• Ambaum, M. H. P., , and B. J. Hoskins, 2002: The NAO troposphere–stratosphere connection. J. Climate, 15 , 19691978.

  • Boer, G. J., , G. M. Flato, , and D. Ramsden, 2000: A transient climate change simulation with historical and projected greenhouse gas and aerosol forcing: Projected climate for the 21st century. Climate Dyn., 16 , 427450.

    • Search Google Scholar
    • Export Citation
  • Chen, P., , and W. A. Robinson, 1992: Propagation of planetary waves between the troposphere and stratosphere. J. Atmos. Sci., 49 , 25332545.

    • Search Google Scholar
    • Export Citation
  • Flato, G. M., , G. J. Boer, , W. G. Lee, , N. A. McFarlane, , D. Ramsden, , M. C. Reader, , and A. J. Weaver, 2000: The Canadian Centre for Climate Modelling and Analysis global coupled model and its climate. Climate Dyn., 16 , 451467.

    • Search Google Scholar
    • Export Citation
  • Fyfe, J. C., , G. J. Boer, , and G. M. Flato, 1999: The Arctic and Antarctic oscillations and their projected changes under global warming. Geophys. Res. Lett., 26 , 16011604.

    • Search Google Scholar
    • Export Citation
  • Gillett, P. G., , M. R. Allen, , and K. D. Williams, 2002: The role of stratospheric resolution in simulating the Arctic Oscillation response to greenhouse gases. Geophys. Res. Lett.,29 (10), 1500, doi:10.1029/2001GL014444.

    • Search Google Scholar
    • Export Citation
  • Houghton, J. T., , Y. Ding, , D. J. Griggs, , M. Noguer, , P. J. van der Linden, , and D. Xiaosu, Eds.,. 2001: Climate Change 2001: Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, 944 pp.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors. 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Latif, M., 2000: On the interpretation of climate change in the tropical Pacific. MPI Rep. 306, Max-Planck-Institut für Meteorologie, Hamburg, Germany, 20 pp.

    • Search Google Scholar
    • Export Citation
  • Limpasuvan, V., , and D. L. Hartmann, 1999: Eddies and the annular modes. Geophys. Res. Lett., 26 , 31333136.

  • Marquardt, D., 1963: An algorithm for least squares estimation of nonlinear parameters. SIAM J. Appl. Math., 11 , 431441.

  • McFarlane, N. A., , G. J. Boer, , J-P. Blanchet, , and M. Lazare, 1992: The Canadian Climate Centre second generation general circulation model and its equilibrium climate. J. Climate, 5 , 10131044.

    • Search Google Scholar
    • Export Citation
  • Nigam, S., 1990: On the structure of variability of the observed tropospheric and stratospheric zonal-mean zonal wind. J. Atmos. Sci., 47 , 17991813.

    • Search Google Scholar
    • Export Citation
  • Pacanowski, R. C., , K. Dixon, , and A. Rosati, 1993: The GFDL modular ocean model users guide. GFDL Ocean Group Tech. Rep. 2, Geophysical Fluid Dynamics Laboratory, Princeton, NJ, 46 pp.

    • Search Google Scholar
    • Export Citation
  • Polvani, L. M., , and P. J. Kushner, 2002: Tropospheric response to stratospheric perturbations in a relatively simple general circulation model. Geophys. Res. Lett.,29 (7), 1114, doi:10.1029/2001GL014284.

    • Search Google Scholar
    • Export Citation
  • Scinocca, J. F., , and P. H. Haynes, 1998: Dynamical forcing of stratospheric waves by tropospheric baroclinic waves. J. Atmos. Sci., 55 , 23612392.

    • Search Google Scholar
    • Export Citation
  • Shindell, D. T., , R. L. Miller, , G. A. Schmidt, , and L. Pandolfo, 1999: Simulation of recent Northern Hemisphere climate trends by greenhouse-gas forcing. Nature, 399 , 452455.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W., , and J. M. Wallace, 2000: Annular modes in the extratropical circulation. Part I: Month-to-month variability. J. Climate, 13 , 10001016.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., , D. P. Stepaniak, , and J. W. Hurrell, 2001: Quality of reanalyses in the Tropics. J. Climate, 14 , 14991510.

  • von Storch, H., , and F. W. Zwiers, 1999: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.

  • Zhang, Y., , J. M. Wallace, , and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900–93. J. Climate, 10 , 10041020.

  • View in gallery

    (left) Time mean and (right) leading PCA modes for the constructed latitudinal jet u(θ,t) = A exp[−(θ − Θ)2/2W2] with fixed width W but variable amplitude A(t) and position Θ(t). (a) A(t) and Θ(t) fluctuating pseudorandomly about constant values. (b) A(t) and Θ(t) fluctuating as in (a) but with the latter also monotonically increasing (i.e., for a poleward drifting jet). Units are m s–1

  • View in gallery

    (top) Observed (NCEP) and (bottom) control simulation climatological zonal mean zonal wind with 4 m s–1 contours (… ,−6, −2, 2, …). Thick contours are negative. Time mean jet axes are indicated by the solid dots

  • View in gallery

    Observed NH variability modes. (top) Standardized principal components. (bottom) Zonal mean zonal wind regression patterns with 0.2 m s–1 contours (… , −0.3, −0.1, 0.1, …). Thick contours are negative. Variances explained in upper left. Hatching indicates significance at the 1% significance level. Time mean jet axes are indicated by the solid dots

  • View in gallery

    Observed NH parameter regression curves. Solid black curves are fast mode 1; gray curves are fast mode 2; dashed curves are slow mode 1

  • View in gallery

    Observed NH MSLP regression patterns with 0.3 hPa (… , −0.45, −0.15, 0.15, …) contours. Thick contours are negative. Outer lat circle is 20°. Inner lat circle is 60°

  • View in gallery

    Same as Fig. 3, but for control simulation NH variability modes

  • View in gallery

    NH slowly varying parameters from an ensemble of climate change simulations. Here aS1 is in units of m;ths–1 and aS2 and aS3 in degrees

  • View in gallery

    Same as Fig. 3, but for climate change simulation NH variability modes

  • View in gallery

    Std devs (filled circles) of the NH fast mode PCs for contiguous 21-yr periods (beginning in 1916) and for each of the climate change simulations. The open circles are for the observed fast mode PCs. The vertical lines represent confidence intervals at the 10% significance level. Run 1 corresponds to the run presented in Fig. 8

  • View in gallery

    Same as Fig. 3, but for observed SH variability modes

  • View in gallery

    Same as Fig. 4, but for observed SH parameter regression curves

  • View in gallery

    Same as Fig. 3, but for control simulation SH variability modes

  • View in gallery

    Same as Fig. 5, but for control simulation SH MSLP regression patterns. Outer lat circle is −20°. Inner lat circle is −60°.

  • View in gallery

    Same as Fig. 7, but for SH slowly varying parameters

  • View in gallery

    Same as Fig. 3, but for climate change simulation SH variability modes

  • View in gallery

    Same as Fig. 9, but for the SH fast modes

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 114 114 0
PDF Downloads 1 1 0

Separating Extratropical Zonal Wind Variability and Mean Change

View More View Less
  • 1 Canadian Centre for Climate Modelling and Analysis, Meteorological Service of Canada, Victoria, Canada
© Get Permissions
Full access

Abstract

Changes in the naturally occurring modes of extratropical annual mean and zonal mean zonal wind variability are investigated using National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalyses and Canadian Centre for Climate Modelling and Analysis (CCCma) global climate model simulations. In the Northern Hemisphere, the first and second modes are primarily stratospheric and tropospheric in character, respectively. The surface pressure manifestations of these modes are intimately linked to the Arctic Oscillation (AO), and together suggest separate stratospheric and tropospheric origins for the AO. In the Southern Hemisphere, the first mode describes north–south shifts in the polar front jet accompanied by polar stratospheric jet fluctuations and Antarctic Oscillation (AAO)-like surface pressure anomalies. The second mode is primarily tropospheric and describes interannual changes in the strength and position of the polar front jet.

The leading observed modes appear unchanged in strength since the 1950s except in the Northern Hemisphere where the second mode shows some evidence of increasing strength. The leading simulated modes appear unchanged in strength since the beginning of the twentieth century, and are predicted to remain so to the end of the twenty-first century. In all cases the leading modes are superimposed upon significant mean change, which when not properly accounted for can lead to erroneous conclusions.

Corresponding author address: Dr. J. C. Fyfe, Canadian Centre for Climate Modelling and Analysis, University of Victoria, P.O. Box 1700, Victoria, BC V8W 2Y2, Canada. Email: John.Fyfe@ec.gc.ca

Abstract

Changes in the naturally occurring modes of extratropical annual mean and zonal mean zonal wind variability are investigated using National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalyses and Canadian Centre for Climate Modelling and Analysis (CCCma) global climate model simulations. In the Northern Hemisphere, the first and second modes are primarily stratospheric and tropospheric in character, respectively. The surface pressure manifestations of these modes are intimately linked to the Arctic Oscillation (AO), and together suggest separate stratospheric and tropospheric origins for the AO. In the Southern Hemisphere, the first mode describes north–south shifts in the polar front jet accompanied by polar stratospheric jet fluctuations and Antarctic Oscillation (AAO)-like surface pressure anomalies. The second mode is primarily tropospheric and describes interannual changes in the strength and position of the polar front jet.

The leading observed modes appear unchanged in strength since the 1950s except in the Northern Hemisphere where the second mode shows some evidence of increasing strength. The leading simulated modes appear unchanged in strength since the beginning of the twentieth century, and are predicted to remain so to the end of the twenty-first century. In all cases the leading modes are superimposed upon significant mean change, which when not properly accounted for can lead to erroneous conclusions.

Corresponding author address: Dr. J. C. Fyfe, Canadian Centre for Climate Modelling and Analysis, University of Victoria, P.O. Box 1700, Victoria, BC V8W 2Y2, Canada. Email: John.Fyfe@ec.gc.ca

1. Introduction

Changes in naturally occurring modes of variability such as the El Niño–Southern Oscillation (ENSO), the Arctic Oscillation (AO), and the Antarctic Oscillation (AAO) have received increasing attention in recent years. Understanding these changes is made difficult by interactions between changes in the mean, or base state, and the variability modes (Houghton et al. 2001). This can be illustrated with a constructed latitudinal jet u(θ, t) = A exp [−(θ − Θ)2/2W2] with fixed width W but variable amplitude A(t) and position Θ(t). In our first example, A(t) and Θ(t) fluctuate pseudorandomly about constant values. Figure 1a shows the time mean (left) and leading variability modes (right) from a principal component analysis (PCA). In this example, the mean and variability structures are straightforward and easy to interpret. As a contrasting example, we take A(t) and Θ(t) fluctuating as before but with the latter also monotonically increasing (i.e., for a poleward drifting jet). Figure 1b shows that since u(θ, t) is nonstationary the PCA seriously confounds the poleward drift with the fast variations identified in Fig. 1a. Nonstationarity is an important open problem in time series analysis but one that is often sidestepped in oceanographic and meteorological analyses. In studies of tropical Pacific sea surface temperature (SST), for example, it is important to distinguish between base-state changes and interannual variability changes associated with ENSO (Latif 2000). Similarly, in studies of hemispheric sea level pressure (SLP) it is important to be aware that base-state change projects across the spectrum of variability modes, including the AO and AAO (Fyfe et al. 1999). With these considerations in mind we turn our attention in this study to annual mean and zonal mean zonal wind u(θ,p,t). Specifically, we ask if the leading variability modes of u(θ,p,t) have changed in the past and how they might behave in the future.

Motivated by the examples above our analysis will be in three easy steps: 1) functional approximation of u(θ,p,t) using a small number of physically based parameters; 2) global (i.e., parameter by parameter) removal of any trend components; and 3) PCA to obtain the variability modes. This procedure removes variability mode trends but still allows for change in their higher-order temporal statistics. Changes in the strength (i.e., variance) of the variability modes will be tested for using standard statistical tests. In the next section we describe the methodology in more detail and then present our results and summary remarks in sections 3 and 4, respectively.

2. Datasets and methodology

This study uses data (primarily annual mean and zonal mean zonal wind) from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalyses (for the period 1958–99; Kalnay et al. 1996) and from the Canadian Centre for Climate Modelling and Analysis (CCCma) coupled GCM (CGCM1; Flato et al. 2000). The atmospheric component of the coupled model is a global primitive equation spectral model with T32 truncation and 10 unequally spaced vertical levels with the top level at 12 hPa (McFarlane et al. 1992). The ocean component is a global primitive equation gridpoint model with 1.875° resolution and 29 vertical levels (Pacanowski et al. 1993). A 201-yr control simulation and an ensemble of three independent climate change simulations forced with changing greenhouse gas (GHG) concentrations and aerosol loadings are available for the period 1900 to 2100. This model's climate, variability, and response have been documented extensively in the literature. For example, Fyfe et al. (1999) show that the model exhibits a robust and realistic AO and AAO, with positive trends predicted for the future. The model results of Shindell et al. (1999) indicate a stronger AO response with a higher upper boundary, however, the more recent results of Gillett et al. (2002) suggest otherwise. Of course, there is no a priori guarantee that different model predictions will match each other in every detail, however, one hopes for and expects significant qualitative agreement.

Our approach begins with writing uU where
i1520-0442-16-5-863-e1
and a(p, t) = (a1, a2, a3) are pressure-and time-dependent parameters representing the magnitude, position, and width of the extratropical westerlies, respectively. These parameters are obtained hemisphere by hemisphere, level by level, and year by year using a standard numerical technique (Marquardt 1963), which produces an excellent approximation (e.g., the u and U time mean spatial correlations exceeds 0.98). Next, we isolate any trend component in the parameters a(p, t). To this end, a slowly evolving base state aS is obtained as the best cubic fit in time to a. While this fitting procedure is not perfect (e.g., it can not handle “regime shifts” and is record-length dependent) it was found, a posteriori, to perform similarly to a wide array of other temporal fits, smoothing operators and filters. The variability about aS is given by aF = aaS. Finally, we obtain the principal components (PCs) of the temporal (standardized) covariance matrices of aS and aF, with the corresponding spatial patterns obtained by regressing the (standardized) PC time series against u(θ, p, t). The PCs and corresponding spatial patterns derived from aS and aF are termed slow and fast modes, respectively. Our focus is on the leading slow and fast modes.

This global approach has a number of attractive features. First, the dimensionality is much smaller than the local (spatial) approach since parameters ≪ latitudes. Second, the mean and variability structures of a jet are very naturally and directly described with this set of parameters (i.e., magnitude, position, and width). Third, for the cases considered, global detrending was found to be much more efficient than local detrending (for the same level of accuracy). To see this consider that a linearly shifting jet requires a much higher-order (nonmonotonic) function than linear to be locally detrended. Finally, we note that the globally derived variability modes shown in the next section are very similar to their locally derived counterparts (albeit not in perfect one-to-one correspondence).

3. Results

Figure 2 compares the observed and control simulation climatological zonal mean zonal wind u(θ, p). The solid dots [given by a2(p)] identify the jet structures whose variability and mean change are the subject of this study. The Northern Hemisphere (NH) model winds are somewhat too strong in the polar stratosphere and the subtropical jet is too northward. The Southern Hemisphere (SH) model winds are much too strong in the polar stratosphere and the subtropical jet is not well separated from the polar front jet. Notwithstanding these differences the model produces reasonably realistic climatological zonal mean zonal winds.

In what follows we consider each hemisphere separately. The parameters in the NH are derived using data north of 20°, and in the SH using data south of −40°. Using SH data to −20° produces the same results in all cases except the climate change simulations. In the climate change simulations (1) is problematical owing to the development of a secondary subtropical structure towards the end of the simulations—a complication which is avoided using data to −40°.

a. Northern Hemisphere

1) Observations

Figure 3 shows the leading fast and slow modes of the observed circulation. Fast mode 1 has a dominant extratropical center in the polar stratosphere, which extends with significant amplitude to the surface. Fast mode 2 consists of a midlatitude dipole with maximum amplitude in the upper troposphere (note that the stratospheric anomalies are not statistically significant in this case). Patterns very similar to these were found by Nigam (1990) using a much shorter record of northern winter data. The fast mode PCs show year-to-year fluctuations in magnitude and sign with no obvious indication that either is systematically changing over the period. In fact, applying a standard statistical test for the equality of variances (von Storch and Zwiers 1999, p. 118) to the first and second halves of the record suggests the variance is unchanged in fast mode 1 but is possibly increased in fast mode 2 (at the 10% level of significance). Slow mode 1 has a dominant center near the tropical tropopause and a midlatitude feature extending from 10 mb to the surface at about 60°. We note that the slow mode 1 PC does not change monotonically, but rather it increases till about 1992 and then decreases. Determining whether these slow changes are real or arise from a time-evolving observing system (Trenberth et al. 2001) is unclear at this time. In this regard it was noted by an anonymous reviewer that the slow mode 1 pattern is similar to that associated with the slow “ENSO-like” tropical Pacific SST variations described in Zhang et al. (1997).

For more physical insight into the fast and slow modes we regress their PCs against the parameters themselves. Figure 4 shows that fast mode 1 (solid black curve) associates strong polar stratospheric winds with weak midlatitude upper-tropospheric winds. It is well known that the vertical propagation of wave activity into the stratosphere is very sensitive to changes in the zonal mean zonal wind, especially the vertical shear, around the tropopause (e.g., Chen and Robinson 1992). With fast mode 1 we assume that the reduced vertical shear around the midlatitude tropopause causes vertically propagating waves to be deflected equatorward. Consequently there would be less wave breaking in the stratospheric jet, which therefore becomes stronger. Fast mode 2 (gray curve) describes a northward shift in the tropospheric jet, which is presumably driven by zonal momentum eddy forcing (Limpasuvan and Hartmann 1999). Slow mode 1 (dashed curve) associates a southward-shifted polar stratospheric jet with a northward-shifted tropospheric jet. The reasons for these changes are unknown but enhanced GHG forcing is one possibility that will be explored in the next section.

Figure 5 shows the mean sea level pressure (MSLP) regression patterns associated with these modes. The fast mode 1 anomalies are confined to the central Arctic and are consistent with a mechanism laid out by Ambaum and Hoskins (2002). Accordingly, a strong polar stratospheric jet elevates the polar tropopause, which spins up the tropospheric column and lowers the surface pressure over the central Arctic. The fast mode 2 anomalies are more midlatitude in nature with opposing centers in the North Pacific and North Atlantic. Slow mode 1 describes generally decreasing surface pressure over the Arctic and increasing surface pressure over the midlatitude landmasses. These u-based MSLP patterns are related to the MSLP patterns obtained following a direct PCA of equal area–weighted MSLP data. The first so-obtained MSLP pattern (not shown) correlates with the slow mode 1 MSLP regression pattern at r ≈ 0.99. The second MSLP pattern (not shown) correlates with the first and second fast mode MSLP regression patterns at r = 0.90 and 0.95, respectively. It is worthwhile noting that the second MSLP pattern is the AO (which in monthly mean data appears as the first mode, see Thompson and Wallace 2000). Interestingly, a multiple linear regression shows that the AO spatial pattern correlates with 0.4 × (fast mode 1) + 0.9 × (fast mode 2) at r ≈ 0.99. Similarly, the AO PC correlates very highly with a linear combination of the leading fast mode PCs. This suggests that the AO superimposes two distinct physical entities: one involving central Arctic MSLP variations associated with fluctuations in the stratosphere polar vortex and another involving midlatitude MSLP variations driven by north–south shifts in the tropospheric jet. Finally, we note that these relationships (as elsewhere in the paper) are as robust for winter half-year averages (or December–February averages) as they are for the present annual averages.

The fast modes were derived here from aF = aaS where aS is a slowly evolving base state. Using a time mean base state yields the following approximate correspondence: mode 1 ↔ fast mode 1, mode 2 ↔ fast mode 2, and mode 3 ↔ slow mode 1. In other words, the slow component of u projects only weakly onto the fast modes and so appears as a distinct mode (i.e., mode 3) when using a time mean base state. This need not always be the case, as we shall see.

2) Simulations

The control simulation considered here uses approximately present-day CO2 concentration and is described fully in Flato et al. (2000). Figure 6 shows the control simulation's leading modes. Since the annual mean radiative forcing is fixed in this simulation the base state is very nearly independent of time. The simulated fast mode 1 compares well with the observed fast mode 1 (at r ≈ 0.82) except for an underestimated polar stratospheric center. The simulated fast mode 2 compares well with the observed fast mode 2 (at r ≈ 0.49) outside of the stratosphere. For the following reasons it is not useful to compare the second fast modes in the stratosphere: 1) as noted earlier the observed stratospheric anomalies are not especially significant, and 2) the simulated stratospheric anomalies are not robust to subsampling of the data. In terms of MSLP we find that the control simulation patterns (not shown) are similar to the observed fast mode patterns (Fig. 5), and superimpose to nearly perfectly explain the AO.

The climate change simulations considered here use a GHG forcing change corresponding to that observed from 1850 to the present, and a forcing change corresponding to an increase of effective CO2 at a rate of 1% per year (compounded) thereafter until year 2100. The direct forcing effect of sulfate aerosols is included by increasing the surface albedo. The simulations are described fully in Boer et al. (2000). Figure 7 shows the ensemble mean base-state parameters aS. The mean response is confined to the upper troposphere and above where the westerlies amplify, shift southward, and broaden. The leading fast modes are computed relative to this ensemble mean base state. Figure 8 shows these modes from one of the climate change simulations (the other climate change simulation modes are very similar). The fast mode structures are statistically indistinguishable from the control simulation fast modes (Fig. 6), correlating as they do at r ≈ 0.96 and r ≈ 0.88 for the first and second modes, respectively.

We now compare the observed and climate change simulation slow modes (Figs. 3 and 8, respectively). In the stratosphere, both slow modes show weakened tropical easterlies and a southward shift in the extratropical westerlies. In the troposphere, the observed extratropical westerlies shift northward while they do not in the model. Importantly, any agreement we see here mostly vanishes when the model calculation is restricted to the observational period (i.e., 1958–99). In other words, the late twentieth century NH zonal mean circulation changes are not consistent with those simulated with this model under historical global warming. Whether this is due to a model deficiency and/or a problem with the observations and/or sampling uncertainty is unknown at this time. It is worthwhile noting, however, that these historical differences do not preclude a zonal mean circulation response in the future similar to that predicted with this model under global warming.

Having isolated the mean response in the model we now ask if the variability modes are sensitive to global warming. Figure 9 presents a statistical analysis of contiguous 21-yr period standard deviations (σ, beginning in year 1916) for the fast mode PCs obtained from each of the three climate change simulations. For comparison we also show the observed standard deviations as the open circles. The vertical lines represent confidence intervals at the 10% significance level. We see that the confidence intervals are always overlapping, which indicates that the standard deviations from period to period are statistically indistinguishable from one another. This is confirmed using Bartlett's test (again at the 10% significance level), which tests the equality of multiple variances (von Storch and Zwiers 1999, p. 180). From these tests we conclude that the strength of the leading modes of NH u-variability in this model are insensitive to global warming. Additionally, we note that the fast mode structures obtained using the first and last 100 yr of data are statistically identical to one another, and to the fast modes shown in Fig. 8.

Finally, we mention that as in Fig. 1 the variability modes (not shown) derived using a time mean state seriously confound the variability and forced structures identified in Fig. 8.

The main findings so far are the following:

  • The first mode of NH u-variability describes polar stratospheric jet fluctuations with significant influence down to the Arctic surface.
  • The second mode is tropospheric and describes north–south shifts in the subtropical jet.
  • The surface pressure manifestations of these modes are intimately linked to the AO, and together suggest separate stratospheric and tropospheric origins for the AO.
  • The first observed mode appears unchanged since the 1950s, while the second mode appears to have gained strength.
  • The leading simulated modes appear unchanged in strength since the beginning of the twentieth century, and are predicted to remain so to the end of the twenty-first century.

b. Southern Hemisphere

1) Observations

Figures 10 and 11 show the leading observed modes in physical and parameter space, respectively. Fast mode 1 shows that a southward-shifted tropospheric jet is associated with a strengthened polar stratospheric jet. Polvani and Kushner (2002) find a similar relationship when the polar stratospheric jet is systematically perturbed in a mechanistic GCM. Fluctuations in the SH polar stratospheric jet are thought to partly owe their existence to transient planetary waves generated by synoptic-scale baroclinic eddies in the troposphere (e.g., Scinocca and Haynes 1998). It is plausible that this planetary wave emission, and hence, wave breaking in the stratosphere, is sensitive to the position of the tropospheric jet. Fast mode 2 is primarily tropospheric where it describes a northward shift and amplification of the polar front jet. A test for equality of variances suggests that the strength of the modes is unchanged since the 1950s (at the 10% level of significance). Slow mode 1 shows that the entire extratropical jet structure has shifted southward over the period. Once again, determining whether these slow changes are real or arise from a time-evolving observing system is beyond the scope of this study.

2) Simulations

Figure 12 shows the leading control simulation modes. The simulated fast mode 1 compares well with the observed fast mode 1 (at r ≈ 0.72) outside the Tropics. This supports the contention that the main mode of stratospheric–tropospheric coupling involves a southward shift in the tropospheric jet accompanying an amplification of the polar stratospheric jet. The simulated fast mode 2 compares favorably with the observed fast mode 2 (r ≈ 0.63) in the troposphere but not in the stratosphere. Figure 13 shows the leading control simulation MSLP regression patterns. Fast mode 1 describes an increased MSLP gradient north of the circumpolar trough (at about −65°) while fast mode 2 describes a northward shift of the trough itself (especially in the Pacific sector). The fast mode 1 pattern is nearly identical to model's version of the AAO (or leading mode of equal area–weighted MSLP data).

Figure 14 shows the ensemble mean base-state parameters aS from the climate change simulations. The mean response involves an amplifying and northward-shifting polar stratospheric jet in conjunction with a southward-shifting tropospheric jet. As before, the leading fast modes are computed relative to this ensemble mean base state. Figure 15 shows that the fast mode structures so computed are statistically indistinguishable from the control simulation fast modes (correlating as they do at r ≈ 0.97 for both modes). Comparing the observed and climate change simulation slow modes (Figs. 15 and 10, respectively) we see that in the troposphere both show a southward-shifting tropospheric jet but in the stratosphere the changes are opposite. It is worthwhile nothing that when the model calculation is restricted to 1958–99 the slow mode 1 pattern is similar but muted when compared to the full period computation.

We now ask if the SH variability modes are sensitive to global warming. Figure 16 presents a statistical analysis of contiguous 21-yr-period standard deviations for the fast mode PCs from each of the climate change simulations. As can be seen, the confidence intervals are always overlapping except for fast mode 1 in run 1 and fast mode 2 in run 2. Consistent with this, Bartlett's test identifies just one period for each mode with a variance unlike the others. Taking into account that at least one sample variance would be expected to differ from the rest at the 10% significance level we conclude that the variances are statistically indistinguishable. From these tests it appears that the strength of the leading modes of SH u-variability in this model are insensitive to global warming. Additionally, we note that the fast mode structures obtained using the first and last 100-yr of data are statistically identical to one another, and to the fast modes shown in Fig. 15.

Finally, we note again that the variability modes (not shown) derived using a time mean state seriously confound the variability and forced structures identified in Fig. 15.

The main findings from this section are the following:

  • The first mode of SH u-variability describes north–south shifts in the polar front jet accompanied by polar stratospheric jet and AAO-like surface pressure fluctuations.
  • The second mode is primarily tropospheric and describes fluctuations in the strength and position of the polar front jet.
  • The leading observed modes appear unchanged in strength since the 1950s, while the corresponding simulated modes are unchanged since the beginning of the twentieth century and are predicted to remain so to the end of the twenty-first century.

4. Conclusions

In the NH, the first mode of u-variability describes polar stratospheric jet fluctuations with significant influence down to the Arctic surface. The second mode is tropospheric and describes north–south shifts in the subtropical jet. The surface pressure manifestations of these modes are intimately linked to the AO, and together suggest separate stratospheric and tropospheric origins for the AO. In the SH, the first mode describes north–south shifts in the extratropical tropospheric jet accompanied by polar stratospheric jet and AAO-like surface pressure fluctuations. The second mode is primarily tropospheric and describes fluctuations in the strength and position of the polar front jet.

The leading observed modes appear unchanged since the 1950s except in the Northern Hemisphere where the second mode shows some evidence of increasing strength. The leading simulated modes appear unchanged in strength since the beginning of the twentieth century, and are predicted to remain so to the end of the twenty-first century. In all cases the leading modes are superimposed upon significant mean change, which when not properly accounted for leads to erroneous conclusions. As the final word, we caution that our conclusions regarding modal change are based upon a single model prediction and need to be verified against other model predictions.

Acknowledgments

This work was begun while visiting the Max-Planck-Institut für Meteorologie in Hamburg, Germany. Many very helpful discussions with Elisa Manzini, Marco Giorgetta, and Mojib Latif are gratefully acknowledged. We also thank Greg Flato, Nathan Gillett, Slava Kharin, Francis Zwiers, and the three anonymous reviewers for their many useful suggestions.

REFERENCES

  • Ambaum, M. H. P., , and B. J. Hoskins, 2002: The NAO troposphere–stratosphere connection. J. Climate, 15 , 19691978.

  • Boer, G. J., , G. M. Flato, , and D. Ramsden, 2000: A transient climate change simulation with historical and projected greenhouse gas and aerosol forcing: Projected climate for the 21st century. Climate Dyn., 16 , 427450.

    • Search Google Scholar
    • Export Citation
  • Chen, P., , and W. A. Robinson, 1992: Propagation of planetary waves between the troposphere and stratosphere. J. Atmos. Sci., 49 , 25332545.

    • Search Google Scholar
    • Export Citation
  • Flato, G. M., , G. J. Boer, , W. G. Lee, , N. A. McFarlane, , D. Ramsden, , M. C. Reader, , and A. J. Weaver, 2000: The Canadian Centre for Climate Modelling and Analysis global coupled model and its climate. Climate Dyn., 16 , 451467.

    • Search Google Scholar
    • Export Citation
  • Fyfe, J. C., , G. J. Boer, , and G. M. Flato, 1999: The Arctic and Antarctic oscillations and their projected changes under global warming. Geophys. Res. Lett., 26 , 16011604.

    • Search Google Scholar
    • Export Citation
  • Gillett, P. G., , M. R. Allen, , and K. D. Williams, 2002: The role of stratospheric resolution in simulating the Arctic Oscillation response to greenhouse gases. Geophys. Res. Lett.,29 (10), 1500, doi:10.1029/2001GL014444.

    • Search Google Scholar
    • Export Citation
  • Houghton, J. T., , Y. Ding, , D. J. Griggs, , M. Noguer, , P. J. van der Linden, , and D. Xiaosu, Eds.,. 2001: Climate Change 2001: Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, 944 pp.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors. 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Latif, M., 2000: On the interpretation of climate change in the tropical Pacific. MPI Rep. 306, Max-Planck-Institut für Meteorologie, Hamburg, Germany, 20 pp.

    • Search Google Scholar
    • Export Citation
  • Limpasuvan, V., , and D. L. Hartmann, 1999: Eddies and the annular modes. Geophys. Res. Lett., 26 , 31333136.

  • Marquardt, D., 1963: An algorithm for least squares estimation of nonlinear parameters. SIAM J. Appl. Math., 11 , 431441.

  • McFarlane, N. A., , G. J. Boer, , J-P. Blanchet, , and M. Lazare, 1992: The Canadian Climate Centre second generation general circulation model and its equilibrium climate. J. Climate, 5 , 10131044.

    • Search Google Scholar
    • Export Citation
  • Nigam, S., 1990: On the structure of variability of the observed tropospheric and stratospheric zonal-mean zonal wind. J. Atmos. Sci., 47 , 17991813.

    • Search Google Scholar
    • Export Citation
  • Pacanowski, R. C., , K. Dixon, , and A. Rosati, 1993: The GFDL modular ocean model users guide. GFDL Ocean Group Tech. Rep. 2, Geophysical Fluid Dynamics Laboratory, Princeton, NJ, 46 pp.

    • Search Google Scholar
    • Export Citation
  • Polvani, L. M., , and P. J. Kushner, 2002: Tropospheric response to stratospheric perturbations in a relatively simple general circulation model. Geophys. Res. Lett.,29 (7), 1114, doi:10.1029/2001GL014284.

    • Search Google Scholar
    • Export Citation
  • Scinocca, J. F., , and P. H. Haynes, 1998: Dynamical forcing of stratospheric waves by tropospheric baroclinic waves. J. Atmos. Sci., 55 , 23612392.

    • Search Google Scholar
    • Export Citation
  • Shindell, D. T., , R. L. Miller, , G. A. Schmidt, , and L. Pandolfo, 1999: Simulation of recent Northern Hemisphere climate trends by greenhouse-gas forcing. Nature, 399 , 452455.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W., , and J. M. Wallace, 2000: Annular modes in the extratropical circulation. Part I: Month-to-month variability. J. Climate, 13 , 10001016.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., , D. P. Stepaniak, , and J. W. Hurrell, 2001: Quality of reanalyses in the Tropics. J. Climate, 14 , 14991510.

  • von Storch, H., , and F. W. Zwiers, 1999: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.

  • Zhang, Y., , J. M. Wallace, , and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900–93. J. Climate, 10 , 10041020.

Fig. 1.
Fig. 1.

(left) Time mean and (right) leading PCA modes for the constructed latitudinal jet u(θ,t) = A exp[−(θ − Θ)2/2W2] with fixed width W but variable amplitude A(t) and position Θ(t). (a) A(t) and Θ(t) fluctuating pseudorandomly about constant values. (b) A(t) and Θ(t) fluctuating as in (a) but with the latter also monotonically increasing (i.e., for a poleward drifting jet). Units are m s–1

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 2.
Fig. 2.

(top) Observed (NCEP) and (bottom) control simulation climatological zonal mean zonal wind with 4 m s–1 contours (… ,−6, −2, 2, …). Thick contours are negative. Time mean jet axes are indicated by the solid dots

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 3.
Fig. 3.

Observed NH variability modes. (top) Standardized principal components. (bottom) Zonal mean zonal wind regression patterns with 0.2 m s–1 contours (… , −0.3, −0.1, 0.1, …). Thick contours are negative. Variances explained in upper left. Hatching indicates significance at the 1% significance level. Time mean jet axes are indicated by the solid dots

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 4.
Fig. 4.

Observed NH parameter regression curves. Solid black curves are fast mode 1; gray curves are fast mode 2; dashed curves are slow mode 1

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 5.
Fig. 5.

Observed NH MSLP regression patterns with 0.3 hPa (… , −0.45, −0.15, 0.15, …) contours. Thick contours are negative. Outer lat circle is 20°. Inner lat circle is 60°

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 6.
Fig. 6.

Same as Fig. 3, but for control simulation NH variability modes

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 7.
Fig. 7.

NH slowly varying parameters from an ensemble of climate change simulations. Here aS1 is in units of m;ths–1 and aS2 and aS3 in degrees

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 8.
Fig. 8.

Same as Fig. 3, but for climate change simulation NH variability modes

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 9.
Fig. 9.

Std devs (filled circles) of the NH fast mode PCs for contiguous 21-yr periods (beginning in 1916) and for each of the climate change simulations. The open circles are for the observed fast mode PCs. The vertical lines represent confidence intervals at the 10% significance level. Run 1 corresponds to the run presented in Fig. 8

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 10.
Fig. 10.

Same as Fig. 3, but for observed SH variability modes

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 11.
Fig. 11.

Same as Fig. 4, but for observed SH parameter regression curves

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 12.
Fig. 12.

Same as Fig. 3, but for control simulation SH variability modes

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 13.
Fig. 13.

Same as Fig. 5, but for control simulation SH MSLP regression patterns. Outer lat circle is −20°. Inner lat circle is −60°.

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 14.
Fig. 14.

Same as Fig. 7, but for SH slowly varying parameters

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 15.
Fig. 15.

Same as Fig. 3, but for climate change simulation SH variability modes

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Fig. 16.
Fig. 16.

Same as Fig. 9, but for the SH fast modes

Citation: Journal of Climate 16, 5; 10.1175/1520-0442(2003)016<0863:SEZWVA>2.0.CO;2

Save