Journal of Geophysical Research: Atmospheres

Sensitivity of GCM tropical middle atmosphere variability and climate to ozone and parameterized gravity wave changes

Authors


Abstract

[1] This paper describes the impact of changing the current imposed ozone climatology upon the tropical Quasi-Biennial Oscillation (QBO) in a high top climate configuration of the Met Office U.K. general circulation model. The aim is to help distinguish between QBO changes in chemistry climate models that result from temperature-ozone feedbacks and those that might be forced by differences in climatology between previously fixed and newly interactive ozone distributions. Different representations of zonal mean ozone climatology under present-day conditions are taken to represent the level of change expected between acceptable model realizations of the global ozone distribution and thus indicate whether more detailed investigation of such climatology issues might be required when assessing ozone feedbacks. Tropical stratospheric ozone concentrations are enhanced relative to the control climatology between 20–30 km, reduced from 30–40 km and enhanced above, impacting the model profile of clear-sky radiative heating, in particular warming the tropical stratosphere between 15–35 km. The outcome is consistent with a localized equilibrium response in the tropical stratosphere that generates increased upwelling between 100 and 4 hPa, sufficient to account for a 12 month increase of modeled mean QBO period. This response has implications for analysis of the tropical circulation in models with interactive ozone chemistry because it highlights the possibility that plausible changes in the ozone climatology could have a sizable impact upon the tropical upwelling and QBO period that ought to be distinguished from other dynamical responses such as ozone-temperature feedbacks.

1. Introduction

[2] For a combination of reasons, future evolution of the global ozone distribution is currently a subject of keen interest. There is an ongoing requirement to assess the impact upon Antarctic ozone depletion of restrictions in release of anthropogenic halogen compounds imposed by the Montreal Protocol. Ozone absorption of shortwave solar radiation throughout the stratosphere (and emitted longwave radiation in the lower stratosphere) provides diabatic heating which stabilizes the region and this dynamical interaction makes ozone prediction key for accurate representation of middle atmosphere climate change in general circulation models (GCMs). The common GCM approach, whereby ozone is specified through a monthly zonal mean climatological distribution to enable diabatic heating from radiative processes to affect the temperature structure, is satisfactory under present-day conditions for which observation-based data sets are available but introduces increasing uncertainty for future projections. A desire for a more integrated approach, which can encompass both the radiative impact of ozone upon temperature and the temperature-sensitive chemical and dynamical processes that determine ozone concentrations, has driven the recent proliferation [Eyring et al., 2006] of coupled chemistry climate models (CCMs). Recent results from Son et al. [2008] show differences in behavior between climate GCMs from the Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC/AR4) and coupled CCMs that result from improved representations of stratospheric ozone recovery and highlight the relevance of ozone representation to future tropospheric climate change [Eyring et al., 2007].

[3] This paper will focus upon one aspect of dynamical variability, the tropical Quasi-Biennial Oscillation (QBO), which recent studies have shown might be impacted by feedbacks arising from the representation of chemical species (in particular ozone) in CCMs [Cordero and Nathan, 2000; Butchart et al., 2003; Shibata and Deushi, 2005a; Tian et al., 2006]. The QBO is a dynamical phenomenon of the tropical middle atmosphere which has generated interest as a prominent source of internal variability ever since it was first identified in zonal wind observations from equatorial regions of the lower stratosphere [Ebdon and Veryard, 1961; Reed et al., 1961]. A comprehensive review of the QBO [Baldwin et al., 2001] describes a long history behind the current attribution to the interaction with the mean flow of upward propagating atmospheric waves at a broad range of spatiotemporal scales as the most probable cause of the oscillation. GCM simulations with sufficient resolution, vertical in particular, in the stratosphere have reported spontaneous behavior driven by resolved wave forcing that resembles the QBO [Takahashi, 1996; Horinouchi et al., 2003; Watanabe et al., 2008; Kawatani et al., 2009]. Other studies have demonstrated the importance of unresolved (parameterized) nonorographic gravity waves in maintaining a QBO which closely resembles that observed [Scaife et al., 2000; Giorgetta et al., 2002; Shibata and Deushi, 2005b]. How well these models represent those fundamental processes which drive QBO variability in the earth's atmosphere still remains subject to debate but they are sufficiently inexpensive to act as components in CCMs.

[4] Thus, after Cordero and Nathan [2000] reported indications that ozone feedbacks in a two-dimensional equatorial stratosphere model can enhance QBO amplitudes, Butchart et al. [2003] described a QBO in ozone in a relatively simple global CCM, where modulation of tropical upwelling over the QBO cycle affected ozone transport and hence modulated ozone concentrations. They reported increased duration of QBO phase and increased temperature amplitudes below 22 hPa, but no impact on zonal wind amplitudes, and concluded that accurate modeling of both chemical and dynamical processes is required in order to capture well the tropical interannual variability. Shibata and Deushi [2005a] also reported prolonged QBO periods after allowing ozone interaction with radiative processes in their model but saw a smaller temperature amplitude response than Butchart et al. [2003]. A study of Tian et al. [2006] confirmed that radiative coupling of more complex interactive chemistry to the GCM dynamics tends to prolong the QBO period and reported increased QBO amplitudes. Tian et al. [2006] noted additional radiation-temperature feedbacks on ozone concentrations that become significant above midstratosphere, including temperature-dependent ozone photodissociation rates and reaction rates for ozone depleting species. The increased complexity of CCMs, with these added feedbacks, makes interpretation of their behavior both more interesting and more difficult to unravel.

[5] Once it is possible to run simulations of a CCM that compare with its host atmospheric GCM configuration (for example, Morgenstern et al. [2009] describe a CCM developed within the MetUM (Met Office U.K. Unified Model) atmospheric GCM) the question arises as to what role might be retained by the GCM with no chemistry and fixed ozone climatology. One important consideration is the added cost of transporting and interacting numerous chemical species in a CCM. Thus where aspects of dynamical behavior are well reproduced in the atmospheric GCM, sensitivities may be explored more rapidly and understanding transferred to CCM scenarios. With regard to the QBO where resolution is an issue, it is possible to explore behaviors at much higher resolutions with the GCM than can be afforded with the CCM and hence to gauge the extent to which choice of resolution might influence results. Another advantage of comparison is that the reduced complexity of the GCM can help to distinguish between changes in the CCM that result from feedbacks and those forced by differences in climatology between the fixed and interactive ozone distributions. Naturally, the focus of CCM experiments referenced above was the attribution of differences to chemical coupling feedbacks. However, a general problem, which is encountered in climate modeling when introducing time varying formulations for previously prescribed quantities, is that the additional freedom allows the model climatology to drift away from that previously constrained and this change in equilibrium state can be sufficient to account for dynamical responses that we would like to be able to pick out from more complex feedback mechanisms.

[6] In order to probe this aspect of ozone dynamical interaction with the QBO further, this paper investigates sensitivities in the MetUM GCM to different representations of the zonal mean ozone climatology under present-day conditions. For pragmatic reasons, the control adopted here is that used in a standard MetUM climate configuration (details in sections 23), constructed from a data set, referred to here as SPARC2000, which was obtained from the Stratospheric Processes And their Role in Climate (SPARC) Data Center. The ozone climatology for the sensitivity test was prepared, from a data set of Rosenlof and coworkers that retains interannual variability, by Dall'Amico et al. [2010] who reported improved temperature trends in a coupled atmosphere-ocean GCM with 40 km top when using it. Differences between the two estimates of present-day conditions arise from differences in instrument measurements used, time frames over which observations are available and choices made in processing the data into a single data set. As such, we submit that they represent the level of change which might be expected between different acceptable model realizations of the global ozone distribution and therefore provide a useful indication of the extent to which more detailed investigation of climate shift issues might be required when assessing CCM ozone feedbacks.

[7] This paper is laid out as follows. Section 2 provides a description of the model used in this study. Section 3 compares the SPARC2000 and Rosenlof climatological ozone data sets with reference to a third climatology of Li and Shine [1995] that was previously used as the standard ozone climatology for the MetUM. Section 4.1 presents results from a sensitivity test in which the Rosenlof climatology replaces that of SPARC2000 in order to assess the influence of prescribed ozone on the MetUM climate, and section 4.2 investigates the influence of prescribed ozone on the model's dynamical QBO. Section 4.3 presents results from a second sensitivity test where in addition the strength of parameterized nonorographic gravity waves in the MetUM is enhanced. We conclude by summarizing results in section 5.

2. Model Description

[8] The middle atmosphere climate configuration of the MetUM GCM used for the study described here is based upon the Hadley Centre Global Environmental Model (HadGEM) package [Collins et al., 2008; Martin et al., 2006; Ringer et al., 2006] and runs as an atmosphere-only model with sea surface temperature and sea ice specified as for the second Atmosphere Model Intercomparison Project (AMIP-II) [Gates et al., 1999]. We use 60 vertical levels from the surface to 84.1 km (∼0.01 hPa). The standard resolution for the model is 1.25° latitude × 1.875° longitude but, in order to reduce costs, sensitivity tests are routinely carried out at the lower resolution of 2.5° × 3.75° as here. Comparison of MetUM behavior at the two resolutions [Hardiman et al., 2010; Osprey et al., 2010] reveals some differences, for instance a tendency to shorter QBO period at the coarser resolution, but similar sensitivities to model parameters suggest a robustness to moderate resolution changes.

[9] Two parameterizations, designed specifically to address middle atmosphere processes, were employed. The methane oxidation scheme provides a source of water vapor in the stratosphere [Untch and Simmons, 1999; Simmons et al., 1999]. Nonorographic gravity wave activity at scales which are smaller than the model resolution are parameterized using the approach of Warner and McIntyre [1996, 1999, 2001] with further modifications to launch an unsaturated spectrum from a level close to the surface [Scaife et al., 2000] and to impose at the launch level a homogeneous total vertical flux of horizontal wave pseudomomentum [Scaife et al., 2002]. Recent refinement to the parametrization allows identification in advance of which part of the launch spectrum will (or will not) be eroded by the local saturation spectrum depending on the density and mean wind Doppler shift relative to the launch level [Bushell et al., 2007]. A condition of zero vertical wave flux passing up out of the top layer (“opaque lid”) is imposed to prevent parameterized gravity waves carrying momentum out through the model top and generating spurious circulations (as per Shepherd and Shaw [2004] and Shaw and Shepherd [2006]) where net momentum is no longer conserved.

3. Climatological Ozone Data Sets

[10] Three 12 year experiments were carried out from a fixed start date of 1 September 1978. The control simulation used the current standard HadGEM ozone (SPARC2000) climatology, which comprises the annual cycle derived from multiyear regression of a stratospheric ozone data set [Randel and Wu, 1999] that combines Stratospheric Aerosol and Gas Experiment (SAGE) I/II and Total Ozone Mapping Spectrometer (TOMS) satellite data with ozonesonde polar observations above the tropopause for the period 1979–1997. As described by Karoly [2000], the SPARC2000 climatology merges this stratospheric component with a hybrid tropospheric data set generated from a combination of observations and chemistry model output [Kiehl et al., 1999]. We refer to this data set as SPARC2000 in order to distinguish it from a data set made available from the SPARC Data Center more recently than the definition [Martin et al., 2006] of our climate control configuration. The two sensitivity tests used an alternative climatology constructed from an ozone data set of Rosenlof and coworkers [Dall'Amico et al., 2010, appendices A–C] for the period 1979–2003. Stratospheric ozone in this data set is derived mainly from satellite observations: Halogen Occultation Experiment (HALOE), Microwave Limb Sounder (MLS), Solar Backscatter Ultraviolet (SBUV) instrument, SAGE-II, and TOMS. Below 215 hPa, a scaled version of the Fortuin and Kelder [1998] tropospheric ozone climatology is used, adjusted such that the combined total column ozone 1990s mean climatology matches that of TOMS.

[11] Figure 1 shows mean equatorial zone (5°S–5°N) height profiles of annual mean climatological ozone concentrations from the SPARC2000 and Rosenlof data sets. A third present-day climatology of Li and Shine [1995] is included for reference. Relative to the other ozone data sets, SPARC2000 shows 20%–25% larger peak concentrations which occur 2 km higher. In addition, the SPARC2000 ozone peak is less spread in the vertical, resulting in significantly lower concentrations in the lower stratosphere. The close comparison between profiles for the Rosenlof and Li and Shine data sets suggests that the HadGEM standard ozone climatology derived from the SPARC2000 data set may be somewhat of an outlier. Note that as the Li and Shine data set is the only one of the three with information reaching as high as 0.01 hPa, the climatologies used by the 60 level model in these experiments have been constructed with data from Li and Shine blended in where higher levels would otherwise be missing. The impact upon total column ozone from this merging is expected to be small as the column above 60 km contributes less than 1% to the total.

Figure 1.

Annual mean equatorial (5°S–5°N) mean profiles of ozone volume mixing ratios for climatologies of SPARC2000 (control, solid line), Rosenlof (dashed line), and Li and Shine (dotted line).

[12] Annual mean zonal mean latitude-height cross sections (Figure 2) again show largest peak concentrations for the SPARC2000 climatology and rather more similar behavior for those of Rosenlof and Li and Shine. The picture from Figure 1 broadly holds across the tropics in difference plot Figure 2d, with concentrations in the Rosenlof climatology exceeding those in SPARC2000 over the lower stratosphere by up to 2.0 ppmv, while the SPARC2000 data set has larger concentrations by up to 2.7 ppmv from 30 to 40 km. As density weighting favors peak concentrations lower in the atmosphere, the net impact that this has on the annual mean equatorial minimum estimated total column ozone for SPARC2000 (210 Dobson Units (DU)) is an increase for Rosenlof of 43 DU which hardly varies over the annual cycle. Comparison of total column ozone climatologies by Eyring et al. [2006, Figure 14] gives ranges which are closer to the Rosenlof values. In extratropical midlatitude regions around 50°N and 50°S, where the SPARC2000 total column ozone peaks, the two data sets have very similar values.

Figure 2.

Annual mean zonal mean ozone mixing ratio climatologies of (a) SPARC2000 (control), (b) Rosenlof, (c) Li and Shine (contour intervals 1 ppmv), and (d) difference of Rosenlof from SPARC2000 (contour intervals 0.25 ppmv; dashed contours represent negative values).

4. Results

4.1. Response of Model Climatology to Changes in Prescribed Ozone

[13] The sole change from the control configuration in the first sensitivity experiment (S:Ozone) was to replace the SPARC2000 ozone climatology with that of Rosenlof. A similar test is described by Dall'Amico et al. [2010], but the experiments differ because the 60 level model here, as described in section 2, has the added parameterizations and top at 84 km not 40 km, which allow it to represent dynamical circulations up to the mesosphere and a dynamical QBO responsive to the ozone changes. Zhong et al. [2008] have shown evidence that shortwave radiative heating rates in the MetUM are very sensitive to changes in ozone absorption coefficients and, by inference, concentrations. Radiative heating changes represent the direct impact of an altered ozone climatology. Dynamical changes in the tropical climate, however, must arise indirectly as a consequence of diabatic heating changes altering the local temperature and circulation and thus influencing the mean flow, wave activity or interactions between them that are hard to quantify.

[14] In contrast to the approach of Butchart et al. [2003], where responsive ozone interacts with the CCM dynamics but differences from the GCM climatology are specifically neglected, this paper focuses upon changes due to the imposed climatology of ozone, which varies monthly through a repeat annual cycle but does not respond to changes in model temperature or vertical velocity. In the absence of a transient response to ozone changes that are linked to the QBO phase and hence can interact constructively [Butchart et al., 2003] or destructively [Shibata and Deushi, 2005a] to influence the QBO evolution, any impact capable of sustained influence upon the QBO dynamics has to persist over the timescale of multiple QBO periods, which leads us to examine the equilibrium response in radiative heating.

[15] The mean annual equatorial 5°S–5°N profiles of instantaneous net clear-sky radiative heating (Figure 3a) for the control and S:Ozone sensitivity experiments are generally similar, reducing almost to zero in the upper stratosphere but otherwise tending to warm the middle atmosphere as the effects of longwave cooling and shortwave heating combine. Monthly differences from the annual mean are relatively small. Ultraviolet absorption by ozone generates a peak in shortwave radiative heating (Figure 3c) at around 45 km, which is above the peak in ozone mass mixing ratio (Figure 1). The picture for longwave radiative heating (Figure 3e) is more complicated for two reasons. First, longwave radiation emitted by the warmer underlying atmosphere significantly heats the stratosphere. Second, the dominant contribution to the cooling from energy reemitted at wavelengths characteristic of the local layer temperature is the 15 μm carbon dioxide band: ozone adds less than 20% at the peak near 50 km and even contributes a slight warming near the tropopause.

Figure 3.

Annual mean (1979–1990) equatorial 5°S–5°N profiles for control versus S:Ozone clear-sky radiative heating rates (K day−1) (a and b) net, (c and d) shortwave, and (e and f) longwave. On difference plots (Figures 3b, 3d, and 3f), crosses and thick lines denote differences significant at the 99% level.

[16] The ozone profile differences (Figure 2d) between S:Ozone and the control experiment led to a net clear-sky radiative warming in the equatorial (5°S–5°N) stratosphere above 50 km and below 30 km that is statistically significant at the 99% level (using Student's T test), with net cooling up to 0.36 K day−1 in between (Figure 3b). This pattern emerges from a combination of differences in the clear-sky shortwave radiative heating (Figure 3d), which shows significant warming where S:Ozone has significantly more ozone than in the control, and longwave heating (Figure 3f), which shows significant warming where S:Ozone has significantly less. The differences largely cancel to leave a residual net heating that is dominated by the shortwave heating response below 60 km and the longwave heating response above 65 km.

[17] The mean annual 5°S–5°N equatorial zone temperature profile (Figure 4a) shows systematic differences between the control and S:Ozone experiments above the tropopause (Figure 4b) that are consistent with differences in the clear-sky net radiative heating rates (Figure 3b). In the troposphere, the slight cooling of S:Ozone relative to the control does not appear to be significant. In the lower stratosphere between 100 and 10 hPa, the S:Ozone temperature profile is warmer by up to 6K, a response which is significant at the 99% level and consistent with the change in net radiative forcing, as is the cooling in the upper stratosphere (10 to 2 hPa) and the warming around the stratopause at about 1 hPa. An approximate height scale (proportional to log [pressure]) is indicated to aid comparison with Figure 3. This pattern of changes is relatively uniform throughout the entire (30°S–30°N) tropical region (not shown) and not just restricted to the equator.

Figure 4.

Annual mean (1979–1990) equatorial 5°S–5°N profiles for (a) temperature (K) and (b) S:Ozone experiment temperature differences from control (K), marked with crosses and thick lines when significant at the 99% level.

[18] Under the assumption of thermal wind balance as per the equatorial beta-plane approximation [Andrews et al., 1987], tropical temperature changes should be proportional to the vertical shear of zonal wind such that increased/decreased temperatures are associated with eastward/westward vertical shear. This is confirmed by annual mean zonal wind profiles in the 5°S–5°N equatorial region (Figure 5a), which show differences in predominantly westward flow between S:Ozone (dashed line) and the control (solid line) that are significant (Figure 5b, dashed line) in the lower stratosphere (40–70 hPa), where the maximum eastward bias reaches 3.1 m s−1, and above the stratopause (0.2–0.03 hPa), where added westward flow with maximum of 4.8m s−1 can actually reverse flow direction, which is eastward above 0.3 hPa in the control. The height at which S:Ozone is coolest with respect to the control corresponds well with the location of zero wind difference between the two. Over the entire 30°S–30°N tropical pipe region (Figures 5c and 5d) this picture is even more pronounced, with a maximum 4.7 m s−1 decrease in westward flow below 5 hPa and maximum westward enhancement of 6.5 m s−1 above.

Figure 5.

(a) Annual mean (1979–1990) equatorial 5°S–5°N zonal wind (m s−1) profiles for control, S:Ozone, and S:GW+Ozone; (b) differences from annual equatorial control; (c) same as Figure 5a but for tropical pipe 30°S–30°N; and (d) differences from annual tropical control. Differences significant at the 99% level are marked with crosses and thick lines.

[19] A version of the thermodynamic equation at the equator [Randel et al., 1999; Andrews et al., 1987] predicts that equation imagediab ≃ [(∂equation image/∂t) + (equation image*)(∂equation image/∂y)] + equation image*(H/R)N2. We choose to ignore the terms in square brackets (mean temperature drift and meridional transport of temperature) which are small in an equilibrated climate model in the equatorial lower stratosphere. Thus the change in diabatic heating, equation imagediab, due to ozone should be approximately proportional to the change in upwelling residual vertical velocity (equation image*) with a ratio dependent on buoyancy frequency N (s−1), R = 287.05 J K−1 kg−1 and scale height H ≃ 7000 m. A profile of the change in upwelling residual vertical velocity over the 5°S–5°N equatorial band (Figure 6a), shows an increase, of up to 0.14 mm s−1, for the S:Ozone experiment (dashed line) throughout the stratosphere between 100 hPa and 3 hPa. A similar adjustment of the basic model climatology in response to the ozone distribution change was reported by Dall'Amico et al. [2010]. Figure 6b shows the difference between S:Ozone and control in the upwelling term from the thermodynamic equation for comparison with net clear-sky radiative heating rate differences in Figure 3b. The key conclusion from this comparison is that significantly increased upwelling between approximately 100–4 hPa (Figure 6a) is consistent with dynamical balancing of a warming contribution with maximum 0.12 K day−1 between 15–35 km (Figure 6b) that is a sizable proportion of the collocated ozone induced radiative heating change (Figure 3b). This strongly supports the concept of a relatively localized equilibrium response of the dynamics to climatological ozone changes in this region.

Figure 6.

Annual mean (1979–1990) equatorial 5°S–5°N differences from control profiles of S:Ozone and S:GW+Ozone for (a) residual vertical velocity and (b) heating rate term calculated from residual vertical velocity and static stability.

4.2. Response of Model Dynamical QBO to Changes in Prescribed Ozone

[20] The next question to address is how ozone-induced changes in model climatology might affect tropical variability and in particular the QBO. Figure 7a shows the control time series of equatorial (5°S–5°N) monthly mean zonal wind profiles with the classic signal of alternately descending bands of westward (negative) and eastward (positive) winds over much of the 100 hPa to 1 hPa pressure range. The control configuration at 10 hPa has a QBO period range of 24–29 months (mean 26.5 months) which lies within the 22–35 months range observed over a 12 year period (1994–2005) of the Met Office operational stratospheric analysis record. This agreement was expected as the 60 level climate configuration is relatively mature and has evolved with settings for parameterizations, such as the nonorographic gravity wave scheme, which have been chosen, within a plausible range, on the basis of the model's agreement with observation. Based on current understanding that the QBO descent rate is a measure of how effectively wave energy deposited in the wind shear zone around zero velocity is able to counteract the equatorial upwelling, a 0.14 mm s−1 enhanced upwelling in the lower stratosphere tropical pipe (Figure 6a) equates to a 4.4 km yr−1 reduction in descent rate. Just such a reduction is noted in the descending easterly shear zone (westward winds above eastward), which stalled more often and slowed from 8.4 km yr−1 to just 4 km yr−1 in the S:Ozone experiment and increases the QBO period (Figure 7b) although it appears, on closer inspection, that the descending westerly shear zone actually descended at a similar rate to the control case. A greater tendency for stalling in descending easterly conditions has been reported from observations [Dunkerton and Delisi, 1997]. Relative to the control, the longer mean QBO period, 38.5 months with a 38–42 month range, inevitably appears to reduce the agreement with reality.

Figure 7.

Monthly mean equatorial 5°S–5°N zonal wind for (a) the control experiment, (b) the S:Ozone experiment, and (c) the S:GW+Ozone experiment. Contour intervals are at 10 m s−1, and dashed contours represent negative (westward) wind values.

[21] As the value of the QBO period is itself an observable characteristic of the stratosphere, there is a need to enable internal adjustment of the model atmosphere either to counteract the increase in upwelling or to compensate by enhancing the wave energy deposition. In the MetUM, a key source of wave energy deposition that governs the QBO behavior is the nonorographic gravity wave parametrization. Hence, in the absence of tighter constraints from direct measurements of gravity wave activity, an increase in parameterized nonorographic gravity wave activity in order to reduce the QBO period toward that observed would be justified.

4.3. Sensitivity to Model Gravity Wave Parameters

[22] At this point, therefore, we introduce a second sensitivity experiment, S:GW+Ozone, which used S:Ozone as its control and additionally enhanced the launch spectrum amplitude for nonorographic gravity waves in order to bring the QBO period back within the observed range (requiring a 36% amplitude increase). S:GW+Ozone has a dual function. It enables the Rosenlof ozone behavior to be assessed against the control without the distraction of dynamical differences that arise from phase mismatches due to the change in QBO periods and, by comparison with S:Ozone, it allows exploration of the extent to which the original sensitivity to a direct change of ozone data set might be attributed to an indirect impact on the parameterized gravity waves.

[23] In S:GW+Ozone the additional momentum available to the parameterized gravity waves for deposition in the descending easterly shear zone accelerates the QBO descent and reduces its period back to a range of 23–29 months, with mean 26.8 months (Figure 7c). Note (Figure 6a) that the reduction is achieved without significantly affecting upwelling (dash-dotted line) in the 100–3 hPa region and although there appears to be improved agreement with the control zonal wind (Figures 5b and 5d, dash-dotted line) at the stratopause, the differences are not significant at the 99% level.

[24] Time series at selected pressure levels enable a more quantitative description of the changes between experiments. Examination of the equatorial zone monthly mean zonal wind at 100 hPa for the control climatology run (not shown) reveals a multiyear mean seasonal cycle of westward winds with an approximate peak-to-peak variation of 4 m s−1 from minimum around February to maximum in August. When this mean seasonal cycle is subtracted month by month from the 12 year sequence to produce a deseasonalized time series, the model has no readily discernible QBO signal at the 100 hPa level. At 30 hPa, however, the same procedure yields a clear QBO signal (Figure 8a) with an amplitude in the control of approximately 18 m s−1, very similar to that observed in Met Office stratospheric analyses. The period lengthening from a mean 26.5 to 38.5 months once the ozone climatology is substituted (S:Ozone) and the reduction back to 26.8 months in response to the added gravity wave input (S:GW+Ozone) are clearly visible. The asymmetry in amplitudes of the eastward and westward phases in Figure 7 appears to arise from a relatively symmetric oscillation in zonal wind offset by a persistent westward flow around the equator, which has a mean seasonal cycle that peaks around 10 m s−1 in August (Figure 8b). By comparison, the seasonal cycle in the Met Office analyses (black diamonds on plot) also shows a maximum mean westward wind in August (12 m s−1) but its minimum of only 2 m s−1, which occurs in March rather than January, gives a much wider range (10 m s−1) than is seen in any of the MetUM runs. At just 3 m s−1, the S:Ozone sensitivity experiment has the lowest seasonal cycle range, as against values of 4.3 m s−1 for the control, and the August maximum close to 6 m s−1 is reduced by 3.1 m s−1 relative to the mean control result. The range of 4.4 m s−1 for S:GW+Ozone is closer to that for the control and its minimum, although occurring in January rather than March as with all the model runs, is closest to that of the observations.

Figure 8.

Equatorial 5°S–5°N mean zonal wind at 30 hPa (control, solid line; S:Ozone, dashed line; and S:GW+Ozone, dash-dotted line) as (a) deseasonalized monthly means 1979–1991 and (b) multiyear (1979–1991) mean seasonal cycle. (Diamonds denote assimilated observations from Met Office operational stratospheric analyses over the 1994–2005 period).

[25] At 1 hPa, the QBO signal (not shown) is no longer discernible from noise, having given way to the SAO of the stratosphere and mesosphere. The mean wind in Figure 9 shows a clear semiannual variation with westward peaks in January (50 m s−1) and July (45 m s−1). The Met Office analysis data peaks at the same months but with values of 34 and 14 m s−1, respectively, which accords with observation climatology results [Garcia et al., 1997] that the “first westward cycle” of northern hemisphere winter is the stronger. The presence of a much stronger annual component relative to the semiannual than is present in the MetUM simulations is also supported by a comparison with European Centre for Medium-Range Weather Forecasting Reanalysis (ERA40) assimilated data and HadGEM runs at both standard horizontal resolution and the coarser resolution used here [Osprey et al., 2010].

Figure 9.

Equatorial 5°S–5°N mean zonal wind at 1 hPa multiyear (1979–1991) mean seasonal cycle (see Figure 8).

[26] The stratospheric SAO arises as a consequence of variation between westward flow generated by horizontal advection at the solstices as the atmosphere responds to the off-equatorial ozone heating maximum and eastward flow generated by wave activity around the equinoxes of which Hitchman and Leovy [1988] assessed Kelvin waves might account for up to 70%. In a study of equatorial waves excited by convective heating, Garcia and Sassi [1999] identified a contribution to both eastward and westward SAO phase, that increases from the stratosphere into the mesosphere, from variable-scale inertio-gravity waves and suggested that smaller scale westward gravity waves with lower phase speeds should be susceptible to damping by the westward QBO phase that can hence modulate the SAO. In discussing the relative contribution of large and small-scale waves, Shepherd [2000] emphasized the fact that GCMs are generally able to simulate an SAO even when failing to produce a QBO. Figure 9, certainly, shows little difference between either sensitivity or the control simulations although, interestingly, the greatest variation does appear at the westward peak in northern hemisphere winter. Further analysis would be required to ascertain if the variation is modulated by the QBO as per Garcia and Sassi [1999].

[27] The vertical propagation of nonorographic gravity wave fluxes parameterized by the MetUM can be seen in Figure 10. Vertical flux of horizontal pseudomomentum due to unresolved nonstationary gravity waves is initialized low in the troposphere and absorbed due to Doppler distortion of the wave spectrum as the upward propagating westward (eastward) flux encounters the wind transition to westward (eastward) winds (see contours in Figure 10). The strong flux gradients (indicated in Figure 10 by rapidly changing color shading) in these shear zones represent momentum deposition onto the mean flow. It is clear that the gravity wave spectra are heavily modified in the region where the QBO dominates and both eastward and westward propagating waves influence the flow. However, there are still fluxes that pass upward to interact with the SAO and two points with relevance to the above discussion are worth noting. First, the eastward component fluxes penetrate the QBO region much more effectively than the westward and, second, westward fluxes above 40 km show a modulation with QBO period that results from throttling that is most effective when the westward QBO phase occupies the greatest depth.

Figure 10.

Monthly equatorial 5°S–5°N mean zonal wind (10 m/s contours; dashed contours represent negative values) overlaid on parameterized gravity wave (a–c) westward or (d–f) eastward horizontal pseudomomentum flux (shading) for control, S:Ozone, and S:GW+Ozone experiments.

[28] The shift in annual mean cycle between the control and S:Ozone equatorial zonal mean zonal wind at 30 hPa (Figure 8) is consistent with the earlier conclusion of a shift in the mean flow. Certainly, gravity wave fluxes (Figures 10b and 10e) in the S:Ozone experiment appear to propagate upward in the same way as those in the control (Figures 10a and 10d), despite the longer QBO period. In contrast, when the initial flux is increased by 36% in S:GW+Ozone, fluxes below the QBO shear zones are correspondingly higher (Figures 10c and 10f) but those propagated through the QBO shear zones appear to be similar to the earlier cases which indicates increased deposition of momentum in the QBO region, consistent with the more rapidly descending, reduced period QBO. The fact that this difference is not seen between the control and S:Ozone runs seems to confirm that the original increase in QBO period did not result from in situ changes in parameterized gravity wave propagation.

[29] In order to examine the relative influence of drag sources on the QBO for experiments with varying QBO period, we begin by taking the equatorial zonal mean zonal wind at 10 hPa (middle stratosphere), remove the seasonal cycle (as in Figure 8) and average the time series over the QBO period, defined as the interval between two consecutive eastward-to-westward (descending easterly) zero wind transitions. The mean cycle can then be plotted (Figure 11) normalized to the mean period (running from 0 to 1 with optional repetition to help in visualizing sequences). Assimilated observations from monthly means of daily Met Office analyses in the 12 year period from 1994 to 2005 (Figure 11) are closely balanced between eastward phase (50.7% and a 17 m s−1 maximum) and westward phase (49.3% and a 14 m s−1 maximum). A 1 month discretization error in the data represents 4% of the mean period in this case, which suggests that the deseasonalized QBO derived from observed data can be considered to spend equal fractions of time in each phase. The same is true of the control experiment and of S:GW+Ozone. But 1 month is only 3% of the much longer mean period for the S:Ozone sensitivity experiment, which spends 55% of the time in its westward phase, although as the experiment duration permits only two contributing QBO periods in this case, variability is possibly an issue. One clear difference from observations is that all 3 experiments have enhanced QBO amplitudes (by around 20%–30% in the eastward phase and 40%–60% in the westward phase). This appears to result from individual peaks of lower amplitude in the observed QBO but, as the QBO period from the analysis data is more variable, there is also a tendency for the normalizing process to smear the signal peak shapes which leads to a broadened, reduced peak on meaning.

Figure 11.

Equatorial 5°S–5°N mean zonal wind at 10 hPa averaged over QBO periods extracted from 12 year intervals, model (1979–1991) and assimilated observations from Met Office operational analyses (1994–2005). Solid line is control, dashed line is S:Ozone, dash-dotted line is S:GW+Ozone, and dotted line is assimilated observations.

[30] Contributions to the zonal wind acceleration (momentum budget) in the equatorial zone at 10 hPa can be meaned over the same time periods as the zonal winds in order to highlight their relationship with the descending zonal wind profiles. The 10 hPa momentum budget is shown in Figure 12, which depicts resolved wave (Eliassen-Palm flux divergence), parameterized wave (nonorographic gravity wave), and vertical advection (by mean upwelling) contributions for the control and sensitivity experiments. As described by Scaife et al. [2000], the parameterized waves add significant drag immediately beneath the descending zero wind transition, hence just prior to its arrival at 10 hPa, and deposit maximum momentum just after, when they act to accelerate the zonal wind flow. Maxima in the control (thick line) and S:Ozone sensitivity (medium line) runs are similar but, as deduced from Figures 10c and 10f, the S:GW+Ozone experiment (thin line) results in momentum deposition enhanced to 140% of the control value. By contrast, the resolved waves at 10 hPa appear to contribute only 18% of the total westward acceleration over the QBO cycle while their contribution to the eastward acceleration reduces from 7% in the control to just 3% in the S:GW+Ozone experiment. Interestingly, the relative contribution of resolved waves reported by Scaife et al. [2000] is clearly enhanced with respect to that seen in this case, where a very similar version of nonorographic gravity wave parametrization is used in a significantly upgraded GCM [Davies et al., 2005].

Figure 12.

Equatorial 5°S–5°N mean momentum budget at 10 hPa averaged over QBO period. Acceleration (m s−1 day−1) due to resolved waves (solid lines), parameterized waves (dashed lines), and vertical advection (dash-dotted lines). Zonal wind (m s−1) at 10 hPa (dotted lines) indicates relative phase sequence. Thick lines are control, medium lines are S:Ozone, and thin lines are S:GW+Ozone.

[31] As with the parameterized waves, the acceleration due to vertical advection of the horizontal wind also tends to maximize around the zero wind transition, because the upwelling traverses the strongest wind shear at these points, but it tends to oppose the parameterized wave acceleration especially in the descending easterly case. In the control experiment, 40% of the total eastward acceleration over the QBO cycle is provided by vertical advection and 53% by parameterized waves as opposed to 15% and 67%, respectively, of the total westward acceleration. This balance is shifted toward vertical advection for the S:Ozone sensitivity experiment (50% and 45% eastward and 22% and 60% westward), consistent with stronger upwelling. In the S:GW+Ozone experiment, however, enhanced parameterized gravity wave amplitudes come close to restoring the original balance (42% and 55% eastward and 13% and 71% westward) at the expense of marginally reducing further the relative contribution from resolved waves. Despite these differences, all three experiments present a consistent picture of eastward acceleration over the QBO cycle resulting from a relatively even balance between vertical advection and parameterized waves but westward acceleration having the greatest contribution from parameterized waves (60%), with the remaining acceleration split roughly equally between advection of zonal wind and resolved waves.

5. Summary

[32] Sensitivity experiments S:Ozone and S:GW+Ozone were run versus an 84 km top HadGEM control. The climatological ozone distribution in the sensitivity experiments differed from that in the control, with tropical stratospheric concentrations enhanced by up to 2.0 ppmv between 20–30 km, reduced by up to 2.7 ppmv from 30–40 km and enhanced above. These changes impact the MetUM profile of clear-sky radiative heating in the tropics, in particular warming the stratosphere between 15–35 km. Our results are consistent with a localized equilibrium response of the dynamics in the tropical stratosphere to these climatological ozone changes that generates increased upwelling between 100 hPa and 4 hPa which is sufficient to account for the described increase of mean QBO period from 26.5 months to 38.5 months. This increase is reversed in the S:GW+Ozone experiment by increasing the parameterized launch spectrum amplitude for upward propagating nonorographic gravity waves by 36%. QBO periods in the control (24–29 months, mean 26.5 months) and S:GW+Ozone experiment (23–29 months, mean 26.8 months) lie within the 22–35 months range observed over a 12 year period (1994–2005) of the Met Office operational stratospheric analysis record.

[33] This result has implications for analysis of the tropical circulation in CCMs with interactive ozone chemistry because it highlights the possibility that plausible changes in the ozone climatology, for instance as a response to greenhouse gas induced climate change [Butchart et al., 2006], could have a sizable impact upon the tropical upwelling and QBO period that ought to be distinguished from other dynamical responses such as ozone-temperature feedbacks. In this context, GCM simulations can assist in interpretation of full CCM experiments. For instance, the analysis carried out here could be repeated with an ozone climatology generated by a fully interactive CCM and that used by its radiation scheme when not coupled to the chemistry in order to assess the relative magnitudes of climatology and ozone-temperature feedback changes on the QBO dynamics. Or, if a more precise attribution of dynamical impact to specific features of the ozone perturbation were required, it would be possible to construct climatologies with just those perturbed features included. Such equilibrium type runs are naturally no guarantee that interactive ozone CCMs will behave identically but they can at least identify quantifiable signals for which to probe.

[34] One aspect of the CCM simulation that can be addressed with GCM experimentation is the sensitivity of the QBO dynamics to other influences, such as the strength of nonorographic gravity waves in S:GW+Ozone. This shows that the effects of a step change in the ozone climatology upon the MetUM representation of the tropics can be mitigated, with the obvious caveat that it is important to use the best available estimate of ozone climatology as any errors can lead in compensation to an inappropriate setting of the gravity wave strength for other times or locations. With similar caveats, where interactive ozone in a CCM might equilibrate to a climatology distinctly different from the initial distribution, a similar strategy might work. However, although the magnitude of changes in ozone shown in Figure 2d are similar to reported ranges of ozone variability in the QBO [Butchart et al., 2003; Shibata and Deushi, 2005a] it is not clear how a switch in strength of wave forcing will interact with the time varying radiative heating induced by the ozone cycle. In reality, the waves that drive the QBO are likely to originate from tropical convection sources [Garcia and Sassi, 1999] which vary on short, intraseasonal and interannual timescales that most probably explain the stalling and irregular behavior of the QBO descending shear zones and also affect the transport of ozone into the stratosphere. Such complexity, however, lies beyond the scope of this study.

[35] Whereas Giorgetta et al. [2006] reported eastward phase propagation that depended more strongly on resolved-scale waves than on parameterized, investigation of the contributions to zonal wind acceleration over the equatorial (5°S–5°N) region at 10 hPa indicate a reduced role for resolved waves relative to parameterized waves in this model by comparison with its predecessor [Scaife et al., 2000]. As the horizontal resolutions are similar, differences are currently thought to arise from the numerical formulation of the present MetUM. Damping of fast propagating signals at the resolved length scale can arise from the MetUM's semi-implicit off centering in time and from interpolation prior to the semi-Lagrangian advection. Kelvin waves can also be prevented from upward propagation if the chosen vertical resolution is too coarse but the L60 resolution employed here was selected to avoid this. The greater dependence upon parameterized waves is an area that merits further investigation as results from the high-resolution middle atmosphere GCM study of Watanabe et al. [2008] indicate appreciable contributions to the mean flow from wave forcing on all scales from planetary down to 190 km, which is subgrid in this study. This suggests that there could be a broader range scale than might have been anticipated of resolved waves on the order of the grid length (around 300 km in the tropics) that are not explicitly represented, leaving the subgrid parametrization scheme to pick up their contribution. The disadvantage of such a shift is that resolved waves can interact with tropical convection and thus introduce feedbacks with other physical processes represented by the GCM. These feedbacks could in turn alter the QBO response to changes in ozone or greenhouse gases with implications for chemical transport of source gases into the tropical stratosphere and for wider connections to extratropical phenomena such as the Northern Annular Mode.

Acknowledgments

[36] N. Butchart and S. Hardiman were supported by the Joint DECC and Defra Integrated Climate Programme (DECC/Defra) (GA01101). S. Osprey and L. Gray were supported by the NERC National Centre for Atmospheric Science (NCAS) climate directorate. The authors would like to thank Gill Martin for helpful suggestions and input to development of the 60 level HadGEM configuration.