Tropopause height and zonal wind response to global warming in the IPCC scenario integrations

Authors


Abstract

[1] The change in the extratropical circulation under global warming is studied using the climate models participating in the Intergovernmental Panel on Climate Change (IPCC) fourth assessment report. The IPCC models predict a strengthening and a poleward shift of the tropospheric zonal jets in response to global warming. The change in zonal jets is also accompanied by a strengthening and a poleward and upward shift of transient kinetic energy and momentum flux. Similar changes in circulation are simulated by a simple dry general circulation model (GCM) when the height of the tropopause is raised. The similarity between the simple GCM and the IPCC models suggests that the changes in midlatitude circulation are predominantly driven by a rise in the height of the tropopause, and that other factors such as increased moisture content and the change in the low-level pole-to-equator temperature gradient, play a secondary role. In addition, the variability about the ensemble-mean of the zonal wind response is significantly correlated with the variability of the tropopause height response over the polar cap, especially in the Southern Hemisphere.

1. Introduction

[2] In response to increased CO2 concentration, the atmospheric temperature increases in the troposphere but decreases in the stratosphere [Manabe and Wetherald, 1967, 1980]. The temperature decreases in the stratosphere occur because to first order the dominant balance in the stratosphere is between warming due to shortwave absorption by ozone and cooling due to longwave emission by CO2 [e.g., Held, 1993]. Therefore an increase in CO2 leads to more longwave cooling in the stratosphere. The warming in the troposphere and the cooling in the stratosphere decreases the static stability in the region of the tropopause, or, in other words, the cooling in the stratosphere and the warming in the troposphere raises the height of the tropopause. A rise in tropopause height in response to CO2 increase has been seen in both observations and climate model simulations [e.g., Kushner et al., 2001; Santer et al., 2003].

[3] In addition to a rise in the tropopause height, climate models simulate increases in upper-level zonal wind and eddy kinetic energy, which appear to be associated with the rise in tropopause [Kushner et al., 2001; Raisanen, 2003]. The increase in zonal-mean zonal wind occurs in the lower stratosphere as a consequence of thermal wind balance in the presence of stratospheric cooling and tropospheric warming. These temperature changes enhance the meridional temperature gradient at the tropopause level because the tropopause slopes downward toward the poles. The increase in kinetic energy also occurs at the tropopause level, and is consistent with the rise in tropopause height because synoptic waves are trapped in the troposphere. A higher tropopause thus means a higher level favorable for synoptic eddy activity. In addition to the above changes, climate models also predict changes in the zonal wind and eddy kinetic energy in the troposphere. Kushner et al. [2001] find that the Southern Hemisphere midlatitude jet and eddies shift poleward in response to global warming. Yin [2005] finds similar changes in the band-pass filtered storm tracks in all seasons and in both Northern and Southern Hemisphere.

[4] The purpose of this paper is to suggest that changes in tropopause height might be responsible for the poleward shifts in the tropospheric jets and storm tracks, in addition to the changes at and above the tropopause. First we will show the ensemble-mean response to CO2 increase in the models participating in the Intergovernmental Panel on Climate Change (IPCC) fourth assessment report (section 3.1). We then use a simple, dry general circulation model (GCM) model forced as in Held and Suarez [1994] to determine the response to a raised tropopause in a relatively simple setting (section 3.2). These simple GCM experiments are basically the same as the experiments in recent work by Haigh et al. [2005] and Williams [2006]. The changes in the simple GCM are remarkably similar in magnitude and structure to the IPCC models, suggesting that the changes in large-scale winds and storm tracks in the midlatitudes are forced by changes in tropopause height rather than, say, changes in moisture content or low-level baroclinicity. Experiments in which tropopause height is selectively raised poleward or equatorward of the midlatitude jet are presented in section 3.3. We also show that the spread of the modeled changes in polar tropopause height about the ensemble mean explains a significant amount of the inter-model variability in the zonal wind response. A summary and discussion follows in section 4.

2. Models and Data

[5] We use output from climate change scenario integrations prepared for the IPCC Fourth Assessment Report. Future climate data used here comes primarily from the A2 scenario, a “business as usual” scenario in which carbon dioxide (CO2) concentrations rise unchecked and reach 856 parts per million (ppm) by the year 2100. Simulations of present-day climate from the same models were obtained from the Coupled Model Intercomparison Project's “20th Century Climate in Coupled Models” (20C3M) data archive. Output from the 15 models listed in Table 1 was available for both the 20C3M and A2 scenario archives for the variables of interest here. Additional data from the B2 scenario, in which CO2 concentrations are controlled and stabilized at 549 ppm by 2100, were used in Figure 13. With the exception of the daily-mean data used for the fluxes in Figure 7, all data used here is archived as monthly averages. Complete details on the forcing used in the A2 and B1 scenarios are given in Appendix II of the 2001 IPCC report [IPCC, 2001]. In the results presented here, climate change is defined as the difference between the climatology's of years 2080 to 2099 in the A2 scenario and years 1980 to 1999 in the 20C3M simulations. One realization was used from each model.

Table 1. IPCC Models Used in This Study
ModelScenario
BCCR-BCM2.020c3m, A2
CCSM320c3m, A2
CGCM3.1(T47)20c3m, A2, B1
CNRM-CM320c3m, A2, B1
CSIRO-Mk3.020c3m, A2, B1
ECHAM5/MPI-OM20c3m, A2
GFDL-CM2.020c3m, A2, B1
GFDL-CM2.120c3m, A2
GISS-ER20c3m, A2, B1
INM-CM3.020c3m, A2
IPSL-CM420c3m, A2
MIROC3.2(medres)20c3m, A2, B1
MRI-CGCM2.3.220c3m, A2, B1
PCM20c3m, A2
UKMO-HadCM320c3m, A2

[6] The simple GCM is the grid-point model described by Held and Suarez [1994]. In this model, the circulation is driven by Newtonian relaxation to a prescribed equatorially symmetric “radiative equilibrium” temperature profile, with additional thermal and momentum damping in the lowest model levels. The external forcing and boundary conditions are zonally symmetric in this GCM. The horizontal resolution is 5 × 4 degrees in longitude and latitude, with 20 equally spaced sigma levels in the vertical. The GCM runs were repeated at twice the horizontal and vertical resolution and similar results were obtained. The model is integrated on an Arakawa C-Grid using an explicit leapfrog scheme with a 7.2-minute time step. Details of the model's dynamical core of are given by Suarez and Takacs [1996], and the model forcing is described by Held and Suarez [1994]. Each model experiment shown here is integrated for eight years after a 300-day spin-up period.

3. Results

3.1. Climate Change in IPCC Models

[7] The ensemble-mean temperature response to climate change is a warming in troposphere and a cooling in stratosphere (Figure 1). The tropospheric warming is largest in the tropical troposphere because moist convection sets the lapse rate in the tropics to be moist adiabatic [Held, 1993]. In a warmer world, the moist adiabatic lapse rate is smaller, so the temperature different between the surface and upper troposphere is smaller.

Figure 1.

Annual average ensemble-mean change in zonal-mean temperature (Kelvin). The horizontal axis is latitude and the vertical axis is pressure in hPa.

[8] Figure 2a shows the 1980 – 1999 ensemble-mean lapse rate for the 20C3M simulations. The location of the tropopause is easily discernable as the abrupt change from a large tropospheric value to about zero in the stratosphere (negative in the tropical stratosphere). The change in lapse rate is largest at the tropopause, where an increase in lapse rates means an upward extension of the troposphere, or an increase in tropopause height.

Figure 2.

(a) Annual average ensemble-mean zonal-mean lapse rate for the 20th century. (b) Ensemble-mean change in zonal-mean lapse rate. (c) The vertical derivative of the 20th century lapse rate multiplied by 400 m.

[9] One way to see that the change in lapse rate represents an increase in tropopause height is as follows [Kushner et al., 2001]: let the twentieth century lapse rate be Γ(z). A rise in tropopause height corresponds to an upward shift in the lapse rate profile near the tropopause. Thus, the new lapse rate at z is Γ(zδz) where δz is the change in tropopause height. The lapse rate difference between the future scenario and the twentieth century simulation is given approximately by

equation image

An upward shift in the lapse rate should thus have the same structure as the vertical derivative of the twentieth century lapse rate. Figure 2b shows the difference between the ensemble-mean lapse rate for years 2080 – 2099 of the A2 scenario and the 20C3M lapse rate in Figure 2a, while Figure 2c shows the vertical derivative of the 20C3M lapse rate multiplied by 400 m. Comparison of Figures 2b and 2c shows that the vertical derivative of the climatological lapse rate multiplied by 400 m corresponds well to the modeled change in lapse rate at the tropopause, especially in the midlatitudes. The change in lapse rate is therefore consistent with a 400 m rise in the tropopause. (Seidel and Randel [2006] estimate a 64 m change per decade in radiosonde data, and Santer et al. [2003] find a 120 m change in a GCM and a 190 m change in the NCEP reanalysis since 1979.) The largest discrepancy between lapse rate change and lapse rate vertical derivative occur in the tropics, where greenhouse warming implies a reduction of the moist adiabatic lapse rate, an effect not captured in (1).

[10] Figure 3 shows the ensemble-mean changes in zonal-mean zonal wind between the A2 and 20C3M simulations for each season, superimposed on the 20C3M climatological winds. The zonal-mean zonal wind changes occur in all seasons, independent of the presence or absence of the polar night jet. In addition to the stratospheric wind changes, which are expected from the thermal wind argument given in section 1, one sees positive wind anomalies in the troposphere centered slightly poleward of the climatological jet. In addition, there are smaller negative wind anomalies on the equatorward side of the jet.

Figure 3.

Ensemble-mean zonal-mean climatological zonal wind for the 20th century (shaded) and the ensemble-mean zonal-mean change in zonal wind (contours). (a) Winter. (b) Spring. (c) Summer. (d) Fall.

[11] Looking at the horizontal map of the zonal wind change (Figure 4), one sees that in the Northern Hemisphere (NH) the zonal wind changes are strongest in the oceanic storm tracks. Except over the Atlantic Ocean in winter and the Pacific Ocean in summer, the zonal wind change is a dipole pattern with the zero wind line at the location of the climatological jet. This zonal wind change corresponds to a poleward shift of the jet. In the NH Atlantic in winter and the Southern Hemisphere (SH) the nodal line of the zonal wind dipole is displaced slightly equatorward of the climatological jet and thus corresponds to both a poleward shift and a strengthening of the jet.

Figure 4.

Ensemble-mean climatological zonal wind at 850 hPa for the 20th century (shaded) and the ensemble-mean change in zonal wind at 850 hPa (contour). (a) Winter. (b) Spring. (c) Summer. (d) Fall.

[12] Is the change in the stratosphere winds seen in Figure 3 responsible for the tropospheric wind changes? Previous studies have shown that stratospheric wind anomalies can have an important effect on the tropospheric circulation [Baldwin and Dunkerton, 1999; Thompson et al., 2005; Scaife et al., 2005]. These studies focus on the cold season, however, while the tropospheric changes found here occur in all seasons. Moreover, the above studies focus on the variability in the strength of the polar night jet, while the modeled stratospheric changes are not collocated with the polar night jet in winter and thus do not represent variability in the strength of the jet. Thus, if the tropospheric changes are due to changes in the upper atmosphere, then the reasons must be more general than the polar night jet/troposphere connection. In the next section, we argue that the changes in the tropopause height are responsible from the tropospheric wind changes.

3.2. Simple GCM

[13] To help understand the tropospheric jet response to climate change, we use a simple GCM forced by Newtonian relaxation of temperature to a prescribed “equilibrium” temperature [Held and Suarez, 1994]. The equilibrium temperature is warm at the equator and cool at the poles and decreases with altitude until the temperature of 200 K (Figure 5a). Above this altitude, the equilibrium temperature is the constant value of 200 K. The equilibrium temperature of this GCM acts to force a tropopause at the level where the imposed lapse rate changes from positive to zero (Figure 5b). In addition to a control run with Held and Suarez [1994] forcing, we also perform a run where the equilibrium temperature decreases with altitude until the temperature reaches 196.5 K. In this second run the equilibrium lapse rate remains at a tropospheric value at higher altitudes than in the control. In this experiment only the stratospheric “equilibrium temperature” is changed and the effect is similar to increasing greenhouse gas loading in the middle atmosphere alone [Sigmond et al., 2004]. This second GCM run simulates the response to an increase in tropopause height (Figure 5b) and is designed to isolate the effect of the changes in tropopause height from everything else. Recently, Haigh et al. [2005] and Williams [2006] performed identical experiments for a range of stratospheric equilibrium temperatures.

Figure 5.

(a) The “equilibrium” temperature for the control run of the simple GCM (Kelvin). (b) Vertical profile of the “equilibrium” temperature of the control (dotted) and the raised tropopause (solid) at 46 degrees latitude.

[14] The difference in lapse rate between the raised tropopause and the control for the simple GCM is similar to the IPCC models, especially at the extratropical tropopause (Figure 6). The differences between the IPCC models and the simple GCM are in the tropical stratosphere and the troposphere. These are the same regions where the greenhouse gas-induced change in lapse rate is not attributable to a simple upward shift in the vertical lapse rate profile (see section 3.1).

Figure 6.

Annual average change in the zonal-mean lapse rate for (a) the IPCC models and (b) the simple GCM. Annual-average climatological zonal-mean zonal wind (shaded) and the change in the zonal-mean zonal wind for (c) the IPCC models and (d) the simple GCM.

[15] The zonal wind changes that accompany the changes in the tropopause height in the simple GCM are qualitatively similar to the IPCC model changes even though the forcing in the GCM is very idealized. Both plots show positive tropospheric anomalies on the poleward flank of the jet and smaller negative anomalies on the equatorward side of the jet (Figure 6). These changes in the zonal wind correspond to a strengthening and shifting of the midlatitude jet in both the IPCC models and the simple GCM. The magnitude of the zonal wind change as a percent of the daily standard deviation in zonal-mean zonal wind is: 80% for the simple GCM, 100% for the Southern Hemisphere (IPCC) and 15% in the Northern Hemisphere. At first glance, the zonal wind changes in the Northern Hemisphere troposphere seem different until one notices that the climatological jet maximum in the NH tilts equatorward with height. This tilt in the climatological winds means that wind changes corresponding to a jet shift must also tilt equatorward with height. This tilt of the zonal wind anomalies is also evident in SH winter (Figure 3c). This strengthening and poleward shift of the tropospheric jets is a robust response to increases in tropopause height [Williams, 2006].

[16] In addition to the zonal-mean zonal wind changes, we also look at the changes in the submonthly transient statistics in the IPCC models with available daily-mean data. We show the results of one model (GFDL) because it has daily-mean data from 1000 hPa to 10 hPa. The other models only have data up to 200 hPa. We believe the results of the GFDL model are robust because the other models agree at levels where they overlap. Also, because the simple GCM has no imposed zonal asymmetries, we compare the transient statistics (i.e., products of departures from the time-mean, including zonal-mean perturbations) of the IPCC models to the eddy statistics (products of departures from the zonal-mean, including any time-mean zonal asymmetries) of the simple GCM.

[17] In Figure 7a, the positive changes in the transient kinetic energy are above and to the poleward side of the climatological transient kinetic energy. The negative changes in the transient kinetic energy, on the other hand, are smaller in size and located below and to the equatorward side of the climatological kinetic energy. A similar change in the eddy kinetic energy is also seen in the simple model in response to a raised tropopause (Figure 7b). The key point is the similarity in the relationship between the climatological kinetic energy and the change in kinetic energy for both the IPCC models and the simple GCM. The pattern of the observed changes suggests a strengthening and a poleward and upward displacement of the climatological transient kinetic energy. A very similar change is also noted by Kushner et al. [2001] and Yin [2005].

Figure 7.

(a) Annual-average climatological (shaded) and change in (contour) submonthly transient kinetic energy for the GFDL model. (b) Annual-average climatological (shaded) and change in (contour) eddy kinetic energy for the simple GCM. (c) Annual-average climatological (shaded) and change in (contour) submonthly transient momentum flux for the GFDL model. (d) Annual-average climatological (shaded) and change in (contour) eddy momentum flux for the simple GCM. (e) Annual-average climatological (shaded) and change in (contour) submonthly transient heat flux for the GFDL model. (f) Annual-average climatological (shaded) and change in (contour) eddy heat flux for the simple GCM.

[18] For the transient momentum flux, one also sees a pattern that suggests a strengthening and a poleward and upward shift of the climatology in both the GFDL model and the simple GCM. Both models show a large change, which is the same sign as the climatology, on the upward and poleward side of the climatology and a weaker anomaly of the opposite sign on the downward and equatorward side (the contour interval for the simple GCM plot is to large to show this).

[19] For the heat fluxes (Tv′), one sees large increases in the stratosphere for both models. In the troposphere, however, we see significant differences between the GFDL model and the simple GCM. The GFDL model (as well as the other models with archived daily data) shows a weak dipole pattern above 700 hPa in the troposphere while the simple GCM has a stronger monopole on the poleward side of the climatology at most pressure levels except at the tropopause there is a weak anomaly of the opposite sign equatorward of the climatology. Thus, on the equatorward flank of the jet at low levels the heat flux is weakening in the IPCC models but it remains basically constant in the simple GCM.

3.3. Jet Sensitivity to Localized Tropopause Changes

[20] We are currently trying to understand the dynamical reasons for the jet shift in response to an increase in tropopause height. In some preliminary work, we have performed experiments with the simple GCM similar to Haigh et al. [2005], which show that the changes in the tropopause height poleward of the jet are key to the poleward shift of the jet. When the stratospheric “equilibrium” temperature is decreased poleward of 45° (i.e., the tropopause is raised poleward of 45°) the zonal wind response in the troposphere is a much larger poleward shift than the response to a uniform decrease in stratospheric “equilibrium” temperature (Figure 8b). In addition, a decrease in the stratospheric temperature equatorward of 45° causes the jet to make a strong shift equatorward (Figure 8c). These results suggest that the correct way to think about the zonal wind response may be in terms of the meridional temperature gradient rather than the tropopause height: Increases in the upper-level temperature gradient lead to a strengthening and poleward shift of the jet and vice-versa. This theory is also consistent with the uniform tropopause height increase experiment having the smallest zonal wind response because this experiment also has the smallest imposed change in upper-level meridional temperature gradient.

Figure 8.

The change in zonal-mean lapse rate (shaded) and zonal wind (contour) in the simple GCM relative to the control for (a) the tropopause raised everywhere, (b) the tropopause raised poleward of the jet, and (c) the tropopause raised equatorward of the jet.

[21] Additional GCM experiments, however, suggest that this theory might not be entirely correct and that tropopause height changes alone can cause changes in tropospheric winds. In these GCM experiments, we apply a localized heat source at all locations in the latitude/pressure plane. We do this by adding a 5 K increase to the control Held and Suarez [1994] equilibrium temperature over a 20° latitude by 150 hPa localized (i.e., step-function) region of the latitude/pressure plane. If we consider two of these applied heating experiments, one with the heating above the tropopause and another with the heating below the tropopause, then we might be able to determine the correct paradigm to view the tropospheric wind response. In the “lapse rate paradigm” the heating below the tropopause will lead to an increase in the tropopause height while the heating above the tropopause will lead to a decrease in the tropopause height. Thus in the lapse rate paradigm the response in the zonal wind to the two heat sources should be opposite. In the meridional temperature gradient paradigm, both heat sources will change the meridional temperature gradient near the tropopause in the same sense. Thus in the temperature gradient paradigm, the zonal wind response to the two heat sources should be the same.

[22] To summarize the results of these experiments, we project the wind response on the leading EOF of the control run (Figure 9a, this pattern explains over 80% of the variance in the zonal wind response for these experiments). Figure 9b shows the response at the latitude of the jet (45°N) as a function of the pressure level of the imposed heating. The units are the maximum EOF1 wind anomalies associated with the zonal wind response in m/s. There is a dramatic change in the sign of the response as the heating crosses the tropopause suggesting that the change in the tropopause height is an important part of the response. Heating above the tropopause lowers the tropopause and causes the jet to shift equatorward (negative EOF1) while heating below the tropopause raises the tropopause and leads to a poleward shift of the jet (positive EOF1).

Figure 9.

Zonal-mean zonal wind response (relative to a control run) of GCM experiments with applied, localized heating. (a) EOF1 of the control run. (b) The zonal wind response projected on EOF1 of the control run as a function of the pressure level of the applied heating. The heating is applied from 35°N to 55°N and is 150 hPa deep. The units are the maximum EOF1 response in m/s (i.e., the amplitude of zonal wind response at the centers of action of EOF1). (c) The zonal wind response projected on EOF1 of the control run as a function of both latitude and pressure. The applied heating is 20° latitude wide and 150 hPa deep. The tropopause is plotted with a thick dotted line.

[23] Figure 9c shows the response projected on EOF1 for heating applied at all points in the latitude/pressure plane. The tropopause is marked by the thick dotted line. One can see that from 35°N to 55°N, the response changes sign as the heating crosses the tropopause. There are also changes in the response as the heating moves meridionally, particularly in the stratosphere and near the surface implying that meridional temperature gradients also play an important role. The most dramatic feature of Figure 9b, however, is the change in the sign of the response as the heating crosses the tropopause either vertically or meridionally suggesting that there is truth in both the lapse rate and temperature gradient paradigms, but that a complete understanding must consider both.

3.4. Analysis of Ensemble Spread

[24] In this section, we use a linear regression analysis to look at the spread of the 15 models about the ensemble mean. Let xi be the change in zonal-mean temperature at 200 hPa and over the polar cap (poleward of 70 degrees latitude), where i is an index over each of the 15 climate models and each of the months in the “season” of the regression analysis. The choice of temperature over the polar cap for our index, xi, is motivated by the results in the previous section, which show that the tropopause height changes poleward of the jet are key to the zonal wind response. We then regress the change in temperature at each latitude and level on this xi index:

equation image

where n and m are the number of models and months, respectively, and σ stands for the standard deviation. The equation image and equation image have values for each individual month so that the calculation analyzes model-to-model differences rather than season-to-season differences in the response. The same formula is used for a Δu regression calculation.

[25] Figure 10 shows the result of this analysis for the SH (year-round) as well as the correlation between the change in temperature over the polar cap at 200 hPa and Δu and ΔT. The regressions and correlations are multiplied by negative one so that the results correspond to a higher tropopause. The temperature regression plot shows cold temperature anomalies in the lower stratosphere, which rapidly go to zero as one crosses the tropopause. Thus these cool temperature anomalies correspond to a higher tropopause over the polar cap. The zonal wind regression shows that models with a higher tropopause over the polar cap also tend to have stronger winds at 55°S and weaker winds at 35°S. These anomalies correspond to a strengthening and poleward shift of the climatological winds. These results thus support the idea that the tropopause height causes the tropospheric wind response because models with larger increases in tropopause height have larger changes in the zonal wind. However, the results could also be interpreted that the zonal wind changes in the midlatitudes cause the changes in tropopause height in the polar regions. The correlation between the polar cap temperature at 200 hPa and the zonal wind change at 850 hPa is over 0.6, which is much greater than the statistically significant value of 0.3 (95% confidence level). A scatterplot of the low-level zonal wind response and the 200 hPa polar cap temperature response shows that the high correlation is real and is not due to a few large outliers (Figure 11).

Figure 10.

Regressions and correlations over all models on an index of temperature at 200 hPa over the polar cap (SH). The index is negative one times the change in temperature at 200 hPa poleward of 70 degrees latitude. (a) Regression of the change in zonal-mean temperature on the above index. (b) Regression of the change in zonal-mean zonal wind on the above index. (c) Correlation between the change in zonal-mean temperature and the above index. (d) Correlation between the change in zonal-mean zonal wind and the above index.

Figure 11.

Scatterplot of zonal-mean temperature change at 200 hPa poleward of 70 S versus zonal-mean zonal wind change at 55 S and low-levels (vertical average from 1000 hPa to 500 hPa). There is one point for each month and model (=12 · 15 = 180 points).

[26] The regression analysis for the Northern Hemisphere winter shows similar looking zonal wind patterns, however, the correlations barely pass the statistical significance level (not shown). The NH summer has even smaller correlations that do not pass the statistical significance level (not shown). Perhaps the correlations are weaker in the NH because the jets and storm tracks are localized in longitude. We tested this idea by looking at zonal averages over sectors over the Atlantic and Pacific Oceans. We found that the correlations improved significantly only in the Atlantic during winter: the correlation between the polar cap temperature decrease and the zonal winds increase improved from 0.3 to 0.45.

[27] Figure 12 shows the 850 hPa zonal-mean Δu for each model about the latitude of the climatological jet position (= 0 on the horizontal axis). The two thick solid lines are the ensemble-mean response plus or minus one standard deviation. For the SH summer all models except one show a dipole structure with positive anomalies poleward of the mean jet latitude and negative anomalies equatorward of the mean jet latitude. The one model that has the opposite sign wind response also has a tropopause height that lowers in response to CO2 increase (see also Figure 11). (Perhaps the stratospheric ozone levels are higher at the end of the 21st century than in the 20th century in this model. The 20C3M protocol does not specify the inclusion or exclusion of the Antarctic ozone hole.) For the SH winter, almost all models show a dipole wind response with positive anomalies poleward of the jet. The relationship of the u change to the time-mean jet, however, is slightly different than in SH summer. In SH winter the zero u change line tends to be more equatorward of the time-mean jet compared to SH summer. Thus, for SH winter, the Δu represents a greater amount of strengthening compared to the SH summer. For the NH summer and winter the spread of the model response is greater than the SH. Nevertheless, very few models have positive u changes equatorward of the time mean jet and negative u changes poleward of the jet.

Figure 12.

The thin lines are the change in zonal-mean zonal wind for each of the 15 IPCC models and the two thick lines are the ensemble-mean response plus or minus the standard deviation. The horizontal axis is the latitude relative to the climatological jet in each particular model and the positive (negative) values on the horizontal axis are degrees latitude poleward (equatorward) from the climatological jet.

4. Summary and Discussion

[28] The IPCC models predict a strengthening and a poleward shift of the tropospheric zonal jets in response to increased CO2 concentration. The change in zonal jets is also accompanied by a strengthening and a poleward and upward shift of transient kinetic energy and momentum flux. These changes in climate under global warming are simulated by a simple GCM when the height of the tropopause is raised. This similarity between the simple GCM and the IPCC models suggests that the changes in midlatitude circulation are predominantly driven by a rise in the height of the tropopause and that changes in moisture content and the low-level meridional temperature gradient, for example, play a secondary role.

[29] In addition, the spread in the model zonal wind response in the Southern Hemisphere is explained well by the spread in the model tropopause height response over the polar cap. The correlation between the zonal wind response and the tropopause height response, however, is significantly weaker in the Northern Hemisphere.

[30] Interestingly, Cai et al. [2003] found that the response of the Southern annular mode is present only in the transient response to CO2 increase and that as the CO2 concentration is stabilized the annular mode change disappears. While the IPCC simulations of the A2 forcing scenario end at 2100, the B1 scenario continues at a stabilized CO2 concentration for two hundred years until 2300. In contrast to Cai et al. [2003], we find that the stabilized u response at 2300 is essentially the same as the u response at 2100 (Figure 13). Our result is consistent with the idea that the tropopause height determines the u response because the tropopause should remain high under CO2 stabilization. The reasons for the difference between this work and Cai et al. [2003] are not clear at present.

Figure 13.

The annual average ensemble-mean zonal-mean change in the 850 hPa zonal wind relative to 1980–1999 for 2080–2099 (solid) and for 2280–2299 (dotted). The calculation is done for the 7 models in the B1 forcing scenario with data out to 2299.

[31] In addition, we are currently looking at simple model experiments to understand the dynamics of the strengthening and poleward shift of the midlatitude jet in response to the rise in tropopause height. The dynamical mechanisms responsible for the jet response to global warming may also be implicated in the midwinter suppression in the Pacific storm track (J. Yin, personal communication) and also the effect of polar night jet variability on the tropospheric annular mode as seen by Baldwin and Dunkerton [1999] and Thompson et al. [2005].

Acknowledgments

[32] This research was supported by the NOAA Postdoctoral Program in Climate and Global Change, administered by the University Corporation for Atmospheric Research. Support for E. DeWeaver was provided by the Office of Science (BER), U.S. Department of Energy, under grant DE-FG02-03ER63604. We thank Max Suarez for providing us with the simple GCM used in this study. In addition, we acknowledge the international modeling groups for providing their data for analysis, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for collecting and archiving the model data, the JSC/CLIVAR Working Group on Coupled Modeling (WGCM) and their Coupled Model Intercomparison Project (CMIP) and Climate Simulation Panel for organizing the model data analysis activity, and the IPCC WG1 TSU for technical support. The IPCC Data Archive at Lawrence Livermore National Laboratory is supported by the Office of Science, U.S. Department of Energy.

Ancillary