A high-resolution three-dimensional off-line chemical transport simulation has been performed with the SLIMCAT model to examine transport and mixing of ozone depleted air in the lower stratosphere on breakup of the polar vortex in spring/summer 2000. The model included ozone, N2O, and F11 tracers and used simplified chemistry parameterizations. The model was forced by T106 European Centre for Medium-Range Weather Forecasts analyses. The model results show that, by the end of June, above 420 K, much of the ozone-depleted air is transported from polar regions to the subtropics. In contrast, below 420 K, most of the ozone-depleted air remains poleward of approximately 55°N. It is suggested that the influence of the upper extension of the tropospheric subtropical jet provides a transport barrier at lower levels, while strong stirring on breakup of the polar vortex is important at upper levels. The mean meridional circulation modifies the distribution of ozone-depleted air by moving it up the subtropics and down in the extratropics. The model simulation is validated by comparing vertical profiles of ozone loss against ozonesonde measurements. The model results are consistent with many of the features present in the ozonesonde measurements. F11-N2O correlation plots are examined in the model and they show distinct canonical correlation curves for the polar vortex, midlatitudes, and the tropics. Comparison against balloon and aircraft measurements show that the model reproduces the separation between the vortex and midlatitude curves; however, the ratio of N2O to F11 lifetimes is somewhat too small in the model. It is shown that anomalies from the midlatitude canonical correlation curve can be used to identify remnants of polar vortex air which has mixed with midlatitude air. At the end of June there is excellent agreement in the position of air with anomalous F11-N2O tracer correlation and ozone-depleted air from the polar vortex.
 In the last decade a significant number of years have been characterized by cold stratospheric Arctic temperatures in winter and early spring with consequent enhanced activation of chlorine and destruction of ozone [World Meteorological Organization (WMO), 1999]. Upon breakup of the polar vortex in spring (sometimes referred to as the final stratospheric warming) much of the ozone-depleted air is transported into midlatitudes. Observations from Total Ozone Mapping Spectrometer indicate that there has been a decline in midlatitude ozone (30°–60°N) in May of 6.8% from 1979 to 1997. Knudsen and Grooss  estimate, from calculations done for 1995 and 1997, that approximately 40% of this trend can be accounted for by transport of polar ozone-depleted air into midlatitudes. Spring and summer are the key periods for health risk from increased UV radiation. Hence there is a clear need to understand the processes which control the transport of ozone-depleted air during this period.
 There have been relatively few studies that examine the transport and mixing associated with the breakup of the Arctic polar vortex. Hess  examined the evolution of ozone, N2O and potential vorticity (PV) from Limb Infrared Monitor of the Stratosphere (for 1979) and general circulation model output. He found that long-lived anomalies of tracers and PV were still observed 2 months after the breakup of the polar vortex. Waugh et al.  performed high-resolution contour advection calculations to follow the evolution of polar vortex fragments as they mixed into midlatitudes. The position of ex-vortex air after integration from 20 April to 7 May 1993 was consistent with aircraft measurements of low N2O made during the Stratospheric Photochemistry Aerosol and Dynamics Experiment campaign.
 A potentially powerful approach to diagnosing mixing between the polar vortex and midlatitude air is by using tracer-tracer relationships [e.g., Rex et al., 1999; Kondo et al., 1999; Abrams et al., 1996]. Scatterplots of two long-lived tracers produce compact relations referred to as canonical curves [Plumb and Ko, 1992]. Waugh et al.  identified an anomalous line on the concave side of the canonical curve from aircraft measurements during April–May 1993. They interpreted this curve as a mixing line that arose from mixing between the polar vortex and midlatitude air after the breakup of the polar vortex. However, Plumb et al.  have shown that this interpretation is incorrect. They show that weak continuous mixing across the edge of the polar vortex during winter (combined with diabatic descent and somewhat stronger mixing inside the vortex) gives rise to a separate canonical curve inside the vortex. Plumb et al.  examined output from two chemical transport models and these showed some indication of a separate vortex canonical curve. However, there was not as much separation between the polar vortex and midlatitude canonical curves in the models compared with observations. This was possibly due to too much mixing across the edge of the vortex in the model simulations which were both run at relatively low resolutions.
 The winter of 1999/2000 was characterized by polar temperatures which were cold enough to form PSCs until early March. From January to March there was substantial ozone loss inside the polar vortex which was observed and modeled during the SOLVE/THESEO-2000 campaign [Richard et al., 2001; Sinnhuber et al., 2000]. The aim of this paper is to characterize the transport of this ozone-depleted air after the breakup of the polar vortex. This will be done in two ways. First, the evolution of polar ozone loss simulated in a chemical transport model is examined. This will be compared with polar and midlatitude ozonesonde measurements. Second, long-lived tracer correlations are used to diagnose mixing between polar vortex and midlatitude air. The canonical tracer-tracer curves will be compared against balloon and aircraft measurements made during SOLVE/THESEO-2000 and previous campaigns.
 The emphasis of this study is to represent the tracer transport as accurately as possible. Hence we use a three-dimensional (3-D) chemical transport model run at very high horizontal and vertical resolutions. The computational expense of doing this necessitates the use of simplified chemistry parameterizations in the model. The details of the model and integration are given in section 2. In section 3 we present results showing the evolution of polar ozone loss by the model. Section 4 compares the profiles of the polar ozone loss tracer with measurements from ozonesonde stations. Section 5 examines the model simulation of F11-N2O tracer correlations and compares with balloon and aircraft measurements. It is shown how anomalous tracer correlations can be used to identify remnants of polar vortex air which have mixed into midlatitudes. The results of the paper are discussed in section 6.
2. Model and Simulations
 In this study we used the 3-D off-line chemical transport model SLIMCAT [Chipperfield et al., 1996; Chipperfield, 1999]. SLIMCAT has an isentropic vertical coordinate; here we use 23 vertical levels with the top and bottom boundaries at 2970 K (∼64 km) and 350 K (∼14 km), respectively. The vertical resolution is approximately 1 km in the lower stratosphere (model levels ∼15 K apart at 400 K) with decreased resolution in the mid and upper stratosphere. The horizontal resolution is approximately 100 km (T106 Gaussian grid). Horizontal wind and temperature fields are obtained from the operational 60-level European Centre for Medium-Range Weather Forecasts (ECMWF) analyses at T106 resolution. Tracer advection is performed by the accurate second-order Prather scheme [Prather, 1986]. Cross isentropic (diabatic) motion is calculated using the MIDRAD radiation scheme [Shine, 1987].
 Four chemical species are advected in the model: N2O, CFC-11 (F11), and two O3 tracers. The tropospheric source species, N2O and F11, have mixing ratios representative of the troposphere specified in the tropics on levels below 380 K. The chemistry for these two tracers is parameterized with a simple photolysis scheme as a function of solar zenith angle and pressure (coefficients calculated using a 2-D full chemistry model). Both ozone tracers have gas phase chemistry parameterized with the Cariolle scheme [Cariolle and Deque, 1986]. This scheme is based on linearization of the O3 photochemical sources and sinks using relaxation coefficients calculated with a 2-D full chemistry model. One ozone field has additional polar ozone loss to account for perturbed chlorine chemistry. This is implemented by means of “cold tracer,” X, which is set to unity in regions below polar stratospheric clouds formation temperature. The cold tracer then decays with an e-folding time of 5 days. In the presence of sunlight, ozone is destroyed by ∂O3/∂t = −κXO3, where 1/κ is 20 days. This simplified ozone parameterization has previously been used by Hadjinicolaou et al. . The difference between the two ozone tracers represents the amount of polar ozone loss.
 The tracer fields were initialized on 10 October 1999 from a T42 (∼300 km) version of the model which had previously been run for 1 year. This T42 integration had in turn been initialized from a multiyear low-resolution full chemistry integration (M.P. Chipperfield, personal communication, 2001). The model integration extended through the SOLVE/THESEO-2000 winter and spring of 1999/2000 and ended on 30 June 2000.
3. Transport of Ozone-Depleted Air
Figure 1 shows the distribution of polar ozone loss in the model on 3 days from spring to summer 2000 and contrasts the behavior on the 470 and 400 K isentropes. Figures 1a and 1b show the distributions on 1 March. Analysis of the PV at 470 K (not shown) shows that on this date there is still a strong coherent vortex, slightly displaced off the pole. Most of the polar ozone loss at 470 K is contained inside the vortex except for a filament which wraps around the vortex. The maximum ozone loss is approximately 1.2 ppmv. On the 400 K level the distribution of ozone loss (Figure 1b) is similar to that at 470 K. There has been substantial ozone loss that is confined to the pole even though this is in the subvortex region where there are no steep PV gradients.
 Around 15 March the polar vortex became very distorted and two large pieces broke off (not shown). On 15 April there remains a small vortex near the pole; Figures 1c and 1d show the distribution of the ozone loss. Because of cold temperatures in early March, there has been further ozone loss which is indicated by the increase in maximum values in the remnant of the polar vortex. At both 470 and 400 K there are large filaments of ozone-depleted air in midlatitudes. However, the character of these are rather different. At 470 K the filaments extend further into the tropics and are generally thinner (looking somewhat shredded). This indicates there has been stronger stirring at 470 K compared with 400 K during this period.
 From PV fields (not shown), the remnant of the polar vortex can be followed until mid May. As found by Hess , the PV slowly loses correlation with the tracer field. Figures 1e and 1f show the ozone loss on 30 June. The distribution on the two levels is dramatically different. At 470 K most of the ozone-depleted air is in a ring around subtropics. In contrast, at 400 K most of the ozone-depleted air is confined poleward of 45°N, except for a tongue which extends deep into the tropics around the eastern hemisphere. Prominent at 400 K are wave-like structures at the edge of the ozone depleted air. Rogers et al.  found similar tracer features at 400 K and showed that they arise from transport by tropospheric baroclinic eddies.
 The difference in transport between upper and lower levels can also been seen in Figure 2 which shows latitude-height (potential temperature) sections of zonal mean ozone loss. Figure 2a (1 March) shows most of the ozone loss is confined poleward of 60°N at all levels. Figure 2b (15 April) shows transport to lower latitudes with a region of large ozone loss below 450 K at 55°N; however, above 450 K there is more transport into the subtropics. Figure 2c (30 June) shows that below approximately 420 K ozone-depleted air is mostly confined poleward of 50°N, while above 420 K there has been significant transport into the subtropics. Note that the total amount of ozone loss decreases between 15 April and 30 June because the Cariolle scheme is relaxing the ozone back toward climatological values.
Figure 2 also suggests that the diabatic circulation is having a significant effect on the distribution of ozone-depleted air by 30 June. During the March–June period the distribution of diabatic heating (not shown) has ascent in the tropics/subtropics and descent poleward of approximately 40°N. This tilts the zonal mean distribution of ozone–depleted air moving it down at the poles and up in the tropics/subtropics. An adiabatic integration of SLIMCAT (i.e., only isentropic transport) over the period 15 March to 30 June (not shown) shows a very similar distribution of polar ozone loss as Figures 1e and 1f. The only differences are that at 470 K there is more ozone-depleted air at high latitudes (but the largest values remain in the subtropics), and at 400 K there are slightly higher values of ozone-depleted air in the tropics.
4. Comparison With Ozonesondes
 Features with small vertical scales are routinely observed in tracer profiles obtained from ozonesondes [e.g., Reid and Vaughan, 1991]. High-resolution transport model studies [e.g., Orsolini et al., 1995] have shown that these ozone laminae can arise from vertical shear tilting of the ozone filaments which break off the polar vortex. We have examined ozone measurements from a number of ozonesonde stations for March–May 2000 for evidence of ozone-depleted layers which may indicate polar vortex air. We have compared these with the polar ozone loss from the model in order to validate the tracer transport in the model over the period of polar vortex breakup.
Figure 3 shows selected ozone profiles from three ozonesonde stations (thin solid lines). In each of the panels there is also the profile of model polar ozone loss corresponding to the same date and grid point nearest to the station (thick-dashed line). Superimposed on each observed ozone profile is the mass-weighted average on the model levels (squares). One of the stations is at high latitudes (Ny Ålesund: latitude 78.9°N, longitude 11.9°W) and the other two are at midlatitudes (Hohenpeissenberg: latitude 47.8°N, longitude 11.7°W and Payerne: latitude 46.8°N, longitude 6.9°W). The dates for which we show observations were chosen among the available sounding dates for each station to be closest to 1 March, 1 April, and 15 April 2000.
 A filament of ozone-depleted air originating from the polar vortex causes a relative minimum in the ozone profiles. However, in comparing the two profiles, one should keep in mind that intrusions of low-ozone tropical air can also cause local minima in the observations. However, such tropical intrusions are obviously less likely at high-latitude stations. Figure 3a, showing ozonesondes from Ny Ålesund on 1 March is representative of high-latitude late winter observations. The model output suggests considerable polar ozone loss has occurred, and this is consistent with the rapid decrease in the ozonesonde profile around 500 K. In Figure 3d (31 March) the match between model and ozonesonde profile is evident. There is a deep layer of ozone-depleted air between 425 and 525 K and a peak in the model polar ozone loss over the same levels. In Figure 3g (12 April), most of the model polar ozone loss is at low levels; this cannot be validated from the ozonesonde profile, though it is not inconsistent either.
Figures 3b and 3c, showing ozonesondes from Hohenpeissenberg and Payerne, are representative of late winter midlatitudes. Here the match between model and data is also reasonably good. In the ozonesonde profile in Figure 3b there is an intrusion of ozone-depleted air between 400 and 475 K which corresponds to a layer of model ozone loss. The same can be said for Figure 3c where ozone loss is confined to a shallow layer at 400 K. This intrusion is also evident in Figure 1b where a filament of ozone loss can be seen to extend over Payerne. The agreement is similarly good in Figure 3e (2 April). In Figure 3f (31 March) observations show a broad layer, centered at 400 K, with relatively low ozone levels while, in the model, ozone loss is confined to a shallow layer centered at roughly 375 K.
 In Figure 3h (12 April) the simulated ozone loss shows two peaks at 475 and 400 K, while the observations show an ozone-depleted layer at 400 K only. Finally, Figure 3i (17 April) shows evidence of moderate ozone depletion around 400 K in both observations and model output.
5. Tracer-Tracer Correlations
 The correlation between long-lived tracers can give important information on stratospheric transport processes. In this section we examine the correlation between F11 and N2O.
Figure 4a shows a density scatterplot of F11 versus N2O from the model on 1 March 2000 for all grid points between 350 and 800 K, and the equator and North Pole. The different shades of gray correspond to different number density of occurrence (light shades are low-number density, dark shades are high-number density). The number densities were calculated in bins of size 1 ppbv (for N2O) by 1 pptv (for F11). It is apparent that most points cluster about three canonical curves (labeled A, B, and C). Scatterplots obtained from subsets of the latitude range (not shown) show that curve A forms from points in the polar vortex, curve B forms from midlatitudes, and curve C forms from the tropics. The position of curve C is noticeably not as well defined as the other two due to the spread in the model data. Figure 4a also shows other less steep lines of points joining the canonical curves; this is particularly apparent between the midlatitude and tropical canonical curves. These lines form from points on each isentropic model level, with higher vertical resolution; presumably, all the space between the polar vortex and tropical canonical curves would be filled in.
 The model canonical curves can be compared with Figure 4b which shows a scatterplot of F11 versus N2O from measurements made by the LACE and Aircraft Laser Infrared Absorption Spectrometer (ALIAS) instruments (squares and diamonds) from five flights of the OMS balloon platform between 1996 and 2000 [Ray et al., 1999; Richard et al., 2001] and in situ airborne measurements made by the Airborne Chromatograph for Atmospheric Trace Species (ACATS) instrument (pluses) in January–March 2000 during SOLVE/THESEO-2000. The lines on Figure 4b are canonical curves A, B, and C from Figure 4a. The measured F11 versus N2O scatterplot shows three clear canonical curves. The model tropical canonical curve agrees well with the curve from measurements. However, the gradient of the model midlatitude canonical curve is slightly less steep than the curve from measurements. The model vortex canonical curve is also shifted away from the curve from measurements, but the separation between vortex and midlatitude curves is similar between the model and measurements.
Plumb and Ko  show that for two tracers the slope of the canonical curve multiplied by the inverse ratio of their atmospheric abundance is equal to the ratio of their lifetimes. Calculating this from Figure 4 (using the region of the midlatitude curve at high N2O values, i.e., near the tropopause where the tracers are very long-lived and the slope of the canonical curve is linear) gives N2O/F11 lifetime ratios of 2.4 and 2.1 for the measurement and model midlatitude curves, respectively. The measured lifetime ratio is in agreement with previous published values (see Tables 1–7 given by WMO ). The shorter lifetime ratio given by the model could be due to errors in the simple N2O and F11 photolysis schemes or errors in the overall model transport.
 Remnants of polar vortex air can be diagnosed from F11-N2O tracer correlations through into summer. Figure 5 shows the F11 versus N2O scatterplot for 30 June. By this time, the vortex canonical curve has collapsed onto the midlatitude curve (see arrow). However, some remnants of the polar vortex can be detected as points in the scatterplot that lie off the curve on the concave side. Figure 6a shows the spatial distribution of these points on 30 June. Figure 6b shows the distribution of N2O for reference. The quantity plotted in Figure 6a (referred to as the “anomalous tracer correlation”) is the perpendicular distance away from the midlatitude canonical curve normalized by the standard deviation of the distribution (only negative departures, i.e., to the left of the curve, are plotted since these are the points where there has been vortex-midlatitude mixing).
 The regions of anomalous tracer correlation are mostly in a ring around the subtropics. This agrees remarkably well with the ozone loss field in Figure 1e. What is more, the regions of anomalous tracer correlation cannot be readily identified in the N2O field (Figure 6b).
 A key result from this paper is the difference in transport above and below approximately the 420 K level on breakup of the polar vortex. At upper levels much of the ozone-depleted air in the polar vortex is transported into the subtropics, while at lower levels the ozone-depleted air is confined poleward of approximately 55°N. The reasons for this can be seen from examining the distribution of ozone loss in conjunction with the horizontal winds shown in Figure 7. The prominent feature in the winds on 1 March at 470 K (Figure 7a) is the polar night jet associated with the polar vortex (band of westerly winds at 60°N). At 400 K (Figure 7b) there is evidence of the polar night jet but the most prominent feature is the upward extension of tropospheric subtropical jet (band of westerly winds at roughly 30°N latitude). On 15 April, at 470 K (Figure 7c) a small core of cyclonic winds can be seen associated with the remains of the polar vortex. There are also significant meridional winds in midlatitudes which are shredding the tracer field. At 400 K (Figure 7d) the subtropical jet is again dominant and the ozone-depleted air extends out to the core of the subtropical jet. On 30 June at 470 K (Figure 7e) the ozone-depleted air is embedded in the subtropical easterlies. At 400 K (Figure 7f) the subtropical jet is less prominent but can still be seen confining the ozone-depelted air to mid and polar latitudes. A notable feature in Figure 7f is the large anticyclone in the eastern hemisphere. This is the upward extension of the Asian summer monsoon, and a tongue of ozone-depleted air has been advected around this feature deep into the tropics.
Figure 7 indicates that at lower levels the upward extension of the tropospheric subtropical jet provides a transport barrier to equatorward movement of the ozone-depleted air. At upper levels, where transport is not influenced by the subtropical jet, most of the tracer is transported into the subtropics where the summer stratospheric easterlies provide a transport barrier to further equatorward movement. The mean meridional circulation then modifies the distribution of ozone-depleted air by moving it up the subtropics and down in the extratropics. The latitude-height section of the ozone loss tracer in Figure 2c is complementary to distributions of stratospheric aerosol (e.g., Figure 4b of Hitchman et al. ). These show confinement of aerosol to the tropics at levels above 20 km with transport from the tropics into midlatitudes at lower levels.
 Transport into the subtropics of much of the ozone-depleted air above 420 K is significant for risks to human health because of the very strong summer incident solar radiation at these latitudes. Further analysis is underway to determine if this transport occurs in every year.
 The model comparisons with ozonesonde profiles indicate that the distribution of polar ozone loss in the model is not inconsistent with the observations. In particular, Figures 3c, 3d, and 3e show ozone-depleted layers in the ozonesonde profiles which agree very well with the model polar ozone loss. However, quantitative comparison between the model and ozonesondes is difficult because the model cannot capture the smallest vertical scale structures, and even small shifts in the timing or position of ozone filaments can lead to large discrepancies. Nonetheless, the comparison does give some confidence to the model simulation of transport of ozone-depleted air over the period of vortex breakup.
 The F11-N2O tracer correlations shows that the model can produce very compact canonical curves for the polar vortex, midlatitudes, and tropics. The slight difference in ratio of F11/N2O lifetimes between the model and measurements makes quantitative comparison of mixing (during winter) across the vortex edge difficult. However, the separation of midlatitude and vortex canonical curves is comparable in the model compared with measurements, indicating that mixing of midlatitude air across vortex edge is reasonably well simulated.
 We have examined transport of ozone-depleted air on breakup of the polar vortex in spring/summer 2000. This study suggests that correlations of long-lived tracers can be used to identify remnants of polar vortex air as they mix with midlatitude air in spring and summer. New remote sensing observations of long-lived tracers, for example, from High-Resolution Dynamics Limb Sounder on the AURA satellite, to be launched in 2003, offer exciting possibilities for future model/measurements comparisons of the transport and mixing processes involved.
 We are grateful to Martyn Chipperfield for use of and advice on the SLIMCAT model. We thank Peter von der Gathen for the ozonesonde data from Ny Ålesund, Hans Claude for the ozonesonde data from Hohenpeissenberg, and Pierre Viatte for the ozonesonde data from Payerne. We thank Christ Webster for the ALIAS data. We would like to thank Alan Plumb, Yvan Orsolini, and an anonymous reviewer for their comments.
 This work was funded by the European Commission under the SAMMOA project (contract EVK2-CT-1999-00049) and by the UK Natural Environment Research Council.