We present results using a coupled chemistry/climate model to study changes in tropospheric ozone between 2000 and 2100. We assess changes first due to the increased emissions of NOx and VOCs (using the IPCC SRES scenario A2) and then due to both emission changes and the anticipated climate change with doubled CO2. In 2100, with the scenarios used, there is a substantial calculated increase in tropospheric O3, but in contrast to earlier studies, the increase is larger for doubled CO2. The increases are most pronounced in the extratropical middle and upper troposphere; changes in circulation and a chemically induced increase in lower stratospheric O3, mainly due to reduced temperatures there, both enhance stratosphere/troposphere exchange. These changes in the lower stratosphere are crucial to our results.
 Tropospheric O3 is an important trace gas both for its role as the precursor of OH and for enhancing the greenhouse effect. Tropospheric O3 has increased considerably since preindustrial times [see, e.g., Volz and Kley, 1988; Thompson, 1992], and further increases are anticipated. Future changes of O3 in the troposphere depend on changes in the emissions of its precursors; increases in NOx and VOCs are expected to increase O3 production. With expected future climate change, the evolution of tropospheric O3 could be modified as a result of rises in temperature, humidity and convection. An increase of water vapor in the atmosphere leads to more O3 destruction through its photolysis and the subsequent reaction of O(1D) with water vapor, the dominant effect in the calculation of Stevenson et al. . On the other hand, enhanced HOx and NOx could play a significant role in O3 production, especially in the middle and upper troposphere [Prather and Jacob, 1997]. The tropospheric O3 burden also depends crucially on the flux of O3 from the stratosphere. An assessment of these processes requires treatment of the coupling between climate and chemistry. A number of previous studies using 2-D [Fuglestvedt et al., 1995] and 3-D chemical transport models (CTMs) [Brasseur et al., 1998; Johnson et al., 1999; Stevenson et al., 2000] suggest that an enhanced photochemical destruction of O3 leads to a decrease in O3 in the troposphere in response to a warmer and wetter future climate, compared to an unchanged climate. However, we note the enormous range in the tropospheric O3 budgets of models reported in the IPCC Third Assessment Report (see, e.g., Table 4.12 of IPCC [Houghton et al., 2001]) and we might expect the response of particular models to anthropogenic perturbations to depend critically on their O3 budgets. Furthermore, many of the assessments of tropospheric O3 change have ignored the change in O3 in the lower stratosphere. Increases (e.g. following the reduction in stratospheric chlorine [World Meteorological Organization, 2003]) or decreases (e.g., forced by reduced polar temperatures [Waibel et al., 1999] or due to increased stratospheric NOx [Penner et al., 1999]) have both been postulated. In the present paper, we use a global chemistry-climate model to calculate tropospheric O3 for the present day and for the end of the 21st century, including a simplified treatment of the lower stratospheric O3 change.
2. The Climate-Tropospheric Chemistry Model
 We use the U.K. Meterological Office Unified Model for climate and weather prediction (UM) [Cullen, 1993; Johns et al., 1997; Senior and Mitchell, 2000], coupled to a detailed tropospheric chemistry scheme. The model has 19 levels from the surface up to 4.6 hPa at a horizontal resolution of 3.75° by 2.5° and is forced using prescribed sea surface temperatures (SSTs). We have adopted a tracer advection scheme by A. R. Gregory (private communication, 2000), based on Leonard et al. , which is conservative, monotonic, and accurate. Convection is based on a penetrative mass flux scheme [Gregory and Rowntree, 1990]. The chemical mechanism includes 46 species, 24 of which are advected, and 186 reactions describing CO, CH4 oxidation and higher hydrocarbon degradation for ethane and propane. This scheme also includes the main stratospheric cycles with the exception of any halogen chemistry. Chemical integrations are carried out using an implicit time integration scheme [Carver and Stott, 2000] with a 15 minute time step. Dry and wet deposition schemes for trace species are included in the model [see Giannakopoulos et al., 1999]. The model uses the diurnal varying photolysis rates calculated off-line in a 2-D model [Law and Pyle, 1993]. Daily concentrations of the stratospheric species O3 and NOy are specified above 50 hPa (the top three model levels) using output from the 2-D model, to produce a realistic annual cycle of these species in the stratosphere. The identical chemistry scheme has been used in our off-line CTM TOMCAT [Law et al., 1998] and some aspects of that model's performance are shown in the IPCC report [Houghton et al., 2001]. Model performance, in both TOMCAT and the UM, compares well with observations and other models for the main tropospheric species. Indeed, comparison with observed O3 in the upper troposphere is rather better for the UM than for TOMCAT. The Edwards and Slingo  radiation code is used in the UM. The model uses present day climatological O3 fields [Li and Shine, 1995] in the radiation calculations for all the scenarios considered and so does not allow for feedback of calculated O3 into the GCM.
 We perform three calculations: Scenario A with year 2000 emissions and present-day climate forcing, using observed SSTs; Scenario B with year 2100 emissions and present-day climate forcing; and Scenario C with 2100 emissions and 2100 double CO2 climate forcing with SSTs produced by the Hadley Centre coupled ocean-atmosphere GCM run with SRES A2 emissions (C. E. Johnson, private communication, 2001). The data for seasonally varying emissions of NOx and VOCs used in the runs are also based on A2 scenarios from the IPCC Special Report on Emission Scenarios (SRES) for the years 2000 and 2100 [Nakićenović et al., 2000]. The A2 scenarios have large increases in NOx and VOC emissions during this century. This was the scenario used by Stevenson et al.  and is used again here for direct comparison with that study. The total annual emissions are listed in Table 1. Seasonal partitioning factors for biomass burning used in both emission scenarios are from Hao and Liu . Stratospheric upper boundary conditions for O3 and NOy are the same for all model runs. Aircraft emissions are 0.7 Tg(N) yr−1 for 2000 and 2.3 Tg(N) yr−1 for 2100. Lightning-produced NOx is calculated using the parameterization of Price and Rind  and scaled to 4 Tg(N)yr−1 for the present day. This value depends on climate since lightning activity is linked to the frequency of convection in the model. We find a 10% increase in NOx production due to the increase of convection in scenario C compared with B and A. In the radiation scheme in both climate scenarios, all trace gas mixing ratios including O3 are fixed at present day levels except for CO2 which is doubled for 2100 climate forcing. The model was run for 28 months. Initial fields of model species in each run are taken from comparable multi-year runs of our off-line model TOMCAT. CH4 initial fields are the same for runs B and C.
Table 1. Emission Scenarios for the Years 2000 and 2100 (Units are Tg(N) yr−1 for NOx and Tg yr−1 for Other Species)
Including surface, aircraft and lightning emissions (see text).
Figure 1a shows the annual and zonal mean distribution of O3 for run A. The main features of the observed O3 distribution are reproduced; the model compares reasonably well with observations in both structure and magnitude (see, e.g., the climatology produced by Fortuin and Kelder ) The annual and zonal mean changes of tropospheric O3 between 2000 and 2100 with present-day climate (i.e. B–A) are shown in Figure 1b. O3 increases are found throughout the troposphere. The largest increases, exceeding 30 ppbv, occur in the upper troposphere, at middle and high latitudes. At the surface, an increase of over 10 ppbv occurs in the tropics and subtropics with over 20 ppbv in the NH subtropics, controlled by the increase of emissions. This agrees well with Stevenson et al.  using the same SRES A2 emission scenarios.
 A double CO2 climate forcing (Scenario C) leads to temperature rises of between 3K at the surface and over 8K in the upper tropical troposphere. In the lower stratosphere, a cooling reaching 4K at the poles is evident (very similar to that shown in Shindell et al. [1998, Figure 3]). The specific humidity increases by 20% in the boundary layer, by over 50% in the middle troposphere, and by up to 100% in the upper tropical troposphere. The O3 change in C–A (Figure 1c) shows a different pattern to B–A. In the tropics, at the surface and in the upper troposphere, increases in O3 are reduced compared to B–A. However, increases, of over 40 ppbv, are larger in the middle and the upper extratropical troposphere. This contrasts with previous studies [e.g., Fuglestvedt et al., 1995; Brasseur et al., 1998; Johnson et al., 1999; Stevenson et al., 2000] which find a generally enhanced destruction of tropospheric O3 in a warmer and wetter climate, so that the net increase in O3 calculated for the end of this century is reduced when climate change is considered.
 To understand these changes in detail we calculated the global tropospheric O3 budget for the three scenarios (see Table 2). Note that the net flux from stratosphere to troposphere is diagnosed here by assuming that the net flux balances exactly with net in situ chemical production and surface deposition. A very small difference (less than 1%) between the O3 burdens in the last two years of our integration justifies this approximation. We have calculated the fluxes at altitudes below a chemical tropopause defined by the 150 ppbv contour of O3 in run A.
(Budget/burden calculated in regions with O3 < 150 ppbv).
stratosphere-troposphere exchange (Tg/yr).
net tropospheric O3 production (Tg/yr). Values in brackets are chemical production and destruction.
surface deposition (Tg/yr).
800 (4102, 3302)
1486 (7405, 5919)
1090 (8236, 7146)
 Compared to run A, both tropospherically averaged production (following increases in HOx and NOx) and destruction (due to increased water vapor and O3) increase in run B and run C. Production and destruction rates are both largest in run C. Overall, however, the increase of in situ net O3 production is smaller with climate change included. In a warmer and wetter climate, stronger destruction occurs because of a large increase of water vapor and more than compensates for the increased production.
 Despite a decrease in net O3 production in C compared to B, the tropospheric O3 burden in C is greater. This is the result of an 80% increase in STE in the changed climate. There are two reasons for this. Firstly, there is more O3 in the lower stratosphere in run C. Secondly, there is also enhanced downward transport from the stratosphere in run C (see also W. J. Collins et al., The effect of stratosphere-troposphere exchange on the future tropospheric ozone trend, submitted to Journal of Geophysical Research, 2002). In experiment B, there is an O3 increase in the lower stratosphere, compared to experiment A, due to enhanced NOx-driven production of O3. The chemical tendency of O3 in the very low stratosphere is a balance between this production and destruction due to the HOx catalytic cycle. When the temperature is reduced, as it is in experiment C, the OH concentration falls due to the strong negative temperature dependence of the reactions OH + HO2 and OH + HNO3, as first described by Eckman et al. . This reduces the destruction of O3 by HOx, the dominant chemical effect leading to enhanced lower stratospheric O3 in run C (see Figure 1d - away from the tropical lower stratospheric O3 is more than 100 ppbv greater in run C than B). This temperature effect accounts for 25–50%, depending on season, of the total increase of O3 (C–B) in the lower stratosphere. In addition, O3 increases because of an enhanced meridional circulation in the double CO2 climate [see Rind et al., 2001; Butchart and Sciafe, 2001]. There is a decrease in O3 in run C in the tropical lower stratosphere (and upper troposphere). Tracer experiments show that there is increased ascent in this region (a stronger Brewer-Dobson circulation), which, with the O3 concentration increasing with altitude, leads to the O3 reduction there. This enhanced circulation contributes to the increased O3 in the extratropical lower stratosphere, and the enhanced stratosphere to troposphere exchange there.
Figure 1d, which shows differences between runs B and C, highlights the effect of climate change on O3 in the troposphere. O3 destruction due to increased water vapour in C dominates in the tropical lower troposphere. The increased transport from stratosphere to troposphere is most evident at the middle latitude tropopause. This increase dominates over the enhanced O3 destruction in the warmer, wetter climate leading to O3 increases in the free troposphere of about 5 ppbv compared to run B. There is also a net O3 decrease in the tropical upper troposphere, despite enhanced chemical production there, associated with the stronger vertical transport in run C.
 We have calculated changes in tropospheric O3 between 2000 and 2100 using emissions based on IPCC SRES A2 scenarios describing the years 2000 and 2100, with and without the changes of meteorology associated with a doubling of CO2. Increases in NOx and VOC emissions result in an increase of tropospheric O3 production in both cases, with compensating O3 destruction. The net change is climate-dependent. Anticipated future changes in climate are expected to modify changes in tropospheric O3 following changes in temperature, humidity and circulation. Increases in humidity will promote O3 destruction, while increases in HOx and NOx enhance its production. In contrast to previous studies, a larger increase of O3 in the extratropical middle and upper troposphere is found in our calculations in a double CO2 climate, compared to an unchanged climate. A major factor is an enhanced flux of O3 from the stratosphere in the changed climate, following reduced O3 loss by HOx reactions in the colder lower stratosphere and the intensified circulation. Our treatment of lower stratospheric chemistry is not complete since chlorine chemistry is ignored. Inclusion of chlorine could further enhance the O3 increase in the lower stratosphere between run A and runs B and C, as chlorine concentrations decrease during this century. A cooling of the polar stratosphere could however enhance chlorine-induced O3 loss there [Waibel et al., 1999]. Our results in both the troposphere and lower stratosphere are dependent on the emission scenarios chosen, which themselves could be climate dependent, and may also be sensitive to, e.g., model spatial resolution in the UTLS. Furthermore, we have presented only a single, snapshot realisation in experiments A, B and C. Nevertheless, our main finding is robust: changes in the low stratosphere, with its large reservoir of O3, are likely to have a crucial influence on the evolution of tropospheric O3 during this century.
 The response of the change in tropospheric O3 to climate change is complex. The response is dependent on the O3 budget which is currently poorly constrained. Note that for the models reported in the IPCC [Houghton et al., 2001], there is a range of net production values from −855 Tg/yr to +507 Tg/yr, with the large differences in STE being the driving force behind whether a model calculates a chemical sink or source of tropospheric O3. Models can reproduce satisfactorily many observed features of tropospheric chemical composition but nevertheless have very different budgets. The treatment in different models of the upper boundary, and stratosphere to troposphere exchange, must be a major area of uncertainty in assessment modelling. These budget uncertainties limit our ability to predict the future evolution of the chemistry/climate system.
 This work was supported by the NERC UTLS thematic programme and the NERC Centres for Atmospheric Sciences (NCAS). The UK Met Office Hadley Centre is thanked for the use of the UM. We are grateful to Andrew Gregory for providing the new advection scheme in the UM. We acknowledge many other colleagues in Cambridge at the Centre for Atmospheric Science.