Impacts of climate change, ozone recovery, and increasing methane on surface ozone and the tropospheric oxidizing capacity



This article is corrected by:

  1. Errata: Correction to “Impacts of climate change, ozone recovery, and increasing methane on surface ozone and the tropospheric oxidizing capacity” Volume 119, Issue 8, 5028–5036, Article first published online: 17 April 2014

Corresponding author: O. Morgenstern, NIWA, Private Bag 50061, Omakau, 9352, New Zealand. (


[1] Using a stratosphere-troposphere chemistry-climate model, we compare the impacts of climate change, stratospheric ozone recovery, and methane increases on surface ozone and the tropospheric oxidizing capacity by 2050. Methane increases lead to a decreasing OH, particularly in the northern subtropics during summer. Stratospheric ozone recovery causes small increases of surface OH driven by increased stratosphere-troposphere exchange, occurring during parts of the year in the southern extratropics. Tropospheric warming is also associated with increasing OH, maximizing in the Northern Hemisphere in northern summer. In combination, OH is anticipated to decrease by approximately 8% in the tropospheric average by 2050 in the scenario considered here. In conjunction with these changes to OH, we model substantial changes in surface ozone in both hemispheres. Methane increases alone will lead to increasing surface ozone by up to 2–3 ppbv in the zonal mean, maximizing around 30°N. This increase is exacerbated during austral winter when increased stratosphere-troposphere flux of ozone causes an increase in surface ozone in the southern extratropics. Both increases are partially offset by decreases in surface ozone of up to 2 ppbv in the zonal mean, with substantial zonal asymmetries, due to global warming. We model substantial changes in the methane lifetime caused by the three factors. In the Arctic during summer, disappearing sea ice, in an ice-albedo feedback, causes substantially reduced surface ozone. Of the three factors considered here, methane increases are found to exert the strongest influence on surface ozone.

1 Introduction

[2] During the 21st century, atmospheric composition is expected to change due to climate change [e.g., Stevenson et al., 2006], ozone recovery due to the implementation of the Montreal Protocol [e.g., World Meteorological Organization (WMO), 2007], continuing industrialization, land use change, and decreasing anthropogenic emissions of some anthropogenic pollutants due to the implementations of air quality standards. Tropospheric ozone is governed by a chemical-dynamical balance involving chemical production and loss, surface dry deposition, and stratosphere-troposphere exchange (STE) of ozone [Stevenson et al., 2006]. In most scenarios, methane is anticipated to increase at least to midcentury [Intergovernmental Panel on Climate Change (IPCC), 2000; SPARC CCMVal, 2010], driven among other factors by intensifying animal husbandry; such an increase would affect ozone throughout the atmosphere because methane is an important long-lived ozone precursor. All but the most optimistic of the new representative concentration pathway (RCP) projections for greenhouse gases have methane increasing at least to 2050 [Meinshausen et al., 2011]. We expect trends in the oxidizing capacity, i.e., the capacity of the troposphere to cleanse itself of anthropogenic and natural pollutants [Derwent, 1996]. Mostly, such removal occurs due to oxidation by the hydroxyl (OH) radical. OH is formed upon removal of tropospheric ozone by way of O3 +  → O(1D) for λ<328 nm, followed by O(1D) + H2O → 2OH. Production of OH (or HOx = HO2 + OH) through this mechanism hence depends, among other factors, on ozone photolysis, which is a function of the overhead ozone column [Fuglestvedt et al., 1994], and on atmospheric moisture, expected to increase under global warming. Methane is a substantial sink of OH. Hence ozone recovery, climate change, and methane trends are factors that affect the OH abundance. STE of ozone is expected to increase under both climate change [Zeng et al., 2008] and ozone recovery [Hegglin and Shepherd, 2009]; Zeng et al. [2010] suggested that by 2100, both ozone recovery and climate change will contribute about equally to an increase in wintertime surface ozone over the southern extratropics, but find no such significant large-scale change in the Northern Hemisphere (NH). The sensitivity of surface ozone to changes in the overhead ozone column depends on tropospheric pollution levels.

[3] Hegglin and Shepherd [2009] and Zeng et al. [2010] used chemistry-climate models (CCMs) exhibiting certain structural shortcomings affecting the stratospheric influence on tropospheric ozone and the oxidizing capacity. Hegglin and Shepherd [2009] used a stratospheric CCM that does not explicitly calculate chemistry below 400 hPa. In polluted regions, surface ozone is dominated by NOx-catalyzed ozone production, which anticorrelates with the ozone column, whereas in clean, low-NOx regions, ozone photolysis and associated loss anticorrelate with the ozone column, which causes surface ozone to correlate positively with the ozone column [Isaksen et al., 2005]. We note that the model used by Zeng et al. [2010] did not include any dependence of photolysis rates on the overhead ozone column and hence would not have reflected these effects. Also, in their climate-change experiment, Zeng et al. [2010] did not consider methane changes.

[4] John et al. [2012] and Voulgarakis et al. [2012] studied the links between methane and OH changes and climate change. They found that in the most extreme scenario (RCP8.5), the methane lifetime is expected to increase, driven in a positive feedback by increasing methane emissions, but in the other RCP scenarios the methane lifetime is influenced by the emissions of short-lived climate agents, which often drive increases in OH and decreases in the methane lifetime.

[5] Here we revisit the question of how the oxidizing capacity, and tropospheric ozone, will change under the relative influences of projected climate and methane changes and stratospheric ozone recovery. Like Zeng et al. [2010], we ignore future changes in surface emissions of NOx, CO, and nonmethane volatile organic compounds (NMVOCs). Zeng et al. [2008] find that by 2100, emission trends will cause a substantial increase in tropospheric ozone; however, in view of the uncertainty in the projected emissions, we do not study this topic here. They also find substantial tropospheric ozone changes due to climate change, linked to increasing humidity and increasing STE of ozone; we will compare our results to theirs. In contrast to Zeng et al. [2010], we use a troposphere-stratosphere CCM comprising a detailed interactive photolysis scheme [Neu et al., 2007; Telford et al., 2012] accounting for clouds, aerosols, surface albedo, and the overhead ozone column.

2 Model and Simulations

2.1 The Model

[6] We use the Met Office Unified Model coupled to the U.K. Chemistry and Aerosols (UKCA) module [Morgenstern et al., 2008, 2009, 2010]. Telford et al. [2012] presented a description of a recent version of the model in a configuration similar to the one used here; below, we therefore only give a brief description. The background climate model is in a configuration similar to HadGEM3-A [Hewitt et al., 2011] at a resolution of 3.75° × 2.5° and 60 levels extending to 84 km. The vertical resolution of the model varies between 20–240 m below 1 km, ∼1 km in the tropopause region, to ∼6 km at the top of the model domain. Chemistry is a merger between the stratospheric mechanism described by Morgenstern et al. [2009] and the tropospheric mechanism used by Zeng et al. [2008] (but excluding the isoprene oxidation mechanism used by Zeng et al. [2008, 2010]); the scheme used here comprises 62 predicted species, 49 photolysis, and 196 thermal and heterogeneous reactions. Note that the UKCA model is designed as a whole-atmosphere model, i.e., for chemistry no explicit distinction is made between stratospheric and tropospheric processes [Morgenstern et al., 2009]. Therefore, the merger of stratospheric and tropospheric chemistry used here is straightforwardly achieved by essentially extending, within a symbolic solver framework [Carver et al., 1997], the mechanism used by Zeng et al. [2008] with reactions listed by Morgenstern et al. [2009] that are not part of the Zeng et al. [2008] tropospheric mechanism. The model used by Zeng et al. [2008] is also based on the same framework. We emit seven primary pollutants at the surface, namely NO, CO, CH2O, C2H6, CH3CHO, C3H8, and CH3COCH3. As noted above, isoprene chemistry is not included and isoprene is emitted as CO [Zeng and Pyle, 2003]. Surface and aircraft emissions (including those of isoprene) vary with season at a monthly resolution, are annually repeating, are the same for all simulations, and follow Zeng et al. [2010]. In the chemistry part of the model, for N2O, methane, H2, and organic halogens we assume uniform, constant lower boundary conditions (see below). Chlorine and bromine source gases are lumped; here we only explicitly model CFCl3, CF2Cl2, and CH3Br, which are increased versus their real atmospheric abundances to account for the halogen source gases not explicitly included in the chemistry scheme. Aerosol and aerosol precursor emissions are representative of the year 2000. Lightning emissions of NOx are interactively coupled to convective activity in the model [Price and Rind, 1992] and will therefore differ between the different simulations detailed in section 2.2. The model uses the FAST-JX interactive photolysis scheme [Neu et al., 2007; Telford et al., 2012] with a correction added above 60 km for wavelengths shorter than 177 nm [Lary and Pyle, 1991]. Interannually varying sea surface temperature and sea ice data, covering the years 1985–2004 and 2035–2054, respectively, are taken from a simulation of the HadGEM1 atmosphere/ocean climate model following the A1b scenario, the same as used for the CCMVal-2 model intercomparison [IPCC, 2000; Johns et al., 2006; Morgenstern et al., 2010; Stott et al., 2006]. Simulated ozone and water vapor feed back onto radiation and the hydrological cycle, respectively. All other long-lived greenhouse gases (LLGHGs, namely CO2, CH4, N2O, CFCl3, CF2Cl2, CF2ClCFCl2, and CHF2Cl) can differ between radiation and chemistry and are assumed uniform for radiation. This allows us to treat their radiative and chemical impacts separately. CO2 is listed as a chemical agent mainly for completeness; its only significant role in the modeled chemistry is as a source of CO in the upper atmosphere. CF2ClCFCl2 and CHF2Cl are lumped with CF2Cl2 and CFCl3, respectively, for chemistry.

2.2 Simulations and Method of Evaluation

[7] We perform six time-slice simulations. Essentially, the six simulations only differ from each other regarding the assumptions made about (a) radiative forcing and associated sea-surface conditions, (b) the methane loading, and (c) ozone-depleting substances. This means that differencing pairs of these simulations will allow us to isolate the impact of changes in one or more of these processes onto any chosen diagnostic. In particular, this choice of simulations allows us to separate climate-change impacts from chemically induced impacts on tropospheric composition.

[8] All simulations cover 20 years; here we use the last 10 of these for evaluation. Forcing data for halocarbons are taken from the WMO [2007] A1 scenario and those for CO2, methane, and N2O from IPCC [2000] A1b (as used by SPARC CCMVal [2010]). The A1b scenario gives a similar radiative forcing to RCP6.0 but projects substantially more methane for 2050 than RCP6.0 [Meinshausen et al., 2011]. Details are listed in Table 1. The simulations are: a reference (REF) simulation representing the atmosphere of the year 2000 with LLGHGs and ozone-depleting substances as in the above two scenarios; a “high-methane” simulation identical to REF but assuming 2050 methane (2.4 ppmv) for chemistry (HM); two simulations identical to REF but assuming “ozone recovery” (i.e., decreased total chlorine (2.2 ppbv) and bromine (15 pptv) and N2O increased to 350 ppbv), with low (1.8 ppmv) and high (2.4 ppmv) methane, respectively (ORLM and ORHM); a “climate change” simulation with chemistry as in REF but assuming 2050 radiative forcings from LLGHGs and corresponding sea surface forcings (CC); and an “ozone recovery and climate change” simulation (ORCC) with chemistry and climate both representative of the year 2050 (including high methane of 2.4 ppmv). NMVOC, CO, and NOx surface and aircraft emissions are identical in all simulations (section 2.1). As stated before, CO2 is assumed uniform in all cases. In three of the simulations listed below the uniform mixing ratios differ between radiation and chemistry.

Table 1. Simulations and Associated Prescribed Source Gas Loadingsa
  1. a

    Left columns: uniform LLGHG volume mixing ratios assumed in radiation. “0” means the species is not considered. Right column, in bold: surface (in the case of CO2: Bulk) mixing ratios assumed in chemistry if different from those used in radiation. Note the effect of lumping on the CFC mixing ratios for chemistry. The “SST/ice” column refers to the years of sea surface forcing data used here from a HadGEM1 simulation of the IPCC [2000] A1b scenario. CF2ClCFCl2 and CHF2Cl do not figure directly in chemistry, and CH3Br is not considered in radiation. For reference, we cite the values for RCP6.0 [Meinshausen et al., 2011].


[9] Note that the RCP scenarios for 2050 have methane ranging from 1.5 (RCP2.6) to 2.7 ppmv (RCP8.5), N2O from 342 to 367 ppbv, and CO2 from 443 to 541 ppmv [Meinshausen et al., 2011].

3 Basic Validation of the Modeled Ozone Field

[10] Here we present comparisons of the ozone field in REF with available observational data. This is not a complete model validation; a more comprehensive validation, with a focus on stratospheric aspects of an earlier version of the model, has appeared [Morgenstern et al., 2009].

[11] Figure 1 shows the zonal-mean ozone column in REF. The model climatology is similar to that published by Morgenstern et al. [2009] based on an earlier version of UKCA without explicit tropospheric chemistry. The model represents the basic features of the ozone column field, in particular the Antarctic ozone hole, the two maxima in northern and southern midlatitude winter, and the tropical ozone column. As is typical for stratospheric chemistry-climate models, there are some biases in the total-column ozone field [Austin et al., 2010]. In particular, relative to the Total-Ozone Mapping Spectrometer/Solar Backscatter Ultra-Violet (TOMS/SBUV) satellite ozone climatology [Bhartia et al., 1984], the model is biased high, with biases ranging from around 20 Dobson Units (DU) in the tropics to 40 DU in northern winter and in southern middle latitudes during spring. The modeled mean column within the ozone hole (of 140–160 DU) well matches that in the TOMS/SBUV climatology. Also in the Arctic, the model produces a depression in the total ozone field in February-March associated with polar ozone depletion.

Figure 1.

Colors: 10 year, zonal, and monthly mean total column ozone (in Dobson Units, DU) in REF. Contours: TOMS/SBUV climatology, averaged over 1995–2004.

[12] Figure 2 shows the monthly and zonal cross -section of ozone versus available climatologies [Fortuin and Langematz, 1995; Hassler et al., 2009] and the model bias relative to these climatologies. The modeled ozone is similar to that published by Morgenstern et al. [2009]. Ozone in the tropopause region is quite realistic. Note that large relative errors in the tropical tropopause region correspond to only small vertical displacement errors because ozone exhibits steep vertical gradients in this region. A substantial tropospheric high bias of ozone apparent in the earlier version of UKCA [Morgenstern et al., 2009], which appeared despite the absence of NMVOC chemistry, has been corrected in this version; tropospheric ozone is now both high- and low-biased, in different regions. At high latitudes of both hemispheres, tropospheric ozone is now lower than in the Fortuin and Langematz [1995] climatology (see below). The high bias of near-surface ozone apparent at middle and low latitudes is inconsistent with the station-based analysis performed below. This may exemplify the problem of compiling a zonal-mean climatology of tropospheric ozone when tropospheric ozone is highly variable, difficult to observe from space, and poorly covered by ground-based observations.

Figure 2.

(top) Colors: 10 year, zonal, and monthly mean ozone (ppmv) in REF, in (left) February) and (right) September. Black contours: NIWA vertically resolved ozone database [Hassler et al., 2009], averaged over 1995–2004. White contours: Ozone climatology after Fortuin and Langematz [1995]. (bottom) Percentage bias of modeled zonal and monthly mean ozone, relative to the climatologies, in February and September.

[13] Figure 3 shows the mean annual cycle of surface ozone at selected stations, and the corresponding results of simulation REF. The comparison suggests that the model produces a mostly realistic surface ozone field except at high latitudes of both hemispheres. At Arrival Heights (and also at other Antarctic stations not shown here) the model substantially underestimates ozone, and at Barrow, Alaska, the model produces an unrealistic seasonal cycle. Both may be due to missing chemistry such as the role of nitrogen storage in snow, and heterogeneous reactions on the snow surface. More research is needed to understand the processes that determine surface ozone at high latitudes.

Figure 3.

(symbols) Observed monthly- and multiannual-mean surface ozone at selected stations. The data are from the World Data Center for Greenhouse Gases and cover the years 1995–2004 unless indicated otherwise. (lines) Monthly and multiannual mean ozone in REF at the same stations. The model has been sampled at the elevation of the station or the lowest model level, whichever is higher. At sea level the model surface level has a thickness of 20 m.

[14] Figures 1 to 3 suggest that the model produces a satisfactory, albeit not perfect, representation of ozone throughout the troposphere and stratosphere.

4 Sensitivity to Climate Change, Ozone Recovery, and Increasing Methane

[15] In the following, we define “significantly different” to mean that two time series have significantly different means at the 95% confidence level, using Student's t-test.

4.1 Ozone Column

[16] Figure 4 shows the sensitivity of the total column ozone field to the various boundary conditions explored in our experiments. Figure 4a indicates that increasing methane tends to increase the ozone column to a small extent but almost universally. There are substantial increases in ozone both in the troposphere and the stratosphere (section 4.2), but the stratospheric changes generally contribute more to the total-column change (not shown). The increases are statistically significant during summer and early autumn in much of the northern extratropics, and also during the ozone hole season. There is a little less ozone depletion in the Antarctic ozone hole under methane increases but the ozone hole gets more long-lived. As a consequence, in December the methane increase actually causes column ozone to decrease over the Antarctic, but due to large variability this decrease is not statistically significant.

Figure 4.

Difference in zonal, monthly, and multiannual-mean total ozone column between the pairs of simulations as indicated in the titles, in DU. “plus” symbols indicate that the two time series do not have a significantly different mean.

[17] Climate change without the associated chemical changes (Figure 4b) produces increases of the ozone column in the northern extratropics and in southern middle latitudes consistent with stratospheric cooling, a slow-down of gas phase ozone depletion, and a speed-up of the Brewer-Dobson Circulation. Also decreases of model ozone during Antarctic winter and spring are consistent with stratospheric cooling causing more vigorous chlorine activation on more abundant polar stratospheric clouds. However, by 2050 these changes will be generally too small to be significant. Significant changes in the ozone column ensue for the three simulations that assume reduced halogen loadings in the stratosphere (ORHM, ORLM, ORCC; Figures 4c, 4d, 4f, 4h). Of these, ORHM has somewhat more ozone than ORLM (in analogy to Figure 4a), and ORCC (which assumes 2050 LLGHG loadings for radiation, the same as in CC) has more ozone than ORHM at high latitudes but less ozone in the tropics. These results are consistent with CCMVal-2 [SPARC CCMVal, 2010; Austin et al., 2010; Eyring et al., 2010].

4.2 Ozone Cross-Section

[18] Figure 5 shows differences in annual- and zonal-mean ozone throughout the model domain. Increasing methane (Figures 5a and 5e) leads to an increase of ozone in the troposphere but a decrease in the mesosphere. In the troposphere, this reflects the well-known coupling of organic and NOx chemistry to produce ozone. In the mesosphere, the additional methane increases the total hydrogen loading, which promotes HOx-catalyzed ozone depletion. In the stratosphere, the changes are mostly insignificant. Reducing halogen and increasing N2O increases ozone in most of the stratosphere and in the extratropical troposphere (Figures 5d and 5h), reflecting reduced ozone depletion by halogen compounds, which outweighs increased ozone depletion due to increased NOx. In the tropical troposphere and mesosphere the impacts are mostly insignificant. Increases in extratropical tropospheric ozone are due to increased cross-tropopause flux of ozone. Increasing the radiative forcing due to LLGHGs (Figures 5b and 5g) reduces ozone in the tropics and subtropics around 20 km but increases ozone elsewhere in the stratosphere. This is due to cooling of the stratosphere, which slows down ozone depletion, and to a speed-up of stratospheric overturning and tropical upwelling, which transport ozone more quickly from low to middle and high latitudes [Butchart et al., 2010]. In the mesosphere and in much of the troposphere, ozone decreases, but less so under low-halogen conditions. In combination (Figure 5f), the model predicts increased ozone everywhere except in the tropical upper-troposphere-lower-stratosphere region and in the mesosphere.

Figure 5.

Difference in zonal and annual mean ozone between the pairs of simulations as indicated in the titles, in ppmv.

4.3 The Near-Surface Ozone Photolysis Rate

[19] Changes in total-column ozone affect surface composition in a variety of ways, one of which is through changes in ultraviolet radiation and hence the rates of photolysis (see section 1). Ozone recovery in the three simulations assuming reduced halogen (ORLM, ORHM, ORCC) does indeed result in generally decreased rates of photolysis of ozone in the O3 + hv → O(1D) channel by roughly 1% in summer, but those changes are mostly insignificant for the 10 year runs considered here (Figure 6). Significant decreases of approximately 10% occur in the near-surface O3 → O(1D) photolysis rate during summer over sea in the Arctic and Antarctic (in the CC and ORCC simulations); these are due to a reduction of sea ice cover in these runs under climate change (of about 2/3 in August), which causes the surface albedo to decrease [Voulgarakis et al., 2009]. Note that in the model, sea ice albedo is in the range of 0.57–0.8, depending on snow cover. Sea ice albedo is much larger than that of the open ocean, which in the model follows Barker and Li [1995] with updated coefficients Hewitt et al. [2011]. Also, the longer lifetime of the Antarctic ozone hole in HM (Figure 4a) results in an increase in this photolysis rate by more than 10% during early summer over Antarctica although this increase is, like the decrease in the ozone column, insignificant. Elsewhere, it appears variability in the photolysis rate caused by other factors that affect tropospheric photolysis, namely clouds and aerosols, may easily mask any impact of large-scale changes such as ozone recovery on the rate of ozone photolysis (Figure 6).

Figure 6.

Same as Figure 4 but for the difference in zonal, monthly, and multiannual-mean rate of photolysis of ozone to form O(1D), in 10–3 s–1.

Figure 7.

Same as Figure 4, but for the zonal and monthly mean surface to 10 km partial OH column, in 109 molecules cm− 2.

4.4 The Hydroxyl Radical

[20] Figure 8 shows the zonal-mean surface-to-10 km OH partial column in REF. The OH column is calculated off-line, based on the monthly mean fields of the OH mixing ratio, temperature, and pressure. Covariances between these quantities that occur on submonthly timescales are therefore not reflected in the column displayed here. OH peaks at about 2.2 ⋅ 1012 molecules cm− 2 around 30°N during summer, reflecting larger background ozone in this region relative to the equivalent season and latitude in the Southern Hemisphere, causing more OH production in the mechanism explained in section 1. Also, many NMVOCs are net sources of HOx, adding to the greater abundance of OH in the Northern Hemisphere. OH maximizes during summer because of the summertime maximum in UV. By contrast, during winter OH is low because of low radiation, humidity, and ozone.

Figure 8.

Same as Figure 1, but for the surface to 10 km OH column, in 1012 molecules cm–2, in REF.

[21] We assess the changes in tropospheric OH in our sensitivity simulations (Figure 7). The increase in methane in simulations HM, ORHM, and ORCC results in a general decrease of OH maximizing (in absolute terms) at more than 0.2 ⋅ 1012 molecules cm− 2, a ∼10% decrease (Figures 7a, 7c, 7e, 7f). The decrease is significant in regions where there is substantial OH. For reference, Voulgarakis et al. [2012] find a comparable reduction in tropospheric OH of ∼8.5% for the RCP 8.5 scenario (which has a methane mixing ratio of 2.7 ppmv in 2050).

[22] Climate change alone causes OH to increase (by less than 0.18 ⋅ 1012 molecules cm− 2), particularly in northern middle latitudes during summer (Figures 7b and 7g). The main reason for this is increased humidity under climate change. The finding is broadly consistent with John et al. [2012] who also found that climate change alone drives an increase in OH. A quantitative comparison is however complicated by differences in scenario and period studied. Significant decreases of surface OH occur in the Arctic during summer despite an increase in humidity there by <1% (not shown). The decrease in Arctic OH is consistent with the decrease in the ozone photolysis rate (section 4.3) and suggests that the sea ice-albedo effect on OH is more important in our model than the effect of increasing humidity. Voulgarakis et al. [2009] found, for complete removal of late-summer sea ice, a reduction in OH by about 50% in the Arctic; considering that in our simulations, sea ice does not completely disappear in 2050, we find a comparable reduction in OH. OH in ORCC exhibits smaller decreases in OH in the tropics than in ORHM, resulting from a substantial cancellation of methane-driven decreases and climate-change and ozone-recovery driven increases. Stratospheric ozone recovery alone (Figure 7d) drives significant increases of tropospheric OH during summer in the middle latitudes of both hemispheres of up to around 0.1 ⋅ 1012 molecules cm− 2. The increase suggests that enhanced cross-tropopause flux of ozone under ozone recovery discussed below outweighs any impact due to reduced photolysis rates and associated reduced tropospheric ozone production. Tropospheric OH is more sensitive to ozone recovery under low-methane than under high-methane conditions (Figures 7d and 7h), possibly because of the improved buffering of OH under high-methane conditions.

4.5 Surface Ozone

[23] Various feedback mechanisms link changes in surface OH to changing surface abundances of ozone, including the ozone photolysis sink described in section 1, as well as an altered efficiency of the HOx ozone loss cycle (O3 + OH → HO2 + O2; HO2 + O3→ OH + 2 O2) important in the upper troposphere, and possibly a changed rate of decay of volatile organic compounds and the associated production of tropospheric ozone. Figure 9 shows the sensitivities of surface ozone to the perturbations studied here. Methane increases cause increases of surface ozone, maximizing around 30°N at up to 2–3 ppbv (Figures 7a and 7e). Climate change alone causes partially significant decreases of up to 2 ppbv, with substantially larger regional decreases and also some increases, see below); generally, the decreases are larger in the north than in the south (Figures 7b and 7g). Slightly larger decreases are found over the Arctic Ocean in summer; these are likely associated with reduced ozone production due to the ice-albedo effect discussed above. Stratospheric ozone recovery alone increases ozone by up to around 1–2 ppbv in both extratropical regions during autumn and winter (Figure 7d). In the Southern Hemisphere, this feature is very similar to the increase found by Zeng et al. [2010] and is thought to be driven by increased transport of ozone from the stratosphere under ozone recovery. In the Northern Hemisphere, such increased transport may contribute during winter, but particularly in summer the increase is also the result of decreased photochemical loss of ozone under stratospheric ozone recovery in relatively clean continental regions (see below; Isaksen et al. [2005]). As found before for OH, the sensitivity of surface ozone to stratospheric ozone recovery is larger under low-methane than high-methane conditions (Figure 7h). All three factors appreciably affect surface ozone in an approximately additive way in the combined simulation (ORCC; Figure 7f).

Figure 9.

Differences in monthly zonal, and multiannual mean surface ozone (at 10 m) between pairs of simulations as indicated, in ppbv. Stippling indicates insignificance.

[24] While displays of zonal-mean ozone (Figure 9) give an overall impression about the seasonality and magnitude of the effects on surface ozone, they mask a lot of geographic structures in the signals. To address this, Figure 10 shows surface ozone during June-July-August when high-ozone episodes are of concern in densely populated areas of the Northern Hemisphere. As noted before, the methane increase is responsible for increased ozone almost everywhere and over both land and sea (Figures 9a and 9e). Climate change causes a decrease over many low-latitude ocean regions, over some land areas including Greenland, and over the Arctic Ocean (Figures 9b and 9g); these decreases outweigh some regional increases of ozone in polluted regions. The pattern of change caused by climate change is generally similar to that found by Zeng et al. [2008, Figure 10], considering that Zeng et al. [2008] studied 2100 in a more aggressive climate change scenario, but do not account for the ice-albedo feedback, which impacts Arctic surface ozone. Ozone recovery causes an increases of surface ozone over much of the southern extratropics (Figure 9d), very similar to that found by Zeng et al. [2010]. However, about half of the Southern Hemisphere increase in surface ozone found by Zeng et al. [2010] is due to climate change. In the NH, summertime ozone increases over many continental regions; a local decrease over East Asia would be consistent with decreased ozone production as described by Isaksen et al. [2005], but may also be the result of circulation changes. In other seasons, regions with decreased ozone under ozone recovery only are not necessarily characterized by large local emissions of NOx and NMVOCs (see auxiliary material).1

[25] In our CC simulation, for JJA UKCA only produces a lesser (mostly insignificant) reduction in surface ozone than over the tropical and NH oceans. Zeng et al. [2010] attributed their increase in surface ozone to increased STE; this factor appears to be less important in our simulations. This may be a model difference, or it may reflect the different time periods considered (2050 versus 2100).

[26] The impact of ozone recovery on surface ozone is a little smaller under high methane than under low-methane conditions (Figure 9h). This may be because under high-methane conditions chemical production and loss of ozone both increase but STE of ozone (the dominant way in which ozone recovery affects surface ozone) does not (section 4.6). The combined simulation (Figure 9f) suggests substantially increased ozone in regions that already have high surface ozone, particularly Western Europe, the Mediterranean, and the Middle East. A comparison with Figure 9a indicates that methane increases are the dominant cause of this. (Note again that changes in NMVOC, CO, and NOx emissions presently affecting, e.g., East Asia are not considered here.) Zeng et al. [2010] did not find substantial changes in the NH due to ozone recovery; here, ozone recovery causes some regional increases of ozone over the northern continents during summer, but also regional decreases. Neither is as systematic and large-scale as the increase in the Southern Hemisphere. In regions that exhibit much reduced ozone in simulation CC (the tropics, due to intensified HOx catalyzed ozone destruction, and the Arctic Ocean, due to the reduction in sea ice cover), surface ozone changes are often insignificant in ORCC. This suggests substantial cancellations of opposing effects (increases of ozone mostly due to increased methane versus decreases due to climate change), particularly over some of the world's oceans, including the Arctic Ocean and some of the tropical oceans. The decreases under climate change are larger over the ocean than over land, and are inhomogeneously distributed over the oceans, because they are mostly the result of increasing humidity.

[27] In the other seasons, in CC, decreases in surface ozone occur in many regions. Significant increases are only found in the Arctic from September to February. In ORHM, surface ozone will increase almost everywhere, including in southern middle latitudes. ORCC likewise has increasing surface ozone almost everywhere, except over parts of the tropical oceans where seasonally significant decreases are modeled (cf. the auxiliary material).

4.6 The Tropospheric Ozone Budget

[28] In diagnosing tropospheric ozone budget terms, we define the tropopause as the height of the 150 ppbv ozone level, taken from a 10 year mean, monthly mean ozone climatology of the REF simulation. Hence, the height of the tropopause is assumed identical in all six simulations for this diagnostic, and is annually periodic. This definition is as adopted by Stevenson et al. [2006]. The ozone-based tropopause pressure at this level in the ORCC and REF simulations, as well as the one based on lapse rate in REF, are displayed in the figure of the auxiliary material. In the tropics, the two definitions give a similar tropopause pressure in REF, but in the extratropics the ozone-based tropopause is consistently at a higher pressure than the lapse-rate tropopause. This means ozone budget terms based on the lapse-rate tropopause would be larger than those presented here. In the ORCC simulation, in the extratropics the ozone tropopause on average is at a slightly larger pressure than in REF, whereas in the tropics the opposite ensues. The other simulations yield similarly small differences in tropopause height with REF.

[29] Table 2 summarizes the ozone budgets and ozone and methane lifetimes for the six simulations and for two important literature references based on multimodel analyses. For REF, production, loss, and dry deposition are smaller than those published by Stevenson et al. [2006] but compare well to Prather et al. [2001]. Lack of isoprene chemistry in our model is a major reason for the relatively small production and loss terms. STE is smaller than the multimodel average from Stevenson et al. [2006] and Prather et al. [2001], and also smaller than an independent estimate, based on observed tracer-tracer correlations, of 450 Tg/year [Murphy and Fahey, 1994]. Unlike Stevenson et al. [2006] and Prather et al. [2001], here we study results from a whole-atmosphere model in which STE is not an easily tunable process. The ozone lifetime is slightly larger than in the two references, in agreement with reduced production and loss. The methane lifetime is slightly longer but within the uncertainty given by Stevenson et al. [2006]. Including more detailed NMVOC chemistry in the model may bring down the methane lifetime; this is the subject of further investigation. However, also this partly indicates less OH than in the references, which is consistent with the smaller tropospheric ozone burden. Krol et al. [1998] inferred the methane lifetime based on the observed rate of change of methyl chloroform, which constrains OH; their estimate is that the methane lifetime is 8.6 years in 1993. Uncertainties in understanding the present-day methane lifetime, as discerned from the current literature, remain considerable (e.g., ±1.32 years in the case of Stevenson et al. [2006]).

Table 2. Tropospheric Ozone Budget Terms, in Tg(O3)/year
  1. P = chemical production by RO2 + NO, with R = H, CH3, C2H5, etc. L = chemical loss via H2O + O(1D) → 2 OH, HO2 + O3→ OH + 2 O2, and OH + O3→ HO2 + O2. DD = dry deposition. STE = inferred stratospheric flux (L + DD - P). B = annual-mean burden (Tg(O3)). inline image B / (L + DD) (days). inline image is the global methane lifetime (years). The methane lifetime is calculated taking into account all chemical (including stratospheric) sinks and assuming a soil sink of 30 Tg/year (as assumed by Prather et al. [2001] and Stevenson et al. [2006]). Percentage differences (given in brackets) are relative to REF. S06 = multimodel mean and standard deviation in Table 5 [Stevenson et al., 2006]. TAR = Prather et al. [2001].
P (Tg(O3)/a)5110±606342038454128 (7%)4140 (8%)3863 (0%)3963 (3%)4276 (11%)
L4668±727347034513701 (7%)3756 (9%)3519 (2%)3591 (4%)3931 (14%)
DD1003±200770724774 (7%)789 (9%)740 (2%)713 (–2%)782 (8%)
STE552±168770329347 (5%)405 (23%)399 (21%)340 (3%)435 (32%)
B (Tg(O3))344±39300296318 (7%)327 (10%)307 (4%)283 (–4%)316 (7%)
inline image (days)22.3±22425.525.6 (0%)25.9 (2%)25.9 (2%)23.7 (–7%)24.1 (–5%)
inline image (years)8.67±1.328.49.710.9 (12%)10.9 (12%)9.7 (0%)9.3 (–4%)10.3 (6%)

[30] A comparison between HM and REF suggests that increasing methane increases production and loss of ozone about proportionally, leaving the ozone lifetime unchanged but leading to an increased ozone abundance. The global-total sensitivity of tropospheric ozone to changes in methane in our model is 34 Tg(O3)/ppmv(CH4) or 11%/ppmv. Ozone recovery following the A1/A1b scenarios but leaving methane unchanged (ORLM) mainly leads to a 21% increase in STE [Zeng et al., 2010], which causes a 4% increase in the ozone burden. The differences with respect to REF evident in HM and ORLM add up about linearly to the changes discerned from ORHM, except for STE in ORHM, which is less than the sum of those in HM and ORLM. Climate change (simulation CC) results in a reduced ozone burden (by 4%) associated with a reduced lifetime of ozone (by 7%). This is due to increased moisture increasing the efficiencies of the ozone destruction channels quantified in Table 2. The differences in the combined simulation (ORCC) are roughly linear in the chemical changes as found in ORHM and the impacts of climate change as found in CC. Again, the exception is STE, which in ORCC increases by more than the sum of those in ORHM and CC. The results imply that in our model, 2050 increases in ozone STE will be dominated by chemical ozone recovery, with methane increases and any speed-up in the Brewer-Dobson Circulation playing minor roles.

[31] An analysis of the methane lifetimes indicates that there is a substantial positive feedback between increases in methane and increases in the methane lifetime. This nonlinearity in the CH4-CO-OH system is well-known [e.g., Prather, 1994]. Here we investigate the effect of a 0.64 ppmv or 36% increase in methane from 2000 to 2050, which is larger than in any but the most extreme of the RCP scenarios [Meinshausen et al., 2011]. The methane lifetime is linked to the methane abundance with a sensitivity of about 1.7 years/ppmv(CH4) or 16%/ppmv under both present-day (HM, REF) and ozone-recovery (ORHM, ORLM) conditions. Following the formalism of Prather et al. [2001], using the lifetimes of the HM and REF simulations, we obtain a lifetime enhancement factor inline image. Comparing this to the results listed by Prather et al. [2001, Table 4.3], this is almost the same as the resulting enhancement factor for the model that has a methane lifetime closest to the UMUKCA one, namely the Royal Netherlands Meteorological Institute (KNMI) model, which produces a lifetime of 9.8 years and an enhancement factor of 0.31. All other models used by Prather et al. [2001] have slightly smaller values for s.

[32] The result means that increasing methane emissions would lead to more than proportional increases in the methane abundance because of the impact of the additional methane on OH, the dominant sink for methane. In our simulations, this feedback is not affecting the methane abundance because methane is prescribed at the surface (section 2.1). The methane lifetimes respond linearly to methane increases and ozone recovery (as discerned from HM, ORLM, and ORHM) and to chemical and climate changes (as discerned from ORHM, CC, and ORCC). In our simulations, ozone recovery does not appreciably affect the methane lifetime, because the associated change in the global and annual-mean tropospheric ozone burden is too small to much affect the oxidizing capacity. On the other hand, climate change, through increased moisture and temperature, leads to an increase in HOx, which is linked to a decrease in tropospheric ozone and also in the methane lifetime. However, in our simulations methane increases dominate the future changes in tropospheric ozone and the oxidizing capacity, in qualitative agreement with John et al. [2012] and Voulgarakis et al. [2012].

5 Discussion and Conclusions

[33] We have presented a model analysis of six simulations designed to explore the sensitivity of tropospheric ozone and the oxidizing capacity to projected methane increases, stratospheric ozone changes, and climate change between the years 2000 and 2050. Of these three factors, the most important factor affecting the tropospheric oxidizing capacity, given the scenarios employed, is methane. This is due to OH + CH4 being an important sink for OH in the troposphere, causing the methane lifetime to depend on the methane abundance, with more methane causing a longer lifetime. The methane lifetime (which reflects tropospheric OH) increases by 12% under the imposed methane increase, half of which would be offset by projected climate change. Ozone recovery has no appreciable net influence on the methane lifetime. Regarding tropospheric ozone, in the annual and tropospheric mean, climate change and stratospheric recovery are about equal, but opposite in sign, in their influence on the tropospheric ozone burden; the result is a net 7% increase in the tropospheric ozone burden driven by increasing methane (Table 2). For surface ozone, climate change is a bigger influence than ozone recovery (Figure 9). In the Southern Hemisphere, as was found before [Zeng et al., 2010], ozone recovery causes an increase of surface ozone during autumn to spring, caused by increased STE. Ozone recovery affects surface ozone more through impacts on cross-tropopause transport of ozone and via its effects on tropospheric photolysis (as discussed by Fuglestvedt et al. [1994]). Sea ice losses projected under climate change will have a substantial regional impact on surface ozone and OH due to the associated reduction in surface albedo. This means reduced surface ozone over the Arctic during summer [Voulgarakis et al., 2009]. Changing from permanent to seasonal sea ice cover in the Arctic might substantially enhance bromine explosion events, which occur over young ice and would seasonally affect tropospheric ozone [e.g., Bottenheim et al., 1990]; this effect is not included in our model.

[34] In interpreting the results, one needs to consider that the model likely underestimates STE [Stevenson et al., 2006]. This means that in reality, the impacts of stratospheric ozone changes on the troposphere may be more pronounced than in this version of our model, particularly over the Southern Hemisphere. Zeng et al. [2008, 2010] found substantially larger influences of both climate change and ozone recovery on surface ozone. This may be partially explained by them studying the year 2100, meaning stronger climate change and larger ozone changes than in our simulations. However, partially this may also reflect their model tendency to produce larger STE. They did not consider methane changes in their simulations, explaining the lack of substantial changes in the Northern Hemisphere. In our model, reductions in surface ozone induced by climate change (CC) over the northern continents, the Southern Ocean, and Antarctica are partially cancelled by increases induced by ozone recovery in these regions (ORLM). The dominant influence of ozone recovery is via its impact on STE of ozone. In our ORCC simulation, methane increases are responsible for a significant increase in Northern-Hemisphere surface ozone (Figures 9 and 10).

Figure 10.

Differences in seasonal and multiannual mean surface ozone (at 10 m) in northern summer between pairs of simulations as indicated, in ppbv. Dots (stipling) indicate insignificant differences.

[35] Our study elucidates the relative roles of three major factors driving composition change in the coming decades, namely climate change, ozone recovery, and increasing methane. The study suggests that methane growth could outweigh the other two factors in driving up tropospheric ozone. Of the four RCP scenarios described by Meinshausen et al. [2011], only the most extreme (RCP8.5) has more methane in 2050 than the A1b scenario used here. Assuming linearity, the ozone budget numbers and associated quantities listed in Table 2 could be constructed for the other RCPs that have smaller changes in methane; under these more moderate projections of the future, climate change and ozone recovery would be relatively more important drivers affecting tropospheric ozone and the oxidizing capacity.

[36] The results suggest that limiting methane growth would be of considerable benefit beyond avoiding the direct climate change associated with this growth. The nonlinear feedback associated with methane removal means that a reduction in methane emissions would yield, over time, more than proportional reductions in the methane abundance. Considerable production of tropospheric ozone would be avoided, meaning better air quality and less warming [e.g., Fiore et al., 2002; Dentener et al., 2005; West et al., 2007; Fiore et al., 2008; West et al., 2012].


[37] We acknowledge the New Zealand Ministry for Business, Innovation, and Employment for funding this work. We acknowledge NCAS Computational Modelling Services for supporting the MetUM. We thank Michael Prather from the University of Irvine for providing the FAST-JX photolysis code. The development of the UKCA model ( was supported by the Joint DECC/Defra Hadley Centre Climate Programme (GA01101) and the Natural Environment Research Council (NERC) through the NERC Centres for Atmospheric Science (NCAS) initiative. We acknowledge usage of surface ozone data from the World Data Center for Greenhouse Gases (