A model study of the connection between polar and midlatitude ozone loss in the Northern Hemisphere lower stratosphere

Authors


Abstract

[1] A numerical model has been used to quantify halogen-induced ozone loss at the winter pole and at middle latitudes. This loss is compared with that due to nonhalogen ozone-destroying cycles. The three-dimensional off-line chemical transport model, SLIMCAT, was run at 3.75° latitude by 3.75° longitude, with United Kingdom Meteorological Office analyzed winds and temperatures. The contribution of polar processes to ozone loss at middle latitudes was investigated with novel tracers mapped to equivalent latitudes. Novel tracers were also included in SLIMCAT to follow ozone loss by reactions with ClOx, BrOx, NOx, and HOx, and to follow ozone production by oxygen photolysis. The interannual variability of the different processes was studied for five winters in the 1990s, which covered a variety of meteorological conditions in and around the polar vortex. Analysis of the ozone loss tracers shows large interannual variability in the relative strengths of particular chemical ozone loss mechanisms depending on the meteorology of the given year. In all cases, the ClO-BrO cycle dominates at polar latitudes. The role of mixing between the pole and middle latitudes also varies with the meteorological conditions. In winters with a cold, stable vortex, like 1996/1997, there is little impact of polar processes on midlatitude loss. In contrast, the cold, disturbed vortex of 1999/2000 contributed significantly to the ozone loss in middle latitudes. In all winters, ozone loss from cycles involving halogens was an important contributor (40–70%) to the modeled midlatitude ozone loss.

1. Introduction

[2] Ozone decline in both polar and middle latitudes has been well documented in recent years (e.g., see World Meterological Organization [1999]). The springtime polar losses are large, especially in the Southern Hemisphere, where essentially complete depletion occurs in the lower stratosphere (between about 14 and 23 km) each year. The total ozone column in September/October is now typically about 30% of the values in the 1960s. The Arctic polar loss is more variable but, in cold winters since the early 1990s, has also been large [e.g., Guirlet et al., 2000; Sinnhuber et al., 2000]. These polar losses are quite well understood (e.g., see Solomon [1999]), arising from halogen-induced ozone destruction in sunlight, following activation of the halogen species on polar stratospheric clouds (PSC).

[3] In contrast, the losses in middle latitudes are smaller and less well understood. The amount of ozone over a midlatitude observing site will be quite variable depending on the synoptic meteorological conditions [Dobson et al., 1926], so that small trends are difficult to detect. Determination of midlatitude losses depends on the statistical analysis of a long-period data set. For example, World Meterological Organization [1999] reported a springtime trend in the northern middle latitudes of about −3 to −6% per decade since 1979 based on satellite and ground-based data. Similarly, the attribution of this small trend is difficult. Both chemical [Solomon et al., 1998] and dynamical [Hadjinicolaou et al., 1997; Hood et al., 1997; Appenzeller et al., 2000] theories have been suggested. Solomon et al. [1998] argued that the loss arises from halogen activation on midlatitude aerosol, which is strongly enhanced following major volcanic eruptions, explaining the particularly low ozone observed after these events. Another source of activation could be midlatitude cirrus [Borrmann et al., 1997]. On the other hand, Hadjinicolaou et al. [1997] showed that dynamical variability, possibly driven by the radiative effect of the enhanced aerosol, could also explain part of the low ozone after the eruption of Mount Pinatubo. The possibility of a trend or long-period variability in transport could then offer an explanation for the trend in ozone. Model studies [Fusco and Salby, 1999], data studies [Hood et al., 1997], and statistical analysis of the ozone trend using dynamical indicators such as tropopause height or the North Atlantic Oscillation (NAO) index [Appenzeller et al., 2000] all point to an important dynamical contribution to the midlatitude ozone trend.

[4] We have used our three-dimensional chemical transport model, SLIMCAT, to look at various aspects of this problem. In a separate study, Hadjinicolaou et al. [2002] have carried out a multidecadal integration forced by daily meteorological analyses, including only a simple chemical scheme, to study the role of dynamic variability, and dynamical trends on ozone. Here in this study, our primary aim is not to understand the long-term trends but rather to investigate the role of the connection between polar and middle latitudes, using a version of the model with a detailed chemistry scheme. In particular, we aim to elucidate the processes leading to seasonal ozone loss in middle latitudes. Three main mechanisms of midlatitude loss have been suggested. The first is the advection of polar air activated within the vortex being mixed with midlatitude air which then causes local midlatitude ozone loss. The second mechanism involves polar air which has already experienced ozone loss which may then be mixed into middle latitudes causing a general “dilution” of ozone there. In these two cases the cause of the midlatitude loss lies in polar latitudes. In the third mechanism, ozone may be destroyed by in situ reactions occurring wholly at middle latitudes, following, for example, chlorine activation on aerosol or cirrus as discussed above. A discussion of these mechanisms can be found, for example, in the work of Pyle et al. [1995] and Chipperfield and Jones [1999]. Understanding these processes is important if we are to understand trends in the middle latitudes.

[5] Specifically, we have carried out a number of integrations of the winter and spring seasons during several years of the 1990s. The model uses diagnostic tracers, described by Lee et al. [2002]. These tracers allow the contribution to the accumulated ozone loss to be separated into the loss by individual chemical cycles and to be separated by the location of the origin of the loss. We concentrate on the lower stratosphere, where the largest depletion occurs

[6] In the next section we describe the model and the experiments in more detail. We provide an overview of the different years considered and then present results diagnosing the polar and midlatitude losses.

2. Model and Experiments

[7] The three-dimensional chemical transport model, SLIMCAT, has been discussed in detail by Chipperfield et al. [1997], Chipperfield [1999], and Guirlet et al. [2000]. SLIMCAT is an off-line model, driven here by daily United Kingdom Meteorological Office (UKMO) winds and temperatures [Swinbank and O'Neill, 1994] on 12 isentropic potential temperature surfaces from 335 K to 2700 K at a resolution of 3.75° latitude by 3.75° longitude (T31). The model uses the Middle Atmosphere Radiation Scheme (MIDRAD) radiation scheme [Shine, 1987] to calculate heating rates in isentropic coordinates, and hence cross-isentropic flow, and a second-order moments scheme of Prather [1986] for tracer advection. The Prather scheme has low numerical diffusion and maintains strong gradients in tracer distribution, an important consideration in polar wintertime simulations. This feature is particularly important for this study where tracers, often with sharp spatial gradients, will be used to track ozone loss by different reaction cycles or by the location of the loss.

[8] A detailed stratospheric chemistry scheme is included in the model with 49 chemical species integrated with a 15-min time step. In general, photochemical data are taken from DeMore et al. [1997] and the model radiation field is calculated by using a scheme based on that of Lary and Pyle [1991]. The model also contains an equilibrium treatment of reactions on liquid sulfuric acid aerosols and on solid nitric acid trihydrate (NAT) and ice particles [Carslaw et al., 1995a, 1995b].

[9] A suite of chemical ozone budgeting tracers is included (see Table 1), which are described in more detail by Lee et al. [2002] and Lee [1996], to follow chemical contributions to changes in ozone by individual reaction cycles. Loss due to cycles involving halogens (which are mainly of anthropogenic origin) and by reaction cycles which can be thought of as “background” or “natural” cycles (such as those involving hydrogen or nitrogen radicals, although these cycles can also be affected by anthropogenic perturbations) are tracked by individual tracers. These tracers solve a form of the continuity equation which follows individual reaction cycles (equation (1)) at each time step, giving the cumulative change in ozone concentration from initialization until any time in the run:

display math

where u is the flow velocity, χn is a tracer accumulating ozone loss due to an individual mechanism from Table 1, and Ln is the ozone loss frequency of the reaction mechanism given by the rate of the limiting step. Note that tracer 13 tracks the ozone production.

Table 1. Individual Ozone Loss (1–12) and Production (13) Mechanisms Included in the Ozone Budgeting Tracers of SLIMCAT
CycleReaction SequenceLn[O3]
1lO + ClO equation image Cl2O22j1[Cl2O2]
 Cl2O2 + hν → Cl + ClOO 
 ClOO equation image Cl + O2 
 2 × (Cl + O3 → ClO + O2) 
 net: 2O3 → 3O2 
2Cl + O3 → ClO + O22k2[ClO][O]
 ClO + O → Cl + O2 
 net: O3 + O → 2O2 
3NO + O3 → NO2 + O22k3[NO2][O]
 NO2 + O → NO + O2 
 net: O3 + O → and 2O2 
4OH + O3 → HO2 + O22k4[HO2][O3]
 HO2 + O3 → OH +2O2 
 net: 2O3 → 3O2 
5BrO + ClO → Br + ClOO2k5a[ClO][BrO]
 ClOO equation image Cl + O2 
 or (BrO + ClO → BrCl + O22k5b[ClO][BrO]
 BrCl + hν equation image Br + Cl) 
 Br + O3 → BrO + O2 
 net: 2O3 → and 3O2 
6Cl + O3 → ClO + O22k6[HO2][ClO]
 ClO + HO2 → HOCl + O2 
 HOCl + hν → OH + Cl 
 OH + O3 → HO2 + O2 
 net: 2O3 → 3O2 
7Br + O3 → BrO + O22k7[HO2][BrO]
 BrO + HO2 → HOBr + O2 
 HOBr + hν → OH + Br 
 OH + O3 → HO2 + O2 
 net: 2O3 → 3O2 
8NO3 + hv → NO + O22j8[NO3]
 NO + O3 → NO2 + O2 
 NO2 + O3 → NO3 + O2 
 net:O3 + O3 → 3O2 
9Br + O3 → BrO + O22k9[BrO][O]
 BrO + O → Br + O2 
 net:O3 + O → 2O2 
10HO2 + O → OH + O22k10[HO2][O]
 OH + O3 → HO2 + O2 
 net: O3 + O → 2O2 
11H + O3 → OH + O22k11[OH][O]
 OH + O → H + O2 
 net: O3 + O → 2O2 
12O + O3 → 2O22k12[O][O3]
13O2 + hν → O + O2j13[O2]

[10] A suite of geographical ozone budgeting tracers to follow ozone destruction by its location relative to the vortex is also included. Equivalent latitudes, derived from an inert tracer which is initialized at the beginning of the run from the potential vorticity field [Lary et al., 1995], is used to map latitude contours around the polar vortex. The Northern Hemisphere is divided into 5° equivalent latitude bands over all isentropic levels, and the ozone destruction occurring within each band is calculated from the sum of the ozone destruction budgeting terms:

display math

where ϕe is equivalent latitude and am ≤ ϕebm denotes a 5° equivalent latitude band within the Northern Hemisphere.

[11] Both the chemical and geographical ozone budgeting tracers are advected in the same manner as the chemical tracers in the model (although they do not participate in any further reactions).

[12] The tracers provide information on both chemistry and transport, integrated over the length of the run. Thus at any time and location, the chemical tracers describe the contribution to the cumulative ozone loss by particular chemical cycles, wherever that loss occurred. For example, losses driven by the ClO dimer cycle (cycle 1), which occur primarily at high, vortex latitudes, may be found at middle latitudes in the spring after transport from the vortex. In a complementary manner, the geographical tracers describe the contribution, at any location and time, by chemical processes operating within specified equivalent latitude bands. Thus in the case of the dimer cycle introduced above, the geographical tracer will highlight the contribution to the springtime middle latitude loss arising from chemical depletion within the vortex. For convenience, the 5° latitude bands have been grouped to define the vortex core, vortex edge, and midlatitude regions as 90°–70°N, 70°–60°N and 60°–30°N equivalent latitudes, respectively. These definitions are used throughout this paper, and although the vortex edge may not always occur at the same position, the definitions allow the ozone loss between the five winters to be compared.

[13] A series of model runs were restarted each December from a 9-year integration of SLIMCAT from 1991 to 1999, which was run at a resolution of 7.5° × 7.5° on 12 isentropic levels from 335 K to 2700 K (updated from Chipperfield [1999]). Thus we effectively sample the long integration but at the higher resolution of 3.75° latitude × 3.75° longitude (T31). The five experiments began on 1 December 1994, 1996, 1997, 1998, and 1999 and were run until after the final warming or the end of May, whichever occurred later. All the runs had a total chlorine loading of 3.6 ppbv and a total bromine loading of 20 pptv.

[14] An important question to consider is whether our results, presented below, depend on the spatial resolution of the model. We have estimated the sensitivity of the transport out of the vortex to resolution through numerical diffusion at the vortex edge. The numerical diffusion in the model arises when fine-scale structures develop which can no longer be represented by the second-order moments at the model resolution. We have updated previous sensitivity studies by Searle [1997] to estimate the maximum impact of model resolution on transport of ozone loss out of the vortex to middle latitudes. The 1998/1999 winter contained the most severely disturbed polar vortex (discussed later), with the highest degree of complexity, and therefore differences in transport due to model resolution should be largest in this year. We found by 21 May 1999 only a 5% difference between the transport of ozone-depleted air out of the vortex to middle latitudes on increasing SLIMCAT resolution from 3.75° latitude × 3.75° longitude to 1.88° latitude × 1.88° longitude (and Searle [1997] found only a very small influence of higher resolutions). We expect even smaller differences in the less disturbed winters, and thus we believe our results to be robust.

[15] This study will now focus on results in the lower stratosphere from the partial column of 342 K to 510 K potential temperatures, although all results are extracted directly from the full 12-level model.

2.1. Meteorology

[16] The interaction between chemical and dynamical processes in middle and high latitudes can be complex, and the processes leading to ozone loss are correspondingly subtle. For example, while low temperatures are important for PSC formation and chlorine activation, it is the location and extent of the low-temperature region which will be critically important for ozone loss. Low temperatures in the center of the vortex, and close to the pole, may possibly have only a small effect on chemical loss. On the other hand, a low-temperature region at the sunlit edge of the vortex, as in 1991/1992 [Pyle et al., 1994], could have an important influence on chemistry in the middle latitudes. We have therefore chosen five winters from the 1990s (1994/1995, 1996/1997, 1997/1998, 1998/1999, and 1999/2000), spanning a range of different meteorological conditions, for this study of Northern Hemisphere ozone depletion. The winters of 1994/1995, 1996/1997, and 1999/2000 all contained extensive areas in the high-latitude lower stratosphere with temperatures low enough for NAT formation at 50 hPa [Pawson and Naujokat, 1999; Manney and Sabutis, 2000]. The 1999/2000 winter was consistently cold, and temperatures fell below the PSC threshold for periods long enough to allow extensive denitrification to take place within the vortex [Sinnhuber et al., 2000]. The 1996/1997 polar vortex was late to form in December, and temperatures did not fall to PSC existence values until January, but thereafter temperatures were low and PSCs were measured extensively [Kyrö et al., 2000; Donovan et al., 1997]. The final warming did not take place until May, later than the other winters. There is also evidence from the Improved Limb Atmospheric Spectrometer (ILAS) of denitrification in 1996/1997 [Kondo et al., 2000]. Both the 1997/1998 and 1998/1999 winters were disturbed. The 1997/1998 winter was characterized by intermittent NAT PSC threshold temperatures throughout the winter. The 1998/1999 winter was the only one of the 1990s to contain a major midwinter stratospheric warming [Pawson and Naujokat, 1999]. Although average temperatures in 1998/1999 are the highest in the winters considered, the disturbed nature of the 1998/1999 vortex does provide an especially interesting opportunity to study vortex/midlatitude exchange.

[17] Thus the winters chosen provide a useful range of conditions to study polar ozone loss and the connection between polar and middle latitudes. At one extreme, 1996/1997 was a cold polar winter with a long-lived stable vortex. In contrast, 1998/1999 was only rarely cold enough for PSC formation but had a very disturbed vortex. The 1999/2000 winter was perhaps the most consistently cold of the winters, with a polar vortex which became quite disturbed after mid-March.

[18] Results from SLIMCAT studies of several of these winters have been presented previously. Chipperfield [1999] has compared results for the 1994–1997 winters to a variety of measurements and has shown that the model reproduces ozone and its interannual and seasonal variability well. Like any model, there are some discrepancies when compared with observations (for example, in the winter tropical gradient of N2O [see Chipperfield, 1999]), but in general the model behavior is very good and forms a suitable tool for the diagnostic studies presented here. In comparisons of model ozone to ground-based UV-visible spectrometer Systeme d'Analyse par Observation Zenithale (SAOZ) and ozone sonde measurements by Guirlet et al. [2000], the model was found to reproduce mean column ozone amounts within ±10%, with slightly too much ozone in 1996/1997 and too little ozone in 1997/1998 and 1998/1999. Investigation of the 1999/2000 winter by Sinnhuber et al. [2000] showed that SLIMCAT agreed extremely well with ozonesondes and Global Ozone Monitoring Experiment (GOME) measurements. The large ozone loss in 1999/2000 was due to an extended period of high average ClOx (around 2.4 ppbv at 480 K between 20 December 1999 and 1 March 2000) and the large denitrification present in the model in this year. The ozone loss in SLIMCAT in 1999/2000 has been compared to empirical estimates of loss based on the O3–CFC-11 correlation (A. Robinson et al., unpublished manuscript, 2001). There is very good agreement for both the timing and magnitude of the loss.

[19] In 1998/1999 the model overestimates loss compared with empirical estimates. This winter was warm and the model only produces two periods of temperatures low enough for PSCs. During the first period in early December there were observations of PSCs over Scandinavia ([European Ozone Research Coordinating Unit, 1999]), but for the second period in early February when the cold pool was east of Greenland, there is no observational evidence for PSCs. A small negative bias in the temperature analyses (and the UK Met Office temperatures in this region were lower than in the European Centre for Medium-Range Weather Forecasts (ECMWF) analyses) or a limitation in the model treatment of PSCs could easily cause the discrepancy between modeled and derived ozone loss. Comparison between modeled and observed ClO [Klein et al., 2000] suggests that the model does overestimate chlorine activation in February 1999. A warm bias in UKMO analyses was also reported in 1994/1995 of up to 1.7 K at the NAT point [Pullen and Jones, 1997], which could reduce the occurrence of PSCs within the model in that year.

2.2. Contributions to Polar Ozone Loss

[20] This section will examine the contributions to ozone loss found in the polar region. We will concentrate on the lower stratosphere between 342 K and 510 K and within the model equivalent latitude contours 70°–90°N. This region includes both the lower vortex core and the subvortex region, with the depth of the vortex varying between years. As the subvortex region tends to have weak potential vorticity gradients, horizontal mixing at these lower levels may be strong. One of the novelties of our ozone loss tracers is that we can follow the advection of midlatitude air into our 70°–90°N region by noting increases in the contribution to ozone loss by cycles such as cycles 3 and 4 which have earlier destroyed ozone at middle latitudes and have since been advected into the subvortex or vortexcore regions.

[21] Figure 1 summarizes the ozone losses during the five winters considered here. Each panel shows the cumulative loss due to cycles presented in Table 1, as well as the cumulative odd oxygen production by photolysis of molecular oxygen. In addition, the partial ozone column, the partial column of passive ozone, and the difference between these two are also shown. Note the extremely good agreement in the change in ozone determined by, first, the difference between cumulative ozone destruction and production (“O3 change” in Figure 1) and, second, the difference between modeled ozone and the passive ozone (“Pass O3 − O3” in Figure 1). These lines are coincident throughout most of each integration. Note also the differences in the evolution of the passive ozone during the different winters. The cold winters with a stable vortex, like 1994/1995, 1996/1997, and 1999/2000, all have much less passive ozone than the warmer, more disturbed winters (1997/1998 and 1998/1999), when stronger descent brings down ozone-rich air from above.

Figure 1.

Average cumulative ozone loss (Dobson units (DU)) within equivalent latitudes 90°–70°N for the partial column between 342 K and 510 K potential temperature, plotted in color according to the chemical mechanism responsible, for the years (a) 1994/1995, (b) 1996/1997, (c) 1997/1998, (d) 1998/1999, and (e) 1999/2000. The initial ozone on 1 December was also advected, and its partial column is labeled “Passive” ozone. Two different model calculations of the ozone loss are shown and agree very well. One estimate (“O3 change”) is based on chemical production and loss; the second is the difference between modeled ozone and modeled passive ozone (“pass O3-O3”).

[22] Temperatures below the PSC threshold were reported early in the 1999/2000 winter. By January a very cold and strong vortex had formed. Temperatures were less than 195 K for longer and over a larger area at 555 K and 480 K than for any of the previous 21 years [Manney and Sabutis, 2000]. UKMO analyses showed that temperatures (between the potential temperatures 440 K and 555 K) were below 195 K continuously from 14 November 1999 to 9 February 2000, with a further cold period from 11 February to 9 March 2000. PSC-driven halogen activation provided a continuously high concentration of ClOx, averaging 2.4 ppbv at 480 K poleward of 75°N from mid-December until the end of February [see also Sinnhuber et al., 2000]. This large activation provided rapid ozone loss by the fast ClO-BrO and Cl2O2 mechanisms on the return of sunlight (Figure 1e, days 40–80). The ClO-BrO and Cl2O2 mechanisms dominate, accounting for 44% and 33% of the total model ozone loss, respectively, in February, with exceptionally high average ozone loss rates in the vortex core of 1.7 Dobson units (DU) per day. The period with temperatures continuously below the PSC threshold provided extensive areas of denitrification within the model as ice particles fell to lower levels. The model's denitrification in the vortex at the end of the winter hindered the conversion of active ClOx to ClONO2, thus providing continued ozone loss up to the final warming in March [see also Sinnhuber et al., 2000]. This led to a net ozone change with an almost Southern Hemisphere magnitude of approaching 50% within this partial column by 30 April 2000 and net changes locally at 480 K of up to 70%.

[23] Up until the onset of the final warming (approximately day 80), the depletion at high latitudes is essentially produced in situ. After day 80, air from lower latitudes, in which some depletion has occurred since the start of the run, is advected to high latitudes, contributing to the loss there. Figure 1e shows this, for example, in the increase in ozone loss after day 80 by the HO2 and NO2 cycles which are more active at middle latitudes. This increase in mixing can also be seen in the suite of tracers following ozone loss by the equivalent latitude region in which the loss occurred (Figure 2). Again, after day 80, note the increase in loss originating below 60°N.

Figure 2.

The 1999/2000 average ozone loss (DU) within equivalent latitudes 90°–70°N for the partial column between 342 K and 510 K potential temperature, plotted according to the origin of the ozone loss at equivalent latitudes: 90°–70°N, 70°–60°N, 60°–30°N, and south of 30°N. The tracers show cumulative ozone loss from 1 December 1999.

[24] The 1996/1997 winter vortex was unusual for the winters considered in that it was late to form. Isolation of polar air from middle latitudes and subsequent radiative cooling giving rise to PSC threshold temperatures did not take place until January. The increase in ClOx concentrations required for the ClO-BrO and Cl2O2 cycles to dominate ozone destruction also occurred late, at the end of January (Figure 1b). Cumulative loss by this stage was the lowest of any of the years considered. However, the subsequent vortex air was well isolated and the vortex remained stable and circumpolar until May 1997. PSC threshold temperatures were observed until late March providing extensive halogen activation of vortex air within full sunlight, resulting in fast ozone destruction within the vortex core of 1 DU/d in March. At the cessation of NAT threshold temperatures (around day 80, 20 March 1997) the rate of ozone loss fell abruptly to 0.1 DU/d and remained low until after day 120 (early May), when cumulative loss sharply rose following advection of air from middle latitudes. This transition following the end of PSCs is often masked by changes in sunlight or mixing events; however, the stability and strength of the circumpolar vortex in 1996/1997 allowed this effect to be followed. The longer-lived stability of the vortex resulted in an extended dominance of the ClO-BrO cycle until the end of May, when the final warming increased the amount of midlatitude air present.

[25] In 1994/1995 and 1997/1998, temperatures dropped periodically below the PSC threshold temperature and polar vortices were relatively disturbed with several minor warmings. PSC-driven halogen activation (from December in 1994 and November in 1997) led to ozone destruction by halogen mechanisms accounting for 63% and 61% of the total destruction, respectively, by 30 April. The destruction by the ClO-BrO mechanism dominated the ozone loss as shown in Figures 1a and 1c. The onset of the final warming was marked by the increased importance of ozone destruction by the background (nonhalogen) cycles.

[26] In the highly disturbed winter of 1998/1999, analyzed temperatures provided some model ClOx activation in December 1998 and in early February 1999. There were few reports of PSC observations. PSCs were observed over Sodankyla on 2 December 1998 [European Ozone Research Coordinating Unit, 1999] and were also observed over Spitsbergen on 10 January 1999 [Beyerle et al., 2001], but there is no observational evidence to support the February activation, which occurred over the North Atlantic.

[27] The model PSC halogen activation was limited in extent and ClOx concentrations remained low (below 1.6 ppbv poleward of 75°N at 480 K). The dimer cycle was of minor importance (Figure 1d) but halogen loss by the ClO-BrO mechanism was still the largest factor in the model. Contributions by the NO2 and HO2 cycles (denoting advection in of midlatitude air) increased after 31 December 1998 (earlier than in any of the other years) following the major midwinter stratospheric warming and accelerated from the beginning of March with the onset of the final warming. The winter of 1998/1999 is especially notable for the high relative importance of the transport of ozone-depleted air from lower latitudes to polar latitudes. The model experiences an overall net ozone change in the vortex core of 21% (comparable with 1994/1995). This transport into the vortex of air from middle latitudes is a consequence of the weaker vortex during 1998/1999 and can be followed by both the increase in contribution to total ozone loss by air with a middle latitude chemical signature (i.e., ozone loss by the HO2 and NO2 cycles in Figure 1d) and by the substantial increase in ozone loss that has originated outside of the vortex from day 60 (Figure 3). Each of these indicators suggests that 40% of the polar loss is due to “natural” cycles occurring in middle latitudes, easily the highest contribution of any considered here.

Figure 3.

The 1998/1999 average ozone loss (DU) within equivalent latitudes 90°–70°N for the partial column between 342 K and 510 K potential temperature, plotted according to the origin of the ozone loss at equivalent latitudes: 90°–70°N, 70°–60°N, 60°–30°N, and south of 30°N. The tracers show cumulative ozone loss from 1 December 1998.

[28] Our overall estimate of the polar loss in 1998/1999 is much larger than some other estimates. For example, MATCH (as reported by European Ozone Research Coordinating Unit [1999]) indicates a negligible loss. It is important, however, to note that the two methods are quite different. The MATCH technique (e.g., see von der Gathen et al. [1995]) uses a Lagrangian approach and depends explicitly on mixing being ignored so that the in situ loss in an isolated air parcel within the vortex can be derived. However, the effect of mixing is implicit in the SLIMCAT estimates. For a winter like 1998/1999, where we have shown that mixing in of air from low latitudes, which has itself experienced ozone loss, is important, the two methods should not be expected to agree. Like MATCH we do agree that the in situ loss in 1998/1999 is the lowest of the years considered. Note, for example, the very low impact of the dimer cycle, seen in Figure 1d. The model loss will, of course, still be too high if PSC activation is overestimated, as discussed above.

2.3. Contributions to Midlatitude Ozone Loss

[29] Midlatitude ozone loss can be affected by the transport of processed polar air during the winter, by dilution of the polar ozone loss mainly in the spring, and by background in situ chemistry. We have used our diagnostic tracers to study the relative importance of polar and midlatitude processes to the midlatitude ozone loss. The winters studied contain two extremes in ozone loss at middle latitudes: the disturbed, cold vortex of 1999/2000 which produces large midlatitude ozone depletion and the isolated, cold vortex of 1996/1997 which results in much less midlatitude ozone depletion.

[30] In this study we have found that all winters show a midlatitude ozone loss by in situ HO2 and NO2 mechanisms which increases with time (see Figure 4). Most winters also show slow, steady detrainment of air from the vortex, marked by the gradually increasing contribution to ozone loss by the Cl2O2 mechanism which is active inside the vortex. Generally, the most important cycles for modeled midlatitude ozone loss in the partial column between 342 K and 510 K are those involving, first, HO2 and, second, ClO-BrO. The ClO and NO2 cycles are usually next in importance and are comparable. Note that the in situ ozone loss by the halogens either can be due to in situ activation of the halogens on midlatitude aerosol or could arise if activated air is removed from polar latitudes (“processing”). Our tracers do not distinguish between these two possibilities.

Figure 4.

Average cumulative ozone loss (DU) within equivalent latitudes 60°–30°N for the partial column between 342 K and 510 K potential temperature, plotted in color according to the chemical mechanism responsible, for the years (a) 1994/1995, (b) 1996/1997, (c) 1997/1998, (d) 1998/1999, and (e) 1999/2000. The initial ozone on 1 December was also advected and its partial column is labeled “Passive” ozone. Two different model calculations of the ozone loss are shown and agree very well. One estimate (“O3 change”) is based on chemical production and loss; the second is the difference between modeled ozone and modeled passive ozone (“pass O3-O3”).

[31] The winter of 1996/1997 represents one extreme of the influence of polar processes on middle latitudes. Figure 5 shows that the high stability of the vortex in that year leads to a very small contribution from high latitudes (ϕe > 60°N) to the midlatitude loss up until the final warming around day 130. This is also very evident in Figure 4b. The dimer cycle, which plays a strong role in polar latitudes (Figure 1b), has a negligible influence in middle latitudes until the final warming. Clearly, up to that time the two regions are effectively isolated from one another. There is significant midlatitude loss by the ClO-BrO cycle and the classical ClO cycle (cycle 2 in Table 1). This has an obvious implication. The importance of the two halogen cycles can only be due in this winter to in situ activation in the middle latitudes. Nevertheless, the contribution to ozone destruction by the halogens up to the final warming is less than about 20% of the total, which is dominated by the natural HO2 cycle (cycle 4).

Figure 5.

The 1996/1997 average ozone loss (DU) within equivalent latitudes 60°–30°N for the partial column between 342 K and 510 K potential temperature, plotted according to the origin of the ozone loss at equivalent latitudes: 90°–70°N, 70°–60°N, 60°–30°N, and south of 30°N. The tracers show cumulative ozone loss from 1 December 1996.

[32] There is a further small point of interest in Figure 4b: the springtime ozone production in middle latitudes (30°–60°N) is greater than in other years. This arises from the strong symmetric vortex which leads to uniform illumination in middle latitudes (and correspondingly less photolytic ozone production over the pole-centered vortex during this year; see Figure 1b). The combination of larger ozone production and reduced mixing of ozone-depleted air out from the vortex results in a net gain in ozone concentrations at the end of April (days 115 to 130, Figure 4b), prior to the dilution of polar ozone-depleted air following the final warming.

[33] The middle latitude loss in the winter of 1999/2000 contrasts strongly with 1996/1997. Losses occurring north of 60°N are much more significant to middle latitudes in 1999/2000 (Figure 6) both before and, especially, after the warming in March. By the end of May, 50% of the midlatitude decline is due to losses in higher latitudes (Figure 6) and more than 50% is due to halogen cycles (Figure 4e). Uniquely among the years considered, the ClO-BrO cycle is the most important loss cycle of the middle latitudes.

Figure 6.

The 1999/2000 average ozone loss (DU) within equivalent latitudes 60°–30°N for the partial column between 342 K and 510 K potential temperature, plotted according to the origin of the ozone loss at equivalent latitudes: 90°–70°N, 70°–60°N, 60°–30°N, and south of 30°N. The tracers show cumulative ozone loss from 1 December 1999.

[34] These two extreme cases are important for understanding present and possible future midlatitude loss. In four of the winters studied, the model has calculated substantial halogen-driven polar ozone loss. This loss is largest in 1999/2000, when denitrification played an important role. In all these winters, except 1996/1997, there was a corresponding impact on the middle latitudes. These three winters all had, to some extent, disturbed polar vortices and there was mixing into middle latitudes. In contrast, there was negligible leakage from the pole to middle latitudes in 1996/1997, prior to the final warming, and very little impact of the pole on midlatitude loss. Thus we conclude that the stability of the vortex is important for mid-latitude ozone depletion. A cold vortex, with substantial polar loss, can influence the middle latitudes in different ways, depending on the meteorology.

[35] The lower stratosphere is predicted to cool with increased greenhouse gases, and this is expected to lead to enhanced polar loss [see also Waibel et al., 1999]. The impact of this on middle latitudes is not clear and, as we have shown, will also depend on any change in the stability of the vortex.

3. Conclusions

[36] Tables 2a, 2b, and 2c at 30 April (day 120) summarize the results presented earlier for both polar and middle latitudes. Table 2a shows, for the year considered, the modeled ozone partial column and the change in ozone as an absolute value and expressed as a percentage. Table 2b describes the ozone changes calculated for the individual cycles, given in Table 1. Table 2c shows the contribution to the modeled ozone loss, as a percentage, which originates in chemical processes in different latitude regions.

Table 2a. Modelled Ozone, Passive Ozone, Partial Column in DU on 30 April, and Change in Ozone (Chemical Production Minus Chemical Loss) as an Absolute Value and Expressed as a Percentagea
 Polar Vortex Air (90°–70°N)Midlatitude Air (60°–30°N)
1999/20001998/19991997/19981996/19971994/19951999/20001998/19991997/19981996/19971994/1995
  • a

    All values are integrated between 342 K and 510 K.

Passive O322624826618620411613913691110
O31181981921321609212411992102
Change in O3−110−51−56−56−45−25−16−17+1−9
Percent Change in O3−49−21−28−30−22−22−12−12+1−8
Table 2b. O3 Budget Tracers per Dobson Unit With Ozone Loss Calculated for the Most Important Individual Cycles, Given in Table 1a
 Polar Vortex Air (90°–70°N)Midlatitude Air (60°–30°N)
1999/20001998/19991997/19981996/19971994/19951999/20001998/19991997/19981996/19971994/1995
  • a

    The percentage of the ozone loss by cycles involving halogens is found by the sum of cycles 1, 2, 5, 6, and 7 over the total ozone destruction (cycles 1 to 12) from Table 1.

ClO-BrO4717272117138835
Cl2O234414129723<12
ClO1511146755523
Other Cl/Br9886554433
HO27151181189988
NO23673323422
O3 production2121243141820202218
Percent O3 loss by halogens80566175637053544048
Table 2c. Pecentage of O3 Loss Originating From Different Equivalent Latitude Regionsa
 Polar Vortex Air (90°–70°N)Midlatitude Air (60°–30°N)
1999/20001998/19991997/19981996/19971994/19951999/20001998/19991997/19981996/19971994/1995
  • a

    The contribution to the modeled ozone loss is shown as a percentage of the total loss, calculated from equation (2), which originates at different equivalent latitude regions.

90°N–70°N5537526360262118318
70°N–60°N25221922202618171019
60°N–30°N11331613133749516548
30°N–90°S9813271112142215

[37] The first point to note is that there is a very large interannual difference in the high-latitude (and midlatitude) passive ozone partial column, obviously driven by dynamics. For example, the winter with a strong, stable vortex, 1996/1997, also has the lowest high-latitude partial column of ozone, associated with the weak descent (and hence low temperatures) in this winter. The partial ozone column in the more disturbed winter of 1997/1998 is nearly 50% larger than that in 1996/1997.

[38] In contrast, the absolute change in high-latitude ozone varies little from year to year, except for 1999/2000. The changes in the earlier years confirm the conclusions of Chipperfield and Jones [1999], indicating that dynamical processes play the largest role in changing the Arctic springtime ozone column. The exception to that is, of course, the winter of 1999/2000, where the net change in modeled ozone is twice that calculated in any other year. This arises, as indicated by Sinnhuber et al. [2000], owing to the strong modeled denitrification in that winter which allows ozone loss by the halogen chemistry to remain important for a much longer period. Our results would suggest that denitrification would be essential for future, enhanced Arctic ozone loss (e.g., see Waibel et al. [1999]).

[39] Table 2b shows that there is a strong contribution to the modeled chemical ozone loss at the pole from the halogens in all years. In all cases the ClO-BrO cycle dominates, confirming the earlier results by Chipperfield and Pyle [1998] (although some models still show a smaller contribution [Woyke et al., 1999; Daniel et al., 1999]). The largest halogen contribution is the 80% calculated in 1999/2000, and the smallest is the 56% calculated in 1998/1999. In this latter disturbed winter the modeled polar ozone loss has a high contribution (33%, 2–3 times greater than that in other years) arising from transport from the middle latitudes and the smallest contribution from in situ loss (see Table 2c).

[40] The halogens also play a major role in the modeled ozone loss in the middle latitudes. In 1999/2000, the year with largest polar loss, 70% of the midlatitude ozone loss is due to the halogens. In contrast, the strong polar vortex of 1996/1997, with substantial halogen-driven polar ozone loss, leads to the lowest halogen-induced loss (40%) in middle latitudes. In either case, the contribution of chemical cycles involving halogens is very significant and emphasizes the important role that changes in halogen loading could have on midlatitude trends.

[41] The overall modeled change in middle latitudes is very variable, ranging from −25 DU (−22%) in 1999/2000 to a small production of 1 DU (+1%) in 1996/1997. The largest loss corresponds to the largest contribution from the pole and the smallest in situ loss; the year with the small production has least input from polar loss and the largest impact of in situ chemistry. There are clearly important connections between midlatitude and high-latitude ozone change which depend critically on the strength of the polar vortex. A strong, isolated vortex, with possibly large ozone loss, will have only a modest impact on the middle latitudes. The largest connection between the two regions is in disturbed winters (with, for example, about half the midlatitude loss in 1999/2000 originating north of 60°N). Future ozone trends over middle latitudes will be influenced by not just halogen loading but also polar temperatures and, crucially, the stability of the polar vortex.

Acknowledgments

[42] We gratefully acknowledge the following for their support: We thank M. P. Chipperfield for the SLIMCAT code and initialization data. Meteorological analyses have been provided by the United Kingdom Meteorological Office. G. A. Millard thanks EPSRC for a research grant. This work forms part of the NERC UGAMP program. We also thank DETR for support via EPG1/1/83. The Centre for Atmospheric Science is a joint initiative of the Department of Applied Mathematics and Theoretical Physics and the Department of Chemistry at the University of Cambridge.

Ancillary