Results from an 11-year simulation of the present-day atmosphere with a coupled tropospheric chemistry-climate model have been used to investigate the extratropical transport from the stratosphere to the troposphere, the transport and photochemical destruction of ozone from stratospheric origin in the troposphere, and the resulting contribution to tropospheric OH. The simulated stratosphere–troposphere exchange (STE) is based on the model calculated wind fields, and its seasonality and spatial distribution resemble results derived from other methods. The lifetime of stratospheric ozone in the troposphere is largely determined by transport between the extratropical tropopause where it enters the troposphere and the subtropical lower troposphere where its photochemical destruction rate maximizes. Comparison of the simulated ozone and O3s budgets suggests that ozone from stratospheric origin contributes about 15% to the average oxidation capacity in the Northern Hemisphere (NH), with regional contributions up to 40%. STE has a reverse effect on OH in the extratropical upper troposphere related to the influence of water vapor concentrations. The oxidation capacity of the troposphere may therefore be sensitive to atmospheric dynamic effects of climate change.
 The tropospheric ozone budget is determined by photochemical production and destruction, dry deposition and cross-tropopause transport of ozone from the stratosphere. Stratosphere–troposphere exchange (STE) contributes to the distribution of ozone in the troposphere, and accurate knowledge of its impact on tropospheric ozone levels is necessary to assess the climate effects of ozone perturbations due to anthropogenic activities. The problem of STE has often been approached from the general circulation perspective without specific attention to synoptic-scale disturbances [Haynes et al., 1991; Holton et al., 1995]. This approach considers the mean circulation induced by the dissipation of Rossby waves as they propagate from the troposphere to the stratosphere. However, it has been argued that the circulation of the extratropical lower stratosphere cannot merely be described in terms of the zonal mean circulation [Tuck et al., 1997]. Zonally asymmetric synoptic-scale events codetermine the distribution of trace gases, such as ozone and water vapor, in the tropopause region [Bregman et al., 1997; Lelieveld et al., 1997; Kentarchos et al., 2001]. This distribution depends on the location, strength and timing of synoptic-scale events which in the extratropics are associated with tropopause folds and cutoff low events [Danielsen, 1968; Shapiro, 1980; Hoskins et al., 1985; Price and Vaughan, 1993; Lamarque and Hess, 1994; Kentarchos et al., 2000].
 Although the important mechanisms associated with STE have been identified, its contribution to the tropospheric ozone budget remains relatively uncertain. Global models indicate cross-tropopause ozone fluxes between 391 and 846 Tg O3 yr−1, while global ozone production and destruction are estimated to range between 3019–4100 and 2511–4064 Tg O3 yr−1, respectively [Roelofs and Lelieveld, 1997; Hauglustaine et al., 1998; Wang et al., 1998; Crutzen and Lawrence, 1999; Lelieveld and Dentener, 2000; Kentarchos et al., 2001]. We remark that recent model intercomparisons have demonstrated that simulated vertical transport efficiencies and chemical tracer fields in and near synoptic disturbances are characterized by large differences between models [Meloen et al., 2003; G. J. Roelofs et al., Intercomparison of tropospheric ozone models: Ozone transports in a complex tropopause folding event, submitted to Journal of Geophysical Research, 2003, hereinafter referred to as Roelofs et al., submitted manuscript, 2003].
 During STE events relatively dry and ozone-rich air masses from the stratosphere enter the troposphere. Mixing of stratospheric and tropospheric air masses modify tropospheric distributions of ozone, water vapor, and hydroxyl (OH) radicals which control the lifetime of many gases [Levy, 1971]. OH is produced from photodissociation of O3 and subsequent reaction of the resulting O(1D) with water vapor. The possible impact of STE on tropospheric OH levels has not been studied in detail before. From sensitivity experiments carried out with a 3-D tropospheric chemistry model, Lamarque et al.  concluded that the stratospheric ozone flux increases the annual average tropospheric column of ozone by ∼4 DU in the Northern Hemisphere (NH) extratropics. Neglect of the stratospheric source of tropospheric ozone resulted in a 5% decrease of the global OH abundance. Their model uses monthly averaged wind fields as input while the stratospheric ozone source is prescribed as an emission flux that is decoupled from the simulated dynamics. The coupled chemistry-climate model applied in this study (ECHAM) derives cross-tropopause ozone transport directly from the simulated dynamics in the lower stratosphere and upper troposphere [Roelofs and Lelieveld, 1997].
 Our previous studies of STE focused on the analysis of simulated events associated with synoptic disturbances and on the global and hemispheric STE climatology. In this study, we examine the contribution by ozone from stratospheric origin to the oxidation capacity of the troposphere, using data from an 11-year simulation of the coupled chemistry-climate model ECHAM. In section 2, the model is described. Section 3 presents simulated tropospheric budgets of ozone from stratospheric origin, and section 4 presents the spatial distributions of its source and photochemical sink. In section 5, the associated contribution to tropospheric OH is investigated, and section 6 presents a summary and conclusions.
2. Model Description
 The climate model used in this study is the European Centre Hamburg Model version 4 (ECHAM4), with a horizontal resolution of 3.75° × 3.75° and a time step of 1800 s (T30). The model contains 19 hybrid σ − p layers between the surface and the top level at 10 hPa. Average pressure levels relevant for the troposphere and lower stratosphere are 995, 980, 950, 908, 846, 770, 680, 590, 495, 405, 320, 250, 190, 140, 100, and 73 hPa referring to approximate altitudes of 0.03, 0.14, 0.38, 0.77, 1.4, 2.1, 3.1, 4.2, 5.4, 6.8, 8.3, 10, 12, 13.8, 15.7, and 18 km above the surface. An elaborate description of ECHAM and the simulated climate can be found in the study of Roeckner et al. [1996, 1999]. A tropospheric chemistry scheme considering emissions of NO, CO and nonmethane hydrocarbons (NMHC), dry deposition of O3, NO2, HNO3, and H2O2, wet deposition of HNO3 and H2O2, parameterized surface CH4 concentrations, and the CBM4 scheme for the representation of higher hydrocarbon chemistry is coupled to ECHAM [Roelofs and Lelieveld, 1995, 2000a]. Tracer transport is calculated with a semi-Langrangian advection scheme [Rasch and Williamson, 1990]. Although this transport algorithm is not completely mass conserving, we found that the resulting artificial gain or loss of the chemical tracers is insignificant compared to the simulated emission, chemical transformation and deposition budget terms. Additional vertical transports are included through parameterization of vertical diffusion and convection.
 Because the model does not represent chemical reactions specific for the stratosphere, upper stratospheric ozone is parameterized with results from a 2-D troposphere–stratosphere chemistry model [Brühl and Crutzen, 1988]. Additionally, the model applies an ozone potential vorticity (PV; unit in PVU) correlation in the lower stratosphere to preserve longitudinal variability of ozone associated with the meandering and breaking of the jet streams. The correlation is deduced from MOZAIC observations (Measurement of Ozone by Airbus In-Service Aircraft) [see Marenco et al., 1998], as described by Roelofs and Lelieveld [2000b]. The parameterization is applied only in the extratropical lower stratosphere at latitudes higher than 25°. Ozone is not parameterized in the lowermost stratospheric model layers to allow for mixing between the upper troposphere and the lower stratosphere.
 The simulated ozone seasonality at the surface, in the free troposphere and in the tropopause region agrees well with observations [e.g., Roelofs and Lelieveld, 1997, 2000a]. The same is true for tropospheric ozone distributions during particular meteorological conditions, e.g., a synoptic disturbance, for horizontal resolutions of T30 (as applied in this study) and finer [Roelofs and Lelieveld, 2000b; de Laat et al., 1999; Kentarchos et al., 2001]. However, because of the relatively coarse vertical resolution near the tropopause and the rather diffusive semi-Lagrangian transport scheme the model tends to overestimate the downward transport efficiency of stratospheric ozone into the troposphere [Meloen et al., 2003; Cristofanelli et al., 2003; Roelofs et al., submitted manuscript, 2003].
 The model considers a separate tracer for ozone that originates from the stratosphere, which is referred to as O3s. The concentration of this tracer is set equal to ozone in the stratosphere where concentrations are prescribed. The ratio between O3s and O3 reflects the fraction of ozone that originates from the stratosphere. While being transported from the stratosphere into the troposphere along with the calculated air motions, ozone is photochemically destroyed but also produced. However, O3s is photochemically destroyed but not produced. The difference between O3 and O3s, referred to as O3t, reflects therefore the chemical production of ozone. Since the extratropical lower stratosphere is dominated by photochemical destruction of ozone, O3t is predominantly from tropospheric origin. Calculation of the photochemical destruction of O3s is based on the odd-oxygen family concept, with main reactions being the photodissociation of O3s and subsequent reaction of the O(1D) produced with water vapor yielding two OH radicals, and reaction of O3s with OH or HO2. Note that O3 and O3s are also removed by dry deposition.
 We carried out an 11-year simulation after a 2-year spin-up time. By using calculated ozone distributions as input for the radiative transfer scheme of ECHAM4, atmospheric dynamics and chemistry are treated interactively. We note that this enhances the internal consistency of the model compared to previous studies where chemistry and meteorology were not coupled through radiative transfer. However, the radiative impact of tropospheric ozone is small compared to other radiatively active trace gases, such as water vapor, clouds and CO2, so that simulated distributions and hemispheric and seasonal budgets of tropospheric ozone are similar to our previous studies. The average global tropospheric OH concentration in this simulation is 1.01 × 106 mol cm−3, with values of 1.10 × 106 and 0.91 × 106 mol cm−3 in the NH and SH, respectively. This is consistent with estimates from other methods [e.g., Krol et al., 1998].
3. Tropospheric Ozone Budgets
Table 1 shows calculated annual source and sink terms for ozone and for O3s, i.e., the fraction that derives from STE, with the calculated standard deviation. The table refers to the NH troposphere. We note that direct calculation of tracer mass fluxes across the highly variable tropopause is a rather complicated and time-consuming task for a Eulerian model [Meloen et al., 2003]. Therefore, we use the calculated residual of the remaining source, i.e., photochemical formation, and sinks, i.e., chemical destruction and dry deposition, on the NH to represent the net cross-tropopause transport of ozone and O3s.
Table 1. Annual NH Source and Sinks Terms (Tg O3 yr−1) and Tropospheric Content (Tg O3) for Ozone and O3s
437 ± 11
641 ± 10
2676 ± 16
2658 ± 15
554 ± 9
460 ± 3
87 ± 2
174 ± 1
49 ± 1
 The dominant source term for ozone is photochemical production resulting from natural and anthropogenic emissions of ozone precursors. Photochemical production is defined as the reaction of NO with HO2, CH3O2, or XO2 (a CBM4 species that represents higher hydrocarbon peroxy radicals), whereas destruction is defined as photodissociation of ozone and subsequent reaction of O(1D) with water vapor, reaction of ozone with OH or HO2, and reactions with NMHC [Roelofs and Lelieveld, 2000a]. The simulated variability of the chemical terms is rather small because emissions are identical each model year, except for NO emissions from lightning which are linked to the simulation of deep convection [Roelofs and Lelieveld, 1997]. The source of ozone due to STE is smaller, contributing about 14% to the total source.
 The sum of the sinks, i.e., photochemical destruction and dry deposition, balances the sources. One must realize that the simulated STE of ozone is the residual of downward, stratosphere-to-troposphere transport (STT) [Stohl et al., 2003] of O3s, i.e., 641 Tg O3 yr−1, and upward, troposphere-to-stratosphere transport (TST) of O3t, i.e., 204 Tg O3 yr−1. The budget term for STE of O3s, is associated with STT. The simulated variability of this term is about 2%. It is possible that the modeled interannual variability of the intensity of the Brewer–Dobson circulation, of which STE is a part, is relatively small because of the relatively low top layer at 10 hPa. However, on regional scales, the variabilities may be considerably larger as shown in section 5.
 Obviously, since the direction of the instantaneous air mass flux across the tropopause varies not only with latitude and season (as will be shown in Figure 2), but also on relatively small spatial as well as temporal scales (as illustrated, e.g., in the case study presented by Meloen et al. , the net downward flux of O3s presented here is a relatively small residual of larger downward and upward fluxes of O3s across the tropopause. The presented O3s STT budget of 641 Tg O3 yr−1 only reflects this residual, corresponding with the so-called irreversible cross-tropopause transport of O3s. O3s that is reversibly transported from the stratosphere to the troposphere, i.e., it is not destroyed by chemical destruction or dry deposition in the troposphere but transported back into the stratosphere, is accounted for in the tropospheric O3s content and distribution. We emphasize that only the O3s that is chemically destroyed in the troposphere, which corresponds therefore with irreversible mixing, contributes to the photochemical production of OH in the troposphere. Similarly, the presented O3t TTS budget reflects only the irreversible upward cross-tropopause transport of O3t. Note that dry deposition is a relatively small sink for O3s. O3s contributes on average about 28% to the global annual tropospheric ozone content in the NH, and this varies between 20% in late summer to 40% in winter/early spring [Roelofs and Lelieveld, 1997].
Figure 1 shows the simulated seasonalities and variability of the NH monthly net cross-tropopause transport, photochemical destruction, dry deposition and tropospheric content of O3s. The net flux of O3s into the troposphere maximizes in February and March, and a secondary maximum occurs in June. Appenzeller et al.  showed that the cross-tropopause air mass flux in the NH has a maximum in midwinter and a stronger maximum in early summer, the latter being associated with the tropopause height seasonality. During winter when insolation is relatively weak and photochemical activity is low, the influx of O3s exceeds chemical destruction so that the amount of O3s in the troposphere increases. In spring and summer, the photodissociation efficiency increases while the cross-tropopause flux of O3s decreases, and the O3s amount decreases, reaching a minimum in late summer. From the data shown in Figure 1 we calculate that the chemical lifetime of O3s in the NH is 1.5–2 months in winter, and less than a month in summer.
4. Latitudinal Distribution of O3s Sources and Sinks
Figure 2 shows the downward fluxes of air and O3s across the 320 hPa model level (∼8.5 km asl) as function of latitude for January and April when the downward flux is relatively large. The 320 hPa level is generally located below the tropopause, especially at (sub)tropical latitudes where the tropopause height is up to 16 km altitude.
 The computed air fluxes are shown in the top panels. In January, the maximum downward air flux is located between 20°N and 40°N, and a net upward flux is simulated between 40° and 70°. This is consistent with the cross-tropopause flux analysis by Hoerling et al.  for January 1979. The simulated flux at extratropical latitudes displays a considerable year-to-year variability, although it appears smaller than that deduced from 15 years of ECMWF meteorological data by Sprenger and Wernli . The simulated air flux maximum across the 320 hPa level is located about 5°–10° southward from that across the tropopause as calculated by Sprenger and Wernli  and Hoerling et al. . Between January and April, the downward air flux decreases considerably.
 The lower panels in Figure 2 present the computed O3s fluxes. The relatively strong O3s concentration gradient between the troposphere and the stratosphere results in a net downward O3s flux at almost all latitudes. Generally, lower stratospheric ozone concentrations increase poleward. As a result, the computed downward flux of O3s maximizes further northward than that of air.
 In winter and spring when the photochemical activity is relatively low, O3s may be transported over relatively large distances before it is photochemically destroyed. Figure 3a shows the simulated zonal distribution of O3s, discussed previously by Roelofs and Lelieveld . Subsidence of relatively ozone-rich air prevails over NH subtropical latitudes. Ozone transport from more northern latitudes, where ozone lifetimes are considerably longer, contributes to this. Figure 3b shows the simulated distribution of the chemical destruction of O3s, integrated over grid volume and longitude. Because of the resolution dependence of this distribution, the diagram should be considered qualitatively rather than quantitatively. O3s destruction maximizes in the subtropical lower troposphere where insolation and humidity are relatively large. Therefore, the average lifetime of O3s in the troposphere is strongly determined by the transport time between the tropopause and the subtropical lower troposphere.
 We have examined the relationship between the computed lifetime and the latitude of the maximum downward O3s flux more closely for April. The computed lifetime in April varies between 1.19 and 1.27 months. The shortest lifetime, i.e., 1.19 month, occurs in the model year where the maximum STE flux, generally located between 20°N and 50°N, is concentrated relatively southward compared to the other years. However, significant variability occurs also at higher latitudes, and different peaks at different latitudes may occur (Figure 2). This makes the relationship much less pronounced for most of the years.
5. Impact on Hydroxyl Radical in the Troposphere
Figure 4 shows the calculated seasonal cycle of the chemical destruction of ozone integrated over the NH, with the separate contributions by the stratospheric (O3s) and tropospheric (O3t) ozone types. We remark that the ozone destruction pathway leading to OH formation, i.e., photodissociation followed by O(1D) + H2O, contributes only about 60% to the total ozone destruction, whereas the remainder consists mainly of the reactions of O3 with peroxy radicals and hydroxyl radicals. The separate pathways for the destruction of ozone, O3s and O3t all maximize in the (sub)tropical lower troposphere where relatively strong insolation and high water vapor concentrations are favorable for the efficiency of photochemical reactions and production of HOx. Therefore, the relative importance of the pathway leading to OH formation is about the same for O3s and O3t. Consequently, the curves in Figure 4 illustrate the seasonality of the relative contributions by O3s and O3t to the production of OH. The hemispheric contribution from O3t is much larger than from O3s throughout the year, although the latter contribution can be up to 40% in winter. Annually, the contribution by O3s destruction to total OH in the troposphere is in the order of 15%, as also follows from the chemical destruction budgets presented in Table 1. Note that this is larger than the contribution of 5% found by Lamarque et al. . In their study, they compare simulations with and without a stratospheric ozone source for the troposphere. However, neglect of this source is likely to be partly compensated by additional photochemical production of ozone, and thus of OH. The simulated interannual variabilities are orders of magnitude smaller than the absolute seasonalities, reflecting the neglect of yearly emission fluctuations and the stable model dynamics simulated by this model version. It also expresses the effect of negative chemical feedbacks that moderate the impact of chemical variabilities on OH production [e.g., Logan et al., 1981].
 At (sub)tropical latitudes the chemical lifetime of ozone is only of the order of days, and here the production of OH is most efficient. This can be seen in Figure 5 that shows the total zonal OH distribution for April and the part that derives from ozone from stratospheric origin. Maximum OH concentrations are computed in the boundary layer and lower troposphere at approximately 10°N, where insolation and humidity are largest. Between 20°N and 30°N and 2 and 5 km altitude the absolute contribution to OH from ozone from the stratosphere is largest, about 0.5 × 106 mol cm−3 which is between 30 and 40%. This occurs not only over the oceans where humidity is relatively large but also over NE Africa and the Middle East where insolation is strongly enhanced by the relatively high surface albedo.
 STE also has an impact on OH in the extratropical upper troposphere, although here O3s is not a dominant parameter. Before mixing into the troposphere, stratospheric air in tropopause folds is relatively ozone-rich but also dry compared to the surrounding air, and is therefore characterized by relatively low OH concentrations. This suggests that monthly averaged upper tropospheric OH levels are susceptible to the annual variability in STE intensity. To illustrate this we selected the model grid coinciding with Hohenpeissenberg, Germany (48°N, 11°E), for which Figure 6 shows simulated correlations between monthly concentrations of O3, O3s, water vapor and OH at 320 hPa for April. The calculated ozone values range between 75 and 105 ppbv, which agrees with observed concentrations of 107 ± 64 ppbv [Logan, 1999; Roelofs and Lelieveld, 2000a]. The stratospheric ozone component ranges between 32 and 62 ppbv, and this variability, which is highly correlated with the synoptic patterns in the upper troposphere, contributes about 80% to the total ozone variability at this altitude. The correlation coefficient between O3 and O3s is 0.88 with a significance exceeding 99%. Generally, the efficiency of OH production increases with increasing ozone and water vapor levels. However, below the tropopause where STE strongly influences the chemical composition, higher values of O3s (and O3) are associated with lower OH and with lower specific humidity, with rather low correlation coefficients of 0.63 and 0.41, respectively. A correlation coefficient of 0.93 is found between specific humidity and OH, with a significance level exceeding 99%. We conclude that variations in the specific humidity on local or regional scales outweigh variations in ozone in determining the OH production in the extratropical upper troposphere. The same analysis applied to the whole extratropical NH upper troposphere shows that specific humidity accounts for almost 95% of OH variability at this altitude.
 Climate model calculations suggest a future increase of the STE intensity as result of climate change [Butchart and Scalfe, 2001; Land and Feichter, 2003]. We speculate that this may decrease the OH abundance in the upper troposphere, and the effect may add to or attenuate other effects on OH related with climate change, such as an increase of tropospheric water vapor concentrations [Stocker et al., 2001] and of anthropogenic emissions of CO and ozone precursors [Prather et al., 2001]. For example, higher hydrocarbons, more specifically acetone, enhance OH and HO2 levels in the upper troposphere, which may be especially relevant for the impact of subsonic aircraft emissions on the chemical composition of the upper troposphere at NH higher latitudes [Müller and Brausser, 1999; Kentarchos and Roelofs, 2002]. However, this has to be further investigated with simulations from an interactive chemistry-climate model that has a top level high enough to realistically represent the strength and the variability of the global large-scale circulation, e.g., MA-ECHAM [Manzini and Feichter, 1999].
 Results from an 11-year simulation with a 3-D chemistry-climate model were analyzed to investigate the cross-tropopause transport of stratospheric ozone, its fate in the troposphere, and the impact of STE on the tropospheric budget and distribution of ozone and OH. The NH annual cycle of STE of ozone displays a maximum in winter/early spring and a second maximum in early summer, and exhibits an interannual variability of the order 20% of the latitudinal maximum. The latitudinal distribution of the flux across the 320 hPa level displays strong subsidence of relatively ozone-rich air at subtropical latitudes.
 In winter and spring the chemical lifetime of ozone is relatively long in the extratropical troposphere, so it can be transported over large distances before it is photochemically destroyed in the subtropical lower troposphere. This destruction contributes about 15% to the OH abundance in the NH, with the largest relative contribution, ∼30%, in winter. The simulated variability of the contribution by STE appears rather weak, probably because the relatively low top level of the model precludes an accurate simulation of the variability of the Brewer–Dobson circulation. Additionally, chemical feedbacks may efficiently counteract the variability.
 In the upper troposphere the OH abundance strongly depends on water vapor concentrations. These are reduced in areas where downward STE prevails, leading to lower OH concentrations. Since the oxidation capacity is directly linked to the global large-scale circulation, it is susceptible for atmospheric dynamical consequences of climate change. These may add to or counteract other effects associated with global warming, i.e., increasing anthropogenic emissions and water vapor concentrations.
 This study is funded by the EU project STACCATO (EVK2-1999-00316). The authors thank the Dutch Computer Centre SARA (Amsterdam) for use of computer resources.