Sensitivity of climate to dynamically-consistent zonal asymmetries in ozone



[1] Previous investigations into the effect of zonal asymmetries in ozone on climate have compared simulations with prescribed 3-D ozone, in which the ozone is not necessarily consistent with the model dynamics, to simulations with prescribed zonal mean ozone. We assess the impact of zonal asymmetries in ozone by comparing a control simulation of a coupled chemistry version of the Canadian Middle Atmosphere Model (CMAM) in which the ozone and model dynamics are consistent, with a simulation in which only the zonal mean of the ozone is passed to the radiative transfer scheme. These simulations reveal a robust stratospheric zonal-mean temperature and geopotential height response to zonal asymmetries in ozone that is consistent with that identified in previous studies and of a magnitude comparable to observed trends. These results suggest that the inclusion of zonal asymmetries in ozone may be essential for the accurate simulation of future stratospheric temperature trends.

1. Introduction

[2] To date almost all coupled atmosphere-ocean climate models [e.g., Meehl et al., 2007], and many model simulations of the middle atmosphere [e.g., Ramaswamy et al. 2001] have used specified zonal average ozone distributions. Son et al. [2008] demonstrated that the Southern Hemisphere tropospheric circulation response to ozone recovery is larger in a set of coupled chemistry models than in a set of climate models in which zonal mean ozone is specified. One possible reason for this difference is that the coupled chemistry models included zonally asymmetric changes in the ozone distribution. Several recent studies have highlighted the importance of zonal asymmetries in ozone for the simulation of stratospheric and tropospheric conditions in the Northern Hemisphere [Kirchner and Peters, 2003; Gabriel et al., 2007], and in the Southern Hemisphere [Crook et al., 2008]. Zonal asymmetries in ozone are particularly large in the Southern Hemisphere during the breakup of the vortex, when the region of maximum ozone depletion is often displaced from the pole. Crook et al. [2008] demonstrated that accounting for zonal asymmetries in ozone in a simulation using observed ozone from a year with particularly strong zonal asymmetry resulted in stratospheric cooling comparable to that due to ozone depletion itself. However, these studies all specified fixed three-dimensional distributions of ozone: In such simulations the position of the ozone minimum is not constrained by the dynamics, and may not be collocated with the dynamical polar vortex. Further, ozone-dynamics feedbacks [Nathan and Cordero, 2007] are not resolved, since the ozone distribution is specified.

[3] Most studies on the influence of the zonal asymmetry of ozone have focused on the stratosphere. However, large zonal asymmetries in ozone are found in the mesosphere and thermosphere associated with the diurnal cycle. Since ozone concentrations are much higher at night than in the day at these levels, using a zonal mean of ozone has the potential to introduce a bias in the radiative heating rates at these levels. For this reason Paul et al. [1998] restrict their ozone climatology to levels below 0.3 hPa. However the widely-used [Li and Shine, 1995] zonal-mean ozone climatology extends to 0.0011 hPa, but uses daytime ozone values from near-infrared airglow measurements from the Solar Mesosphere Explorer [Li and Shine, 1995]. However, in some cases climate models are run with prescribed ozone from coupled chemistry models, particularly for future simulations, and in these cases the use of zonal mean ozone could introduce a bias in the radiative heating rates.

[4] One way in which the influence of zonal asymmetries in ozone may be examined in a more realistic context is by comparing a simulation of a coupled chemistry model, with a second simulation in which the zonal mean of the calculated ozone is prescribed. Reddmann et al. [1999] carried out such a comparison using the 3-D Karlsruhe simulation model of the middle atmosphere (KASIMA), accounting for the diurnal cycle in ozone above 50 km by prescribing the daytime mean ozone above this level in the second simulation, and with conditions at the 10-km lower boundary prescribed from European Centre for Medium Range Weather Forecasting (ECMWF) analyses. Their simulations start in July, meaning that they are not able to realistically simulate the Antarctic winter vortex or Antarctic ozone depletion. They simulate a single northern winter, and use trace gases representative of approximately 1992. They examine temperature differences between the two simulations in December and find only small differences between the two simulations, which leads them to conclude that ‘for undisturbed ozone conditions (no polar ozone hole) a realistic ozone climatology is sufficient for model simulations’. In this study, we also assess the role of three dimensional ozone variations in a fully-coupled context, but using 40-yr simulations with a realistic annual cycle and stratospheric chlorine levels representative of approximately 1990.

2. Model and Experiments

[5] We use the Canadian Middle Atmosphere Model with coupled chemistry (CMAM) [Scinocca et al., 2008; de Grandpré et al., 2000], a version of the Canadian Centre for Climate Modelling and Analysis Third Generation atmospheric model. CMAM is a spectral model with a horizontal resolution of T31, and 71 levels extending up to 0.0004 hPa, or about 95 km. The chemistry module includes a standard set of stratospheric gas phase chemical reactions and heterogeneous reactions on Type-Ib and Type-II polar stratospheric clouds [de Grandpré et al., 2000]. Tropospheric halocarbon concentrations were fixed at their 1985 values, giving stratospheric chlorine loading representative of about 1990, taking into account the delay due to transport. The control (CTL) for the present study was a 40-yr simulation of CMAM coupled to an ocean general circulation model with a horizontal resolution of 1.875° (V. K. Arora et al., The 20th century carbon budget simulated with the CCCma earth system model CanESM1, submitted to Journal of Climate, 2009).

[6] The impact of zonal asymmetries in CMAM's ozone is assessed by a comparison of 40-yr means from CTL with 40-yr means from a second coupled chemistry simulation in which only the zonally symmetric component of the ozone is passed to the radiative transfer scheme at each time step. A simulation in which the ozone was zonally symmetrized at all altitudes produced a band of heating in the mesosphere due to the averaging of nighttime (high) and daytime (low) concentrations of ozone, which differ by roughly an order-of-magnitude [de Grandpré et al., 2000]. This is illustrated in Figure 1 where the annual mean temperature difference shows a pole-to-pole warming of up to 17 K in the mesosphere. Thus, in the remainder of this study we consider a 40-yr simulation (ZM) in which the zonal symmetrization of ozone in the radiation was made to transition smoothly back to the CMAM's 3-D ozone above 0.3 hPa. Another simpler approach is to simply use daytime mean values of ozone in the mesosphere. This would, for example, parallel the approach followed by Reddmann et al. [1999], where daytime mean values of ozone were specified above 50 km. One drawback of this approach, however, is that the nighttime long-wave terrestrial radiation sees the low daytime ozone concentrations and this could lead to a bias.

Figure 1.

Annual mean zonal mean temperature difference between a simulation in which zonal mean ozone is passed to the radiation scheme throughout the atmosphere and a control.

3. Results

[7] Figure 2a shows the standard deviation of ozone about its zonal mean in CTL. The largest zonal asymmetries in ozone (∼15% of the zonal mean) are found in the lower stratosphere at ∼70°N and ∼65°S, consistent with Crook et al.'s [2008] findings based on the ECMWF 40-yr Reanalysis (ERA-40). However, while CMAM shows larger mean asymmetries in the Northern Hemisphere, Crook et al. [2008] found larger asymmetries in the Southern Hemisphere during the 1990s in ERA-40. Since CMAM slightly underestimates Arctic ozone depletion (A. Karpechko, personal communication, 2008), this difference cannot be explained by differences in the extent of Arctic ozone depletion, but could relate to dynamical variability of ozone or errors in the reanalysis. Since we infer the impact of zonal asymmetries in ozone from differences between CTL and ZM, it is important to verify that the zonal mean ozone in the two simulations is consistent. Figure 2b shows this annual mean difference in ozone mass mixing ratio on the same scale as Figure 2a. Differences in zonal mean ozone mass mixing ratio between the two simulations, likely dynamically-driven, are small compared to variations about the zonal mean, and we expect any difference in the two simulations to be dominated by the effect of zonal asymmetries in the control simulation.

Figure 2.

(a) Annual mean standard deviation of monthly mean ozone mass mixing ratio from the zonal mean in CTL, a measure of the degree of zonal asymmetry of ozone at each latitude and pressure level. (b) The difference in zonal mean ozone mass mixing ratio between CTL and ZM (CTL-ZM).

[8] In the Antarctic, the zonal asymmetry in ozone is largest in November (Figure 3a) coincident with the time of maximum ozone depletion [Crook et al., 2008]. However, there is a secondary maximum in April, and both November and April are associated with maxima in the zonal asymmetry of geopotential height at these levels (not shown). In the Arctic, where ozone depletion is much less marked and at a maximum in January and February in CMAM (A. Karpechko, personal communication, 2008), the largest zonal asymmetry in ozone is found in November (Figure 3b), suggesting that this is associated primarily with dynamical variability rather than with a region of depleted ozone located off the pole. In both hemispheres, zonal asymmetries in ozone are small when the polar stratosphere is undisturbed by planetary waves: during winter and summer in the Antarctic, and during summer in the Arctic.

Figure 3.

(left) Antarctic (90°S–66°S) and (right) Arctic (66°N–90°N) monthly mean zonal means of (a and b) standard deviation of ozone mass mixing ratio about its zonal mean, (c and d) differences in temperature between CTL and ZM, and (e and f) differences in geopotential height between CTL and ZM (CTL-ZM). Hatching in the lower plots indicates differences significant at the 5% level based on a two-sample t-test.

[9] Antarctic mean temperature (Figure 3c) and geopotential height (Figure 3e) exhibit a similar pattern of response to that identified in simulations with prescribed ERA-40 ozone by Crook et al. [2008]: A downward propagating cooling in the lower stratosphere in October–December, overlaid by a warming, corresponding to a delay in the final warming and a strengthened polar vortex in November and December (Figure 3e). In CMAM the magnitude of the response is approximately two-thirds of that shown by Crook et al. [2008], perhaps because the latter prescribed ozone from a year which exhibited particularly strong zonal asymmetry. The response also occurs somewhat later in the CMAM simulations, which is likely related to the Antarctic vortex breaking up too late in CMAM [Eyring et al., 2006; Scinocca et al., 2008]. Nonetheless, these results confirm Crook et al.'s [2008] finding that zonal asymmetries in ozone have a significant effect on zonal mean quantities of a magnitude comparable to the response to ozone depletion. Crook et al. [2008] demonstrated that the temperature response was driven by changes in dynamical heating which they speculated were associated with a change in the shape and location of the vortex influencing the ability of planetary waves to propagate upwards into the stratosphere.

[10] In the Arctic a warming response > 3 K is seen in November and December somewhat above the region of maximum asymmetry in ozone. Although the zonal asymmetry in ozone is largest in these months, it might be expected to have little effect on Arctic temperature because of the absence of sunlight: Presumably the effect is driven by zonal asymmetries in ozone outside of the region of polar night.

[11] Lastly we examine the horizontal structure of the ozone and temperature anomalies. Figure 4a shows ozone anomalies from zonal mean over the Antarctic in October at 50 hPa (month and level chosen for comparison with Crook et al. [2008]). This pattern of ozone anomalies is similar to that shown for October 2000 in the ERA-40 reanalysis by Crook et al. [2008], with a relative minimum in ozone located over the Weddell Sea, and a maximum over the coast of East Antarctica. The magnitude of the anomalies is much smaller in CMAM, but this may be partly because CMAM anomalies are a mean over 40 years, whereas Crook et al. [2008] show one year with particularly strong zonal asymmetry. Figure 4c shows the temperature anomalies from the zonal mean in CTL, also in October at 50 hPa, and while weaker in magnitude, the pattern is in agreement with Crook et al. [2008]. As expected the ozone minimum and temperature minimum approximately coincide. Figure 4e shows the temperature response to the zonal asymmetry in ozone (CTL-ZM) and consistent with Crook et al. [2008], we find cooling approximately collocated with a positive ozone anomaly and warming associated with a negative ozone anomaly, albeit of a smaller magnitude. Note that this antiphase relationship between the ozone anomaly and the temperature response is not preserved over all levels and months, but does appear to be robust in October at this level. This pattern of response may be associated with the delay in breakdown of the vortex in CTL and a tendency for the vortex to remain centered close to the pole for longer.

Figure 4.

(left) Antarctic October 50-hPa and (right) Arctic January 10-hPa ozone and temperature. (a and b) Ozone mass mixing ratio anomalies from the zonal mean in CTL, (c and d) temperature anomalies from the zonal mean in CTL, and (e and f) temperature differences between CTL and ZM (CTL - ZM). Levels and months were chosen for comparison with Crook et al. [2008] in the Antarctic and Gabriel et al. [2007] in the Arctic.

[12] In the Arctic CMAM simulates similar zonally asymmetric ozone features to those shown for the 1990s January mean at 10 hPa in ERA-40 by Gabriel et al. [2007] (Figure 4b). While the magnitude of the zonal asymmetries is somewhat smaller in CMAM, the pattern is well-reproduced, suggesting that it is determined by robust features of the background climate which are reproduced in the model. The pattern of zonally-asymmetric temperature anomalies is also consistent with that simulated [Gabriel et al., 2007]. Further, the simulated response to the zonal asymmetries in ozone (Figure 4f) is also relatively similar to that shown by Gabriel et al. [2007], with a cooling maximum close to Hudson Bay, and a warming simulated over Europe, albeit somewhat further south and east in CMAM. While the response is not in antiphase with zonal asymmetries in ozone, as in the Southern Hemisphere, much of the warming response over Europe is collocated with a region of negative ozone anomalies, as also found by Gabriel et al. [2007]. As in the Southern Hemisphere, the phase relationship between the ozone anomaly and the temperature response varies with month and pressure level. Reddmann et al. [1999] find only a small temperature response to zonal asymmetries in ozone in the Northern Hemisphere at 2 hPa on December 30th, and conclude that the response to zonal asymmetries in ozone is weak. We find no significant temperature response over the Arctic at this level and time (Figure 3d), but we find a significant temperature response at 2 hPa earlier in the year, and at other levels in December. Our results thus disagree with Reddmann et al.'s [1999] conclusion that zonal asymmetries in ozone are unimportant in the Arctic under early 1990s conditions.

4. Discussion and Conclusions

[13] These simulations of the coupled chemistry version of CMAM confirm that the response to zonal asymmetries in ozone simulated in the Northern Hemisphere by Gabriel et al. [2007] and in the Southern Hemisphere by Crook et al. [2008] are robust in a model setting in which ozone is consistent with the model dynamics. The main zonally-asymmetric features of the ozone and temperature distribution are reproduced in CMAM, suggesting that these are robust climate features, likely set by the location of wave sources in each hemisphere. We find significant zonal mean temperature responses to the zonal asymmetry in ozone of up to 4 K in the lower stratosphere. Since the zonally asymmetric components of the ozone distribution have changed over time due to ozone depletion, we conclude that the zonally asymmetric component of the ozone distribution should be included in physical climate models if we wish to simulate realistic zonal mean stratospheric temperature trends.


[14] We thank Larry Solheim (CCCma, Victoria) for extracting diagnostics from the simulations we use here. NPG acknowledges support from the Leverhulme Trust.