Abstract
 Top of page
 Abstract
 1. Introduction
 2. Climate Simulations and Methods
 3. Results
 4. Conclusion
 Acknowledgments
 References
 Supporting Information
[1] Recent climate studies show that the northernwinter wave driving of the BrewerDobson circulation is enhanced if greenhouse gas concentrations increase. An explanation for this enhancement does not yet exist. In this study, the enhanced wave driving, as simulated in a doubledCO_{2} experiment with the MAECHAM4 climate model, is analyzed in detail. The extratropical poleward eddy heat flux increases (decreases) in the stratosphere (troposphere) mainly due to the stationary (transient) heatflux component. The heat flux at 100 hPa is a measure of the stratospheric wave driving, and is found to increase by 12% in the doubledCO_{2} climate. This increase is dominated by the stationarywave 1 heat flux, which is also enhanced in the midlatitude troposphere. The heat flux increase at 100 hPa is almost entirely due to an increase in the longitudinal temperature variability. The latter increase is mainly due to the wellunderstood sharpening of the lowerstratospheric meridional temperature gradient.
1. Introduction
 Top of page
 Abstract
 1. Introduction
 2. Climate Simulations and Methods
 3. Results
 4. Conclusion
 Acknowledgments
 References
 Supporting Information
[2] Globalmean CO_{2} concentrations have risen by about 10% in the 1979–2002 period, but a structural trend in the planetarywave driving of the BrewerDobson circulation (BDC) is not yet observed during northern winter, likely due to a low signaltonoise ratio over this relatively short period [e.g., Haklander et al., 2007]. However, Butchart et al. [2006] recently found that a more substantial increase of greenhouse gas (GHG) concentrations yields an overall strengthening of the BDC and the associated upward propagation of planetarywave activity, with the strongest trend during northern winter. They found this result by comparing the response of the BDC to increasing GHG concentrations in an ensemble of general circulation models (GCM), with some models including interactive chemistry. The causes and nature of the strengthening are not yet clear. Shindell et al. [1999] suggested that increased GHG concentrations enhance the subtropical jet, so that more tropospheric planetarywave activity at midlatitudes propagates into the stratosphere, thereby strengthening the BDC. The physical explanation of this was, that near 200 hPa, the meridional temperature gradient increases due to lowerstratospheric cooling on the poleward side and uppertropospheric warming on the equatorward side of the subtropical jet. The associated enhancement of the subtropical jet would then cause a decrease in the equatorward refraction of tropospheric wave activity at midlatitudes. Eichelberger and Hartmann [2005] studied the qualitative effect of increasing GHG concentrations on the BDC by imposing a tropical tropospheric heat source in their GCM. This dry hydrostatic GCM had a flat lower boundary, and therefore did not simulate stationary waves at all. It was found that the BDC strengthens in response to the imposed tropospheric tropical heating, with an increase in upward propagation of transient wave2 and wave5 activity in the troposphere. However, only the transient wave2 increase extended upward into the stratosphere.
[3] Sigmond et al. [2004] compared the BDC in a control climate and a doubledCO_{2} climate simulation. The BDC was found to be significantly stronger in the doubledCO_{2} run. Although the vertical component of the EliassenPalm (EP) flux increased in the midlatitude lower stratosphere, a decrease was found in the midlatitude troposphere. Sigmond et al. [2004] concluded that this might be caused either by a higher transparancy of the NH midlatitude tropopause for tropospheric wave activity, or by more generation of wave activity near the tropopause. Also, a modification of the gravitywave spectrum could alter the wave driving of the BDC during northern winter. In the present study, the processes via which the resolved wave driving of the BDC could increase in an enhancedCO_{2} climate are examined in greater detail, using the same climate simulations as used by Sigmond et al. [2004].
2. Climate Simulations and Methods
 Top of page
 Abstract
 1. Introduction
 2. Climate Simulations and Methods
 3. Results
 4. Conclusion
 Acknowledgments
 References
 Supporting Information
[4] This study compares two climate simulations which have been performed by Sigmond et al. [2004], using the middleatmosphere (MA) version of the ECHAM4 model [Manzini et al., 1997; Roeckner et al., 1996]. The model has a T42 horizontal spectral resolution (about 2.8° × 2.8°), with 39 vertical levels between the surface and 0.01 hPa (about 80 km). Momentum deposition by subgrid scale waves is parameterized for both orographic gravity waves [McFarlane, 1987] and for a transient gravitywave spectrum [Hines, 1997a, 1997b]. One simulation is a 30yr control run, in which CO_{2} concentrations were fixed at 353 ppmv. In the perturbation run, CO_{2} concentrations were doubled. The prescribed seasurface temperatures were obtained from a control run and a doubledCO_{2} run with the ECHAM4 model coupled to a slab layer ocean model.
[5] We use the zonalmean poleward eddy heat flux [v*T*] as a measure of the net zonalmean upward flux of planetarywave activity, since [v*T*] is proportional to the vertical component of the EP flux for quasigeostrophic waves. We define H_{100} as the average of [v*T*] at 100 hPa over January–February and 40°–80°N. This (or a similar) diagnostic has been used in several studies [e.g., Austin et al., 2003; Haklander et al., 2007]. By partitioning the response of H_{100} to the CO_{2} doubling into its stationary and transient components for different zonal wavenumbers, we examine the causes and nature of the increase in the northern midwinter wave driving in greater detail. Here, the zonal wavenumberk component of [v*T*] is defined as the zonalmean product of the wavek components of v and T.
[6] Additionally, the poleward eddy heat flux is decomposed by noting that it equals the zonal covariance between v and T, i.e.,
where r_{v,T} is the zonal correlation coefficient of v and T, and σ_{v} and σ_{T} are the zonal standard deviations of v and T. Both σ_{v} and σ_{T} are indicators of wave amplitude, while r_{v,T} is a measure of how effectively waves transport the sensible heat poleward. For monochromatic waves, r_{v,T} is directly proportional to the cosine of the zonal phase difference between the v and T fields. CO_{2} doubling could modify both the amplitude of the waves and the efficacy of the poleward eddy heat transport.
[7] By definition, H_{100} is obtained by evaluating the l.h.s. of Equation 1 at 100 hPa and averaging over January–February and 40°–80°N. If the three r.h.s. terms are averaged over January–February and 40°–80°N separately, the product of those averages might be different from the average of the singlevariable product, due to intraseasonal and meridional crosscorrelations between r_{v,T}, σ_{v}, and σ_{T}. Averaging the three factors on the r.h.s. of Equation 1 over January–February and 40°–80°N at 100 hPa, and multiplying them, yields an approximation of H_{100} which can be compared to H_{100} itself. For both runs, the 30year timeseries of this approximation of H_{100} follows that of H_{100} itself very closely: correlation coefficients are 0.96 and 0.89 for the control and doubledCO_{2} runs, respectively. Henceforth, to keep notation simple, r_{v,T}, σ_{v}, and σ_{T} will denote the corresponding averages over January–February and 40°–80°N at 100 hPa. We examine if the difference in H_{100} between the control and doubledCO_{2} runs can be understood in terms of the differences in r_{v,T}, σ_{v}, and σ_{T}. These decompositions of the poleward eddy heat flux have been discussed in more detail by Haklander et al. [2007, section 2.3]. In the present paper, the uncertainties provided along with the estimates represent the 95% confidence intervals. Differences between the control run and the doubledCO_{2} run are considered statistically significant if the confidence level exceeds 95%.
3. Results
 Top of page
 Abstract
 1. Introduction
 2. Climate Simulations and Methods
 3. Results
 4. Conclusion
 Acknowledgments
 References
 Supporting Information
[8] We first examine the effect of CO_{2} doubling on H_{100} by comparing the 30year averages of the control and doubledCO_{2} runs. The result is shown in Table 1, along with the same comparison for several wave components of H_{100}. The total H_{100} controlrun average of 15.0 ± 0.9 K m/s agrees remarkably well with the observed 1979–2002 average of 15.1 ± 1.1 K m/s in the ERA40 dataset [Haklander et al., 2007]. In our model, doubling the CO_{2} concentrations yields a significant and substantial increase in H_{100} of 1.8 ± 1.2 K m/s, or 12% ±8% of the control average.
Table 1. The 30Year Averages for Run C and Run A and Differences Between the A and CRuns for H_{100} and Its Total, Stationary and Transient k = 1–5 Components^{a}Component  CRun Mean, K m/s  ARun Mean, K m/s  Differences, K m/s  Relative Differences, % 


Total  15.0 ± 0.9  16.8 ± 0.8  +1.8 ± 1.2  +12% ± 8% 
Stationary  8.6 ± 1.0  10.6 ± 0.9  +2.0 ± 1.3  +23% ± 16% 
Transient  6.4 ± 0.6  6.2 ± 0.5  −0.1 ± 0.8  −2% ± 12% 
Total 1  8.6 ± 1.0  9.2 ± 0.7  +0.6 ± 1.2  +7% ± 14% 
Total 2  3.2 ± 0.7  3.4 ± 0.5  +0.3 ± 0.8  +8% ± 26% 
Total 3  1.7 ± 0.4  2.1 ± 0.3  +0.4 ± 0.5  +24% ± 27% 
Total 4  0.5 ± 0.2  0.5 ± 0.2  +0.1 ± 0.3  +12% ± 56% 
Total 5  0.5 ± 0.1  0.9 ± 0.1  +0.4 ± 0.2  +86% ± 40% 
Stationary 1  6.5 ± 1.0  8.2 ± 0.7  +1.7 ± 1.3  +27% ± 20% 
Stationary 2  1.7 ± 0.6  1.8 ± 0.5  +0.1 ± 0.8  +5% ± 46% 
Stationary 3  0.5 ± 0.3  0.6 ± 0.2  +0.1 ± 0.4  +26% ± 80% 
Stationary 4  −0.2 ± 0.1  −0.2 ± 0.1  +0.0 ± 0.1  +1% ± 60% 
Stationary 5  0.1 ± 0.1  0.2 ± 0.0  +0.1 ± 0.1  +49% ± 59% 
Transient 1  2.2 ± 0.4  1.1 ± 0.3  −1.1 ± 0.5  −51% ± 25% 
Transient 2  1.5 ± 0.4  1.6 ± 0.3  +0.2 ± 0.5  +12% ± 35% 
Transient 3  1.2 ± 0.3  1.5 ± 0.2  +0.3 ± 0.3  +23% ± 28% 
Transient 4  0.7 ± 0.1  0.8 ± 0.2  +0.1 ± 0.2  +8% ± 30% 
Transient 5  0.4 ± 0.1  0.7 ± 0.1  +0.4 ± 0.1  +99% ± 45% 
[9] In both the control and the perturbation run, stationary waves dominate the influx of planetary wave activity into the lower stratosphere. Doubling the CO_{2} concentration in the model yields a significant stationarywave flux increase of 23% ±16%, determining almost entirely the total H_{100} increase. The stationarywave flux increase occurs mainly in its stationary wave1 component. Of the longest five stationary zonal wave components, only wavenumber 1 exhibits a noticeable change. This suggests that the total wave response to the CO_{2} doubling can mainly be attributed to stationary wave 1. Considering the transient waves, we find a significant and substantial reduction in the transient wave1 flux. However, this reduction is neutralized by small increases in the transient wave 2–5 components. Although the change in the transient wave 2–4 components is not significant, the statistical significance of the increase in the transient wave5 flux is extremely high. Figures 1 and 2 show a large vertical coherence near 100 hPa. At the adjacent model levels (70 and 150 hPa), the increases in ‘H_{70}’ and ‘H_{150}’ are 1.3 ± 1.5 K m/s (8% ± 9%), and 0.6 ± 1.2 K m/s (4% ± 7%), respectively. The significant changes in the stationary wave1, and transient wave1 and wave5 fluxes are also found at 70 and 150 hPa, with the dominant change being an increase in the stationary wave1 flux.
[10] So far, we have only discussed 40°–80 °N averages at 100 hPa, and not yet considered the pattern of the difference in [v*T*] in the meridional plane. It was mentioned in the previous section, that for quasigeostrophic waves [v*T*] is proportional to the vertical component of the EP flux. The meridional pattern of the EPflux vector differences can elucidate the effect of CO_{2} doubling on the propagation of planetarywave activity. Figure 1 shows these difference patterns for the total, stationary and transient components of the EP flux. At midlatitudes between 400 and 10 hPa, the total upward EP flux increases significantly. A significant decrease is found in the troposphere north of about 25°N. In the stratosphere the total wave response (Figure 1a) is similar to the stationarywave response (Figure 1b), whereas in the troposphere it is more similar to the transientwave response (Figure 1c). In the subtropical troposphere, the doubledCO_{2} run exhibits significantly less equatorward propagation of (transient) wave activity, indicated by the poleward orientation of the EP flux difference vectors in Figures 1a and 1c. Starting in the subtropical troposphere and following the difference vectors according to raytracing theory, we see that the subtropical reduction of equatorward wave propagation is likely associated with less transient upward wave propagation in the midlatitude (lower) troposphere to begin with. The latter is expected to be due to a decrease in baroclinicity, since the tropospheric poletoequator temperature gradient is reduced in the perturbation run [Sigmond et al., 2004] (Figure 2a).
[11] Since a significant difference between the control and perturbation runs was found for the stationary wave1 and transient wave1 and wave5 components of H_{100}, we also show the meridional cross sections of the EP flux difference for those three wave components in Figure 2. For stationary wave1 (Figure 2a), a marked increase in the upward EP flux is observed particularly at midlatitudes, although the increase is not significant in the lower troposphere. Nevertheless, the contour pattern in Figure 2a indicates that the increase in stationary wave1 activity flux at 100 hPa is associated with an increase in upward stationary wave1 flux in the entire troposphere. Table 1 suggests that the total wave response to the CO_{2} doubling can mainly be attributed to stationary wave 1. This is confirmed by the strong similarity in the lower stratosphere between the total difference pattern in Figure 1a and the stationary wave1 difference pattern in Figure 2a. Figure 2b shows that the substantial decrease in the transient wave1 component of H_{100} is due to a flux decrease north of about 52°N. The EPflux difference for the transient wave5 component at 100 hPa (Figure 2c) exhibits a significant increase at midlatitudes.
[12] We next examine the effect of CO_{2} doubling on the amplitude of the waves, and their correlation, as described in the previous section. Table 2 shows the 30year averages and differences for H_{100}, its approximate value r_{v,T}σ_{v}σ_{T}, and r_{v,T}, σ_{v}, and σ_{T}. The increase in r_{v,T}σ_{v}σ_{T} in the doubledCO_{2} climate is statistically significant and quantitatively comparable to the increase in H_{100}. The changes in r_{v,T} are very small and not significant. However, the zonal standard deviations of both v and T do show a significant increase. Whereas σ_{v} exhibits only a marginal increase, the increase in σ_{T} is found to be the dominant cause of the increased wave driving. Not only is the change in σ_{T} significant, the relative change is substantial (+12%± 5%) and comparable to the change in H_{100} (+12% ± 8%). A significant increase in σ_{T} is also found at 70 and 150 hPa, and the percent increase in σ_{T} is comparable with the percent increase in the total heat flux at both levels (not shown). In Figure 3, the January–February mean total eddy temperature field at 100 hPa is shown for the control and the perturbation runs. The stationary wave1 signature is well visible for both experiments, and Figure 3 clearly demonstrates the increase in longitudinal temperature variability in the doubledCO_{2} run. The increase in σ_{T} may be due to an increase in the meridional temperature gradient between 40°–80°N at 100 hPa, as meridional air displacements will produce larger zonal temperature asymmetries if the zonalmean meridional temperature gradient increases. Indeed, the 30year average January–February temperature difference between 40°N and 80°N at 100 hPa increases by about 10% in the perturbation run (not shown). Below 400 hPa, the temperature difference between 40°N and 80°N decreases in the doubledCO_{2} run, with the largest reduction in the lower troposphere. This result agrees with the significant reduction in midlatitude upward EP flux at those levels that was discussed above.
Table 2. The 30Year Averages for the Control and DoubledCO_{2} Runs and Differences Between the Two Runs in H_{100} and r_{v,T}σ_{v}σ_{T} and the Separate Factors r_{v,T}, σ_{v}, and σ_{T}^{a}Component  Control  2 × CO_{2}  Differences  Relative Differences, % 


H_{100}, K m/s  15.0 ± 0.9  16.8 ± 0.8  +1.8 ± 1.2  +12% ± 8% 
r_{v,T}σ_{v}σ_{T}, K m/s  14.1 ± 0.9  16.4 ± 0.8  +2.3 ± 1.2  +16% ± 8% 
r_{v,T}  0.213 ± 0.007  0.216 ± 0.013  +0.002 ± 0.015  +1% ± 7% 
σ_{v}, m/s  10.1 ± 0.2  10.5 ± 0.3  +0.4 ± 0.4  +4% ± 3% 
σ_{T}, K  6.5 ± 0.2  7.3 ± 0.2  +0.8 ± 0.3  +12% ± 5% 
4. Conclusion
 Top of page
 Abstract
 1. Introduction
 2. Climate Simulations and Methods
 3. Results
 4. Conclusion
 Acknowledgments
 References
 Supporting Information
[13] In this paper, the effect of CO_{2} doubling on the northernwinter upward wave activity flux at midlatitudes has been analyzed, using the MAECHAM4 climate model. The upward wave flux is quantified by H_{100}, defined as the average of [v*T*] at 100 hPa over January–February and 40°–80°N. Doubling the CO_{2} concentration leads to a substantial and significant increase in H_{100} of 1.8 ± 1.2 K m/s, or 12% ± 8%. This is, at least qualitatively, in agreement with the studies by, e.g., Butchart and Scaife [2001], Rind et al. [1998, 2002], and Butchart et al. [2006]. Table 1 and Figures 1a and 2a indicate that stationarywave 1 likely accounts for almost all of the total increase in H_{100}. This suggests that at least part of the increase is due to more stationary wave1 generation at midlatitudes in the (lower) troposphere. The results indicate that doubling the CO_{2} yields significant changes in transient wave1 and transient wave5 as well. The latter change is in agreement with the result of Eichelberger and Hartmann [2005], but their model setup (no topography) was quite different from that in the present study. Transient wave5 may not seem very relevant due to its small contribution, but its increase amounts to 19 ± 15% of the total H_{100} increase.
[14] Using an alternative decomposition of the poleward eddy heat flux, we found that the increased wave driving can be attributed mainly to a larger longitudinal temperature variability, that is mainly due to the increased meridional temperature gradient at 100 hPa in the doubledCO_{2} climate simulation. With only one climate model at our disposal, we were not able to test the robustness of our results. However, both the increase in the northernwinter wave driving of the BDC and the increase in the 100hPa meridional temperature gradient are robust features in enhancedCO_{2} climate simulations. We would welcome studies in which our analysis method is applied to data from other enhancedCO_{2} simulations.