside the vortex core) with a maximum ozone loss of 45 ± 5 DU (Table 1, third column). In 1997, the polar vortex was still very strong until the end of April. An additional ozone loss occurred due to NOx chemistry. Also, the vortex becomes more and more compact until April and mixing within the vortex air masses presumably leads to uniform ozone mixing ratio profiles throughout the vortex. Inside the outer part of the vortex the mean deduced ozone loss is almost zero in March (Table 1), due to the influence of the mixing of air from outside the vortex inside the outer vortex along isentropic lines.
 For the period from 18 March to 4 April 1997, the mean ozone loss in the vortex core is calculated as 19 ± 6 DU with a maximum of 39 ± 6 DU for the altitude range of 450–550 K (Table 2, third column). The corresponding percentage values are 26% ± 6% with a maximum of 47% ± 4%. (The percentage of ozone loss is calculated as: difference profile Δ O3/ozone proxy Ô3 * 100). Because ILAS and HALOE profiles are located at very nearby positions and a comparison between these measurements is possible, this time period is further discussed below (see section 7).
Table 2. Ozone Loss in Dobson Units in 450–550 K
HALOE (HF) (Standard Deviation)
HALOE (CH4) (Standard Deviation)
ILAS (N2O) (Standard Deviation)
4–16 March 1997
14.4 ± 7.3 (6.7)
12.6 ± 7.0 (8.4)
8.3 ± 5.7 (12.2)
15.9 ± 7.0 (7.1)
14.4 ± 6.7 (8.3)
15.1 ± 5.5 (10.3)
Mean Outer Vortex
13.1 ± 7.6 (6.2)
11.2 ± 7.3 (8.7)
1.2 ± 5.9 (9.8)
25.9 ± 7.3
26.3 ± 6.2
35.4 ± 6.0
18 March to 4 April 1997
18.7 ± 7.2 (15.2)
18.8 ± 6.9 (15.9)
11.3 ± 6.0 (11.3)
27.8 ± 6.9 (10.0)
28.1 ± 6.6 (9.6)
18.8 ± 5.6 (8.3)
Mean Outer Vortex
8.2 ± 7.5 (13.3)
7.8 ± 7.3 (14.7)
5.3 ± 6.2 (9.0)
46.1 ± 6.7
48.3 ± 6.3
39.1 ± 6.2
 In the frequency distributions (Figures 8 and 9) the calculated ozone loss in the vortex core is shown as a solid line, and the calculated ozone loss inside the outer vortex is indicated as a dashed line. In this time period, a separation of profiles inside the vortex core with large ozone loss (30–40 DU) from profiles with moderate ozone loss (5–25 DU) is becoming obvious (Figure 9, panel a, solid line). In the outer vortex, only profiles with moderate or very little ozone loss were observed.
6. HALOE Results
 The calculation of ozone loss from HALOE observations is possible for March, April, and May 1997, when measurements inside the polar vortex are available. CH4 and HF are long-lived tracers measured by HALOE that are both suited for the ozone-tracer method. Further, HALOE took measurements in the high northern latitudes up to a maximum of 49°N in November and 48°N in December 1996. Müller et al. [1997b] derived an ozone tracer reference relation from HALOE data for November 1996 using HF as the long-lived tracer. However, since the vortex was not fully established in November 1996 (Figure 10, black asterisks), this relation is not well suited as a reference relation. A better choice would be to employ the HALOE observations in high northern latitudes in December, but these measurements are located very much toward the edge of the polar vortex. A reference relation deduced from those observations would thus be influenced by air from outside the vortex (Figure 10, red asterisks).
 Thus, we derive here a correlation function for the long-lived tracer CH4 and HF from the ILAS reference function for O3/N2O (equation (1)) using a N2O/CH4 and a CH4/HF relation. Engel et al.  report a linear relation for N2O/CH4 from whole air sampler measurements at different seasons and mid and high northern latitudes between 1988 and 1992:
valid for the range 0.65 ppmv < CH4 < 1.6 ppmv. Ozone mixing ratios obtained by ILAS are systematically larger than HALOE values (0.16 ppmv inside the vortex core and 0.25 inside the outer vortex) see above. Therefore, a shift of −0.21 ppmv ±0.05 ppmv in 3 is applied to the derived early winter reference function, for the HALOE calculations.
 The O3/CH4 relation derived from equation (1) in this way is shown in Figure 10 (bottom panel, black line):
The indicated range of uncertainty, σ = 0.28, (shaded in gray) corresponds to the uncertainty of the reference function, σ = 0.28 as calculated from ILAS early vortex profiles and the uncertainty of the mean difference between HALOE and ILAS ozone mixing ratios, σ = 0.05 ppmv. Vertical profiles of tracer and ozone mixing ratios (by balloon-born cryogenic whole air samplers and concurrent ozone sondes from the University of Frankfurt [Schmidt et al., 1987]) from 11 February 1997 [Kondo et al., 1999] were measured above Kiruna, Sweden, inside the outer vortex. No significant ozone loss took place up to that time inside the outer vortex (Figure 10, bottom panel). This assumption was confirmed by the ILAS observations as discussed above (Figure 7).
 To obtain a reference function for HF as the long-lived tracer, the O3/CH4 relation (equation (3)) is converted to an O3/HF relation (Figure 10, top panel, black line). For this purpose the following CH4/HF relation was calculated from HALOE observations inside the entire polar vortex in March 1997:
in (ppmv/ppbv) valid for the 0.1 ppbv < HF < 1.5 ppbv range with an uncertainty of σ = 0.079 ppmv. The uncertainty σ = 0.079 ppmv was taken into account to calculate the uncertainty of the O3/HF relation.
 HALOE profiles inside the vortex in May and measurements outside the vortex close to the edge in November and December (when no vortex profiles were observed by HALOE) confirm this relation (Figure 11). A deviation of more than 0.2 ppmv in the CH4 mixing ratios from the relation occurs in the period from 4 to 16 March 1997 (Figure 11, e.g., green dots). These profiles are characterized by large beta angles and were not used in the further analysis. Deviations from the derived relations O3/HF and O3/CH4 are noticeable for early March (Figure 10, blue and green diamonds) and are stronger at the end of March (Figure 10, magenta and violet diamonds). The strongest reduction of ozone appears at the end of March (18–31), when ozone mixing ratios as low as 0.9 ppmv (at 0.8–0.9 ppbv HF) were observed.
Figure 12 shows the measured ozone profiles, the proxy for the ozone Ô3 and ΔO3 and the difference of these profiles for the period 18–31 March 1997 over the altitude range of 350–550 K. The ozone loss ΔO3 at the height of 475 K observed for individual profiles ranges from very small deviations up to strong losses with a maximum loss of up to 2.6 ppmv. The derived ozone loss (ΔO3) from January to the end of March 1997 at 475 K can be grouped into three types, those with weak ozone loss (about 0.3 ppmv), those with moderate loss (0.9–1.4 ppmv) and those with strong ozone loss (2.4–2.6 ppmv). At lower altitudes, below 420 K (Figure 12) the scatter of the profiles is stronger, especially when HF is used as the long-lived tracer.
 The mean ozone loss inside the entire vortex for the period 18 March to 4 April 1997 is determined as 44 ± 21 DU using HF and 45 ± 20 DU using CH4 as tracer in the height range 380–550 K (Table 3). The reported uncertainty (Tables 2 and 3) is derived from the uncertainty of the early vortex reference relation. The calculated ozone loss should be independent of the utilized tracer, and indeed the values deduced for CH4 and HF as tracers agree very well within the uncertainty introduced by the uncertainty of the reference function. The ozone loss reported here was calculated as the average of the results from CH4 and HF: entire vortex mean = 44 ± 21 DU and vortex core mean = 61 ± 20 DU. The reported uncertainty includes the variability introduced by employing two different tracers. For the height range of 450–550 K, the ozone loss calculated using the two long-lived tracers is nearly identical (Table 3). Indeed, in this altitude range the vortex is more stable and thus less permeable than above and below so that the variability introduced by mixing of air from outside the vortex is expected to be small. In this altitude range the mean of the ozone loss inside the vortex core for both tracers is about 34% ±8%, corresponding to 28 ± 7 DU, and the maximum loss is 58% ±4% (47 ± 7 DU).
Table 3. Ozone Loss in Dobson Units in 380–550 K
HALOE (HF) (Standard Deviation)
HALOE (CH4) (Standard Deviation)
4–16 March 1997
29.5 ± 21.3 (15.1)
23.0 ± 20.5 (23.5)
29.6 ± 20.8 (18.2)
26.1 ± 20.0 (25.0)
Mean Outer Vortex
29.3 ± 21.6 (13.2)
20.5 ± 20.8 (23.3)
52.7 ± 23.3
58.6 ± 19.7
18 March to 4 April 1997
43.7 ± 20.7 (28.1)
44.5 ± 19.9 (33.4)
59.0 ± 20.2 (19.8)
62.6 ± 19.4 (22.0)
Mean Outer Vortex
25.7 ± 21.4 (25.8)
23.4 ± 20.6 (32.2)
101.9 ± 17.9
112.2 ± 17.8
 There are two groups of losses in the frequency distribution of the calculated ozone loss of HALOE observations (see Figures 8, bottom panels, and 9), a large ozone loss of 90–110 DU was determined mainly in the vortex core for 380–550 K and 40–45 DU for 450–550 K. Smaller losses of between 20 and 80 DU for 380–550 K and 15–30 DU for 450–550 K were calculated for the entire vortex.
 In May 1997 three profiles were found in vortex remnants with very low ozone values (Figure 13). The ozone loss in the vortex remnants was calculated as 69 ± 23 DU in the period 12–13 May 1997, with a maximum of 98 ± 22 DU in 380–550 K, again the average of the result from CH4 and HF. The maximum column ozone loss is not as large as calculated in March, but the ozone-tracer relations are much more compact and the mean ozone loss is about 8 DU larger compared to March. In the altitude range of 450–550 K, the mean ozone loss was calculated as 38 ± 8 DU, which is an increase of 10 DU compared to March.
 Ozone losses from HALOE and ILAS measurements are comparable in March 1997, especially at the end of March, where the instruments made measurements at very close positions. The calculations were based on the same tracer-ozone correlation function derived from ILAS data for January 1997. The systematic difference between HALOE and ILAS mean ozone mixing ratios was considered. A comparison is made for the column ozone loss calculated over an altitude range of 450–550 K. Over the period 4–16 March 1997, the ozone loss of HALOE and ILAS inside the vortex core is in agreement within a range of 2 DU. The average of the result from HALOE and ILAS is 15 ± 7 Dobson units (DU). The maximum ozone loss of ILAS is 9 DU higher than that calculated from HALOE. At the end of March, significantly larger loss values were obtained from HALOE than from ILAS data for the vortex core. The difference between the HALOE and ILAS mean ozone loss is estimated to be about 8 DU (Table 2).
 HALOE and ILAS ozone values are in agreement, excepting a large difference between the two instruments for the lowest ozone values measured at the end of March (see above). In the first half of March, moderate ozone values were measured consistently by HALOE and ILAS and the results are in good agreement. The discrepancy between HALOE and ILAS at the end of March arises owing to the large difference of relatively low ozone values.
 The calculated ozone losses of both sets of satellite data consistently show an inhomogeneity of ozone loss inside the entire vortex. The mean ozone loss in the vortex core differs significantly from the mean ozone loss in the outer vortex, especially in ILAS results (Table 2). Moreover, in the frequency distributions a separation within the vortex core into two distinct parts was found. Large values of ozone loss occur mainly within the vortex core, while smaller losses were found in the vortex core and in the outer vortex. Model calculations of ozone depletion patterns within the vortex using the Chemical Lagrangian Model of Stratosphere (CLaMS) show a good agreement with this pattern of ozone loss derived from HALOE and ILAS observations [McKenna et al., 2002]. For the end of March, the model simulated a filament of vortex air out of the core going through the outer vortex [McKenna et al., 2002]. This air mass is characterized by lower ozone mixing ratios than air masses observed in the outer vortex but not as low as the in the cold parts of the vortex core. Thus the outer vortex is also divided into a part with stronger ozone loss and a part with moderate ozone loss (Figure 9). Greater ozone loss is caused by lower temperatures and stronger denitrification within this region [McKenna et al., 2002]. Ozone loss in March 1996–1997 is spatially much more inhomogeneous than that observed in previous winters [Müller et al., 1996, 1997a]. The range of moderate loss values for 450–550 K inside the vortex core is 15–30 DU for HALOE and 5–25 DU for ILAS and for the strong ozone loss about 40–45 DU for HALOE and 30–35 DU for ILAS.
 In a previous study of the same winter 1996–1997, only HALOE observations were available to derive a reference correlation function [Müller et al., 1997b]. In this winter the vortex was formed very late compared to other winters. Müller et al. [1997b] deduced a reference function from HALOE observations measured in November as described above. Therefore, the maximum column ozone loss inside the vortex was underestimated by up to ≈20% for the altitude range of 380–550 K and the local loss at 475 K was underestimated by 15% (0.4 ppbv). Goutail et al.  used the 3D chemical transport model REPROBUS (Reactive Processes Ruling the Ozone Budget in the Stratosphere) [Lefèvre et al., 1998] to estimate the loss of column ozone in the stratosphere in winter 1996–1997. Chemical ozone loss is estimated as the difference between ozone measurements (ground-based and balloon-borne measurements (SAOZ) and satellite measurements (POAM)) and simulated passive ozone [Deniel et al., 1998; Goutail et al., 1999]. The calculated column ozone loss averaged over the vortex in the winter 1996–1997 is 110 DU. This result from SAOZ agrees with the greatest ozone loss calculated by HALOE inside the vortex core, but is much higher than the vortex average.
 Accumulated ozone loss in the vortex in 1996–1997 was also estimated by the Match technique [Rex et al., 1998; Schulz et al., 2000] and by Knudsen et al.  using the vortex average approach. In this approach, the temporal development of the vortex average of ozone measurements are considered and corrected for dynamical changes (due to diabatic descent and mixing across the vortex edge from model calculations) to deduce chemical ozone change. In January, already a small amount of ozone loss was calculated by Match [Schulz et al., 2000]. This is in accordance with a very small deviation from the reference function at the beginning of January. However, no significant chemical ozone loss has taken place. In the time period between the end of January and the end of March the accumulated ozone loss was calculated in a range of 1.1 ppmv by the Match technique and 0.9 ± 0.2 ppmv by Knudsen at the 455 K potential temperature level [Harris et al., 2002]. Such values are approximately in agreement with the moderate ozone loss deduced here from ILAS 0.5–1 ppmv (±0.2 ppmv) and HALOE 0.9–1.4 ppmv (±0.2 ppmv) at the 475 K level. However, in this winter vortex average losses are difficult to compare because ozone loss inside the vortex was spatially very inhomogeneous. Thus, the deduced averages will strongly depend on what fraction of the data entering the average originates from the vortex region showing the stronger ozone loss.
 The maximum vortex-averaged ozone loss deduced from ILAS observations was calculated for the period 1–15 May 1997 (see Table 1). In May, remnants of vortex air show a large ozone loss (33 ± 7 DU and a maximum of 45 ± 5 DU). The HALOE ozone loss in May was deduced only from three profiles with a maximum of 54 ± 7 DU. Deviations from the reference function are not as large as in March (HALOE) (Figure 13) and April (ILAS) (Figure 7), although the mean ozone loss increases by up to 10 DU (HALOE) and 8 DU (ILAS) compared to March. The low ozone values in these remnants of vortex air are first of all due to the very low ozone values present in the vortex in March. Further, an additional ozone loss will occur in the vortex remnants due to chemistry [Grooß et al., 1998; Hansen and Chipperfield, 1999]. The fact that such low values of ozone are observed in the vortex remnants indicates that they have remained largely intact and that no substantial mixing of outside air has occurred. Further, the ozone-tracer relations in May (Figure 13) are much more compact, consistent with the view that substantial mixing within the vortex remnants has taken place.
 The chemical ozone loss in the Arctic winter 1996–1997 was calculated from ILAS and HALOE measurements employing the ozone-tracer correlation technique. The ILAS instrument made measurements over a period of eight months from November 1996 to June 1997 covering a latitude range from 55–70°N. With the ILAS data set it was possible for the first time to describe in detail the complete life cycle of the polar vortex by the ozone-tracer correlation technique. A special characteristic of winter 1996–1997 was that the vortex formed very late [Coy et al., 1997]. Thus, mixing processes with air from outside the vortex still influenced the ozone-tracer relation at the edge and inside the polar vortex until the end of December 1996. For the beginning of January an early winter reference function valid for chemically unperturbed conditions was derived from ILAS O3 and N2O measurements. Significant ozone loss was calculated for the period from the end of February onwards. Remnants of vortex air were still found until the end of May.
 HALOE data are available inside the vortex for November/December 1996 and for March to May 1997. The early winter observations from HALOE in 1996 are not suitable for deriving an early winter reference function because of the late establishment of a stable vortex in this year. Therefore, the ILAS O3/N2O reference function was converted to an O3/CH4 and O3/HF relation to calculate the ozone loss from HALOE observations.
 The HALOE results of the two long-lived tracers CH4 and HF are in agreement within the uncertainty derived from the early winter reference function. At the end of March, the mean ozone loss calculated as the average of the results from CH4 and for 380–550 K is 61 ± 20 DU for the vortex core. There, large ozone loss occurred in a range of 90–110 DU and moderate ozone loss between 20 and 80 DU. For the altitude range of 450–550 K, the mean ozone loss is 28 ± 7 DU for the vortex core.
 The ozone loss calculated from ILAS observations did not reach such large ozone losses as calculated from HALOE data. However, the mean ozone loss calculated from the ILAS data for 450–550 K is 19 ± 6 DU for the vortex core (maximum: 39 ± 6 DU). HALOE and ILAS results still agree within the range of uncertainty for this altitude range. Also, both data sets consistently show a separation of the ozone loss into moderate ozone losses (HALOE 15–30 DU, ILAS 5–25 DU) and large ozone losses (HALOE 40–45 DU, ILAS 30–35 DU) inside the vortex core at the end of March.
 The peculiarity of the polar vortex in this winter is the strong inhomogeneity of the distribution of the ozone loss inside the entire vortex until March. Further, the vortex existed for a very long time up to the end of April. Very low ozone values were still observed inside the remaining vortex air during May 1997.
 We acknowledge all members of the ILAS and HALOE data processing teams for their work in producing two high-quality data sets, the UK Meteorological Office for providing meteorological analyses, and Andereas Engel for placing Cryosampler data at our disposal. We thank H. Gernandt and P. von der Gathen for the Ny-Ålesund ozone sonde data. We thank Larry Gordely and Takafumi Sugita for helpful comments on the HALOE and ILAS instruments, respectively. We further thank Janet Carter-Sigglow for a grammatical and stylistic revision of the manuscript. Part of this work was funded by the European Community within the SAMMOA project.