• Akahori, K., , and S. Yoden, 1997: Zonal flow vacillation and bimodality of baroclinic eddy life cycles in a simple global circulation model. J. Atmos. Sci., 54 , 23492361.

    • Search Google Scholar
    • Export Citation
  • Ambaum, M. H., , and B. J. Hoskins, 2002: The NAO troposphere–stratosphere connection. J. Climate, 15 , 19691978.

  • Andrews, D. G., , J. R. Holton, , and C. B. Leovy, 1987: Middle Atmosphere Dynamics. Academic Press, 489 pp.

  • Baldwin, M. P., , D. B. Stephenson, , D. W. J. Thompson, , T. J. Dunkerton, , A. J. Charlton, , and A. O’Nell, 2003: Stratospheric memory and extended-range weather forecasts. Science, 301 , 636640.

    • Search Google Scholar
    • Export Citation
  • Barnston, A. G., , and R. E. Livezey, 1987: Classification, seasonality, and persistence of low-frequency atmospheric circulation patterns. Mon. Wea. Rev., 115 , 10831126.

    • Search Google Scholar
    • Export Citation
  • Benedict, J. J., , S. Lee, , and S. B. Feldstein, 2004: Synoptic view of the North Atlantic oscillation. J. Atmos. Sci., 61 , 121144.

  • Cash, B. A., , P. J. Kushner, , and G. K. Vallis, 2002: The structure and composition of the annular modes in an aquaplanet general circulation model. J. Atmos. Sci., 59 , 33993414.

    • Search Google Scholar
    • Export Citation
  • Chang, E. K. M., , and D. B. Yu, 1999: Characteristics of wave packets in the upper troposhere. Part I: Northern Hemisphere winter. J. Atmos. Sci., 56 , 17081728.

    • Search Google Scholar
    • Export Citation
  • Feldstein, S. B., 1998: An observational study of the intraseasonal poleward propagation of zonal mean flow anomalies. J. Atmos. Sci., 55 , 25162529.

    • Search Google Scholar
    • Export Citation
  • Feldstein, S. B., 2000a: Is interannual zonal mean flow variability simply climate noise? J. Climate, 13 , 23562362.

  • Feldstein, S. B., 2000b: The timescale, power spectra, and climate noise properties of teleconnection patterns. J. Climate, 13 , 44304440.

    • Search Google Scholar
    • Export Citation
  • Feldstein, S. B., , and S. Lee, 1998: Is the atmospheric zonal index driven by an eddy feedback? J. Atmos. Sci., 55 , 30773086.

  • Franzke, C., , S. Lee, , and S. B. Feldstein, 2004: Is the North Atlantic Oscillation a breaking wave? J. Atmos. Sci., 61 , 145160.

  • Gerber, E., 2005: A dynamical and statistical understanding of the NAO and annular modes. Ph.D. thesis, Princeton University, 242 pp.

  • Gerber, E., , and G. K. Vallis, 2005: A stochastic model for the spatial structure of annular patterns of variability and the North Atlantic Oscillation. J. Climate, 18 , 21022118.

    • Search Google Scholar
    • Export Citation
  • Gerber, E., , and G. K. Vallis, 2007: Eddy–zonal flow interactions and the persistence of the zonal index. J. Atmos. Sci., 64 , 32963311.

    • Search Google Scholar
    • Export Citation
  • Hurrell, J. W., , Y. Kushnir, , M. Visbeck, , and G. Ottersen, 2003: An overview of the North Atlantic Oscillation. The North Atlantic Oscillation: Climate Significance and Environmental Impact, Geophys. Monogr., Vol. 134, Amer. Geophys. Union, 1–35.

    • Search Google Scholar
    • Export Citation
  • Kornich, H., , G. Schmitz, , and E. Becker, 2006: The role of stationary waves in the maintenance of the northern annular mode as deduced from model experiments. J. Atmos. Sci., 63 , 29312947.

    • Search Google Scholar
    • Export Citation
  • Kravtsov, S., , A. W. Robertson, , and M. Ghil, 2006: Multiple regimes and low-frequency oscillations in the Northern Hemisphere’s zonal-mean flow. J. Atmos. Sci., 63 , 840860.

    • Search Google Scholar
    • Export Citation
  • Kushner, P. J., , and L. M. Polvani, 2004: Stratosphere–troposphere coupling in a relatively simple AGCM: The role of eddies. J. Climate, 17 , 629639.

    • Search Google Scholar
    • Export Citation
  • Lee, S., , and S. B. Feldstein, 1996: Two types of wave breaking in an aquaplanet GCM. J. Atmos. Sci., 53 , 842857.

  • Lee, S., , S-W. Son, , K. Grise, , and S. B. Feldstein, 2007: Mechanism of poleward propagation of zonal mean flow anomalies. J. Atmos. Sci., 64 , 849868.

    • Search Google Scholar
    • Export Citation
  • Lorenz, D. J., , and D. L. Hartmann, 2001: Eddy–zonal flow feedback in the Southern Hemisphere. J. Atmos. Sci., 58 , 33123327.

  • Lorenz, D. J., , and D. L. Hartmann, 2003: Eddy–zonal flow feedback in the Northern Hemisphere winter. J. Climate, 16 , 12121227.

  • Norton, W. A., 2003: Sensitivity of Northern Hemisphere surface climate to simulation of the stratospheric polar vortex. Geophys. Res. Lett., 30 .1627, doi:10.1029/2003GL016958.

    • Search Google Scholar
    • Export Citation
  • Oort, A. H., , and J. J. Yienger, 1996: Observed interannual variability in the Hadley circulation and its connection to ENSO. J. Climate, 9 , 27512767.

    • Search Google Scholar
    • Export Citation
  • Orlanski, I., 2003: Bifurcation in eddy life cycles: Implications for storm track variability. J. Atmos. Sci., 60 , 9931023.

  • Riehl, H., , T. C. Yeh, , and N. E. La Seur, 1950: A study of variations of the general circulation. J. Meteor., 7 , 181194.

  • Ring, M. J., , and R. A. Plumb, 2007: Forced annular mode patterns in a simple atmospheric general circulation model. J. Atmos. Sci., 64 , 36113626.

    • Search Google Scholar
    • Export Citation
  • Robinson, W. A., 1991: The dynamics of the zonal index in a simple model of the atmosphere. Tellus, 43A , 295305.

  • Robinson, W. A., 2006: On the self-maintenance of midlatitude jets. J. Atmos. Sci., 63 , 21092122.

  • Shiogama, H., , T. Terao, , H. Kida, , and T. Iwashima, 2005: Roles of low- and high-frequency eddies in the transitional process of the Southern Hemisphere annular mode. J. Climate, 18 , 782794.

    • Search Google Scholar
    • Export Citation
  • Son, S-W., , and S. Lee, 2005: The response of westerly jets to thermal driving in a primitive equation model. J. Atmos. Sci., 62 , 37413757.

    • Search Google Scholar
    • Export Citation
  • Son, S-W., , and S. Lee, 2006: Preferred modes of variability and their relationship with climate change. J. Climate, 19 , 20632075.

  • Thompson, D. W. J., , and J. M. Wallace, 1998: The Arctic Oscillation signature in the wintertime geopotential height and temperature fields. Geophys. Res. Lett., 25 , 12971300.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W. J., , 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
  • Thorncroft, C. D., , B. J. Hoskins, , and M. E. McIntyre, 1993: Two paradigms of baroclinic life-cycle behavior. Quart. J. Roy. Meteor. Soc., 119 , 1755.

    • Search Google Scholar
    • Export Citation
  • Vallis, G. K., , E. P. Gerber, , P. J. Kushner, , and B. A. Cash, 2004: A mechanism and simple dynamical model of the North Atlantic Oscillation and annular modes. J. Atmos. Sci., 61 , 264280.

    • Search Google Scholar
    • Export Citation
  • Wallace, J. M., , and D. S. Gutzler, 1981: Teleconnections in the geopotential height field during the Northern Hemisphere winter. Mon. Wea. Rev., 109 , 784812.

    • Search Google Scholar
    • Export Citation
  • Yu, J-Y., , and D. L. Hartmann, 1993: Zonal-flow vacillation and eddy forcing in a simple GCM of the atmosphere. J. Atmos. Sci., 50 , 32443259.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    (a)–(f) Linear regression of 〈[u]〉′ at all latitudes against 〈[u]〉′ at a base latitude of 30°. All six M0 runs are illustrated. The contour interval is 0.2 m s−1. Solid contours are positive, dashed contours negative, and the zero contour is omitted. The regressed 〈[u]〉′ values, which are statistically significant at the 99% confidence level, are shaded. The right panel of each frame shows the latitudinal profiles of the time-mean 〈[u]〉 (thick gray line), EOF1 (thin solid line), and EOF2 (thin dashed line). The location of polar front or eddy-driven jet is denoted with a straight line.

  • View in gallery

    (a) Autocorrelation coefficients for PC1 in all six M0 runs, and (b) their corresponding e-folding time scale (ε). In (a), the poleward propagation cases, CS5 and CS6, are discriminated from the other cases with a dashed line. In (b), the upper and lower limits of the error bars correspond to the values of ε in each hemisphere, and open circles indicate an average.

  • View in gallery

    Lag-correlation coefficients between F1 and PC1 for the (a) CS2 and (b) CS6 M0 runs. Negative time lag denotes that F1 leads PC1. The values that are statistically significant at the 99% (90%) confidence level are indicated by dark (light) gray. The number within each panel denotes the time scale of the positive feedback, which is defined as the maximum lag day for which the correlation coefficient is statistically significant at the 90% confidence level.

  • View in gallery

    (a) Autocorrelation coefficients for PC1 in all six M1 runs, and (b) the values of the e-folding time scales in all M0 (circle), M1 (triangle), and M2 runs (square).

  • View in gallery

    Lag-correlation coefficients between F1 and PC1 for (a)–(c) the CS2 and (d)–(f) CS6 M1 runs; (a), (d) total forcing, (b), (e) momentum flux forcing, and (c), (f) mountain torque forcing.

  • View in gallery

    As in Figs. 1b and 1f, but for the M1 runs.

  • View in gallery

    Scatterplots of the WBL and PC1 for the (a) CS2 and (b) CS6 M0 runs. The probability density functions are illustrated at each axis. White crosses in the figure denote the location of jets. Two crosses in (b) correspond to subtropical and eddy-driven jets, respectively. Note that the sign of PC1 in (b) is reversed in order that the sign of the midlatitude EOF1 in the CS6 M0 run matches with that in the CS2 M0 run. While the positive phase of EOF1 in the CS2 M0 run corresponds to positive 〈[u]〉′ at 50° (Fig. 1b), that in the CS6 M0 run corresponds to negative 〈[u]〉′ (Fig. 1f).

  • View in gallery

    Composite time series of (a) PC1 and (b) WBL in the CS2 M0 run. The value of PC1 is normalized by 1 std dev. The composite fields are constructed with respect to the PC1 extrema. A total of 329 positive and 333 negative cases, in which the local extrema are greater or less than 1 std dev of PC1 time series, are utilized. Positive (negative) cases are indicated with solid (dashed) lines. Both PC1 and WBL are slightly smoothed with 10-day low-pass filter.

  • View in gallery

    One-point correlation maps of 250-hPa eddy streamfunction in the (a) CS2 M0, (b), (c) CS2 M1, (d) CS6 M0, and (e), (f) CS6 M1 runs. The contour interval is 0.2. Solid contours are positive, dashed contours negative, and the zero contour is omitted. The shading and thick cross symbol denote the mountain and the base point for the correlation, respectively. The superimposed thin contours are the time-mean qy at 250 hPa. The contour interval is 2.5 × 10−11 m−1 s−1 and zero values are denoted with a thick gray line.

  • View in gallery

    The values of the refractive index for the (a) CS2 M0, (b) CS2 M1, (c) CS6 M0, and (d) CS6 M1 runs. The values are slightly smoothed with a nine-point smoother. The contour interval is 2 × 10−7 m−2. The superimposed shading denotes the mountain. The insets in (a) and (c) show the refractive index for the zonal-mean flow from equator to 60°N. The values displayed in the insets range from −1 × 10−9 to 1 × 10−9 m−2.

  • View in gallery

    The e-folding time scale as a function of (a) mountain height and (b) model resolution. In (a), the M0 and 2-km M1 runs are marked with a circle and triangle, respectively. In (b), only the zonal resolution is varied, while the maximum meridional wavenumber is fixed at 30. A maximum zonal wavenumber of 15 (60) is denoted by K15 (K60).

  • View in gallery

    As in Figs. 1b and 1f, but for the 5-km M1 runs.

  • View in gallery

    The contribution by stationary (long-dashed lines) and transient (short-dashed lines) eddies to the total (solid lines) zonal-mean eddy momentum flux convergence at 250 hPa in the (a) CS2 M0 and (b)–(f) the CS2 M1 runs with varying height. The mountain height in each run is denoted at the top of each panel.

  • View in gallery

    Various quantities from the CS2 5-km M1 run. (a), (b) One-point correlation maps of 250-hPa eddy streamfunction with the base point denoted by a strong cross mark and (c) the refractive index. The shading and contour intervals are the same as in Figs. 9 and 10.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 48 48 1
PDF Downloads 40 40 1

Time Scale and Feedback of Zonal-Mean-Flow Variability

View More View Less
  • 1 Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania
  • | 2 Earth and Environmental Systems Institute, The Pennsylvania State University, University Park, Pennsylvania
  • | 3 Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania
© Get Permissions
Full access

Abstract

The physical processes that determine the time scale of zonal-mean-flow variability are examined with an idealized numerical model that has a zonally symmetric lower boundary. In the part of the parameter space where the time-mean zonal flow is characterized by a single (double) jet, the dominant form of zonal-mean-flow variability is the zonal index (poleward propagation), and the time-mean potential vorticity gradient is found to be strong and sharp (weak and broad). The e-folding time scale of the zonal index is found to be close to 55 days, much longer than the observed 10-day time scale. The e-folding time scale of the poleward propagation is about 40 days. The long e-folding time scales for the zonal index are found to be consistent with an unrealistically strong and persistent eddy–zonal-mean-flow feedback. A calculation of the refractive index indicates that the background flow supports eddies that are trapped within midlatitudes, undergoing relatively little meridional propagation.

Additional model runs are performed with an idealized mountain to investigate whether zonal asymmetry can disrupt the eddy feedback. For single-jet states, the time scale is reduced to about 30 days if the mountain height is 4 km or less. The reduction in the time scale occurs because the stationary eddies excited by the mountain alter the background flow in a manner that leads to the replacement of zonal-index events by shorter-time-scale poleward propagation. With a 5-km mountain, the time scale reverts and increases to 105 days. This threshold behavior is again attributed to a sharpening of the background zonal jet, which arises from an extremely strong stationary wave momentum flux convergence. In contrast, for double-jet states, the time scale changes only slightly and the poleward propagation is maintained in all mountain runs.

* Current affiliation: Department of Applied Physics and Applied Mathematics, Columbia University, New York, New York

+ Current affiliation: Department of Civil and Environmental Engineering, Stanford University, Stanford, California

Corresponding author address: Seok-Woo Son, Department of Applied Physics and Applied Mathematics, Columbia University, 200 S.W. Mudd Building, 500 W. 120th St., New York, NY 10027. Email: sws2112@columbia.edu

Abstract

The physical processes that determine the time scale of zonal-mean-flow variability are examined with an idealized numerical model that has a zonally symmetric lower boundary. In the part of the parameter space where the time-mean zonal flow is characterized by a single (double) jet, the dominant form of zonal-mean-flow variability is the zonal index (poleward propagation), and the time-mean potential vorticity gradient is found to be strong and sharp (weak and broad). The e-folding time scale of the zonal index is found to be close to 55 days, much longer than the observed 10-day time scale. The e-folding time scale of the poleward propagation is about 40 days. The long e-folding time scales for the zonal index are found to be consistent with an unrealistically strong and persistent eddy–zonal-mean-flow feedback. A calculation of the refractive index indicates that the background flow supports eddies that are trapped within midlatitudes, undergoing relatively little meridional propagation.

Additional model runs are performed with an idealized mountain to investigate whether zonal asymmetry can disrupt the eddy feedback. For single-jet states, the time scale is reduced to about 30 days if the mountain height is 4 km or less. The reduction in the time scale occurs because the stationary eddies excited by the mountain alter the background flow in a manner that leads to the replacement of zonal-index events by shorter-time-scale poleward propagation. With a 5-km mountain, the time scale reverts and increases to 105 days. This threshold behavior is again attributed to a sharpening of the background zonal jet, which arises from an extremely strong stationary wave momentum flux convergence. In contrast, for double-jet states, the time scale changes only slightly and the poleward propagation is maintained in all mountain runs.

* Current affiliation: Department of Applied Physics and Applied Mathematics, Columbia University, New York, New York

+ Current affiliation: Department of Civil and Environmental Engineering, Stanford University, Stanford, California

Corresponding author address: Seok-Woo Son, Department of Applied Physics and Applied Mathematics, Columbia University, 200 S.W. Mudd Building, 500 W. 120th St., New York, NY 10027. Email: sws2112@columbia.edu

1. Introduction

Teleconnection patterns impact a very broad range of time scales, from just beyond the period of synoptic-scale variability to interannual and interdecadal time scales. Although teleconnection patterns are a dominant feature of climate variability, the intrinsic time scale of most teleconnection patterns is found to be less than 10 days (Feldstein 2000b). Therefore, one crucial aspect for improving our understanding of climate variability and prediction must involve gaining a deeper knowledge of the atmospheric dynamical processes at this fundamental time scale.

The North Atlantic Oscillation (NAO) (Hurrell et al. 2003) and the very similar northern annular mode (Thompson and Wallace 1998) are the most prominent teleconnection patterns in the Northern Hemisphere (see also Wallace and Gutzler 1981; Barnston and Livezey 1987). Consistent with the relatively short 10-day time scale for the NAO, Benedict et al. (2004) and Franzke et al. (2004) showed that at this intrinsic time scale the NAO pattern comprises the remnants of a breaking synoptic-scale wave. They found that the positive (negative) phase of the NAO arises from anticyclonic1 (cyclonic) wave breaking. In this picture, the NAO time scale is determined essentially by the time scale of wave steepening and breaking.

For the zonal-mean analog of the NAO, known as the zonal index, the observed time scales vary between 8 and 18 days, depending upon the hemisphere and the season (Feldstein 2000a). This time scale is longer than that for synoptic-scale eddies. Several studies have suggested that this increase in the time scale relative to that for synoptic-scale eddies, for both the zonal index and also for the northern and southern annular modes (Thompson and Wallace 2000), arises from a positive eddy–zonal-mean-flow feedback process (Robinson 1991; Yu and Hartmann 1993; Feldstein and Lee 1998; Lorenz and Hartmann 2001, 2003). As a result, depending on the pervasiveness of the positive eddy feedback process, the time scale of the zonal index or the annular modes can potentially be much longer.

In addition to the eddy feedback, there is another element of complexity that needs to be considered in order to better understand the time scale. While transitions between the two phases of the zonal index and the annular modes can be described as being stochastic (Feldstein 2000a, b), it has also been shown that zonal-mean-flow anomalies sometimes undergo quasi-periodic poleward propagation (Riehl et al. 1950; Feldstein 1998; Kravtsov et al. 2006). Son and Lee (2006) show, with a series of climate model runs, that the dominant form of zonal-mean-flow variability depends upon the climate state. They find that quasi-periodic poleward propagation dominates the zonal-mean-flow variability for climate states with a time-mean potential vorticity (PV) gradient that is relatively weak and broad. With the same model, they also find that zonal-index variability dominates if the climate state is characterized by a time-mean PV gradient that is strong and sharp.

In this study, in model runs with a zonally uniform lower boundary, we will find that the time scale of the zonal index is close to 55 days, which is much longer than that observed in the atmosphere. To the extent that these model results are relevant for the atmosphere, this suggests that although an eddy feedback extends the time scale of the zonal index to beyond that of synoptic-scale eddies, there may also be a process in the atmosphere that disrupts the eddy feedback, preventing the zonal-index time scale from being much longer. Because of the zonal symmetry of the model’s lower boundary, it is plausible that some aspect of the atmosphere’s zonal asymmetry accounts for its much shorter zonal-index time scale than that in our model. In this study, we will address this issue by performing model simulations that include an idealized mountain. As we will see, important insight into the impact of zonal asymmetry on the zonal-index time scale can be attained by examining the results from model simulations that extend over a range of climate states whose internal variability is dominated either by zonal-index variability or by poleward propagation.

The organization of this paper is as follows. Section 2 describes the model and analysis methods. Section 3 presents results from base model runs, for which there is no mountain, and from 2-km mountain runs. Explanations for the results are offered in section 4. The sensitivity of the time scale to various mountain heights is evaluated in section 5. Section 6 then presents a summary and discussion.

2. Model and analysis methods

Because most numerical simulations analyzed in this study are essentially identical to those of Son and Lee (2005, 2006, hereafter SL05 and SL06, respectively), only a brief description is provided. The model used is the dynamical core of the Geophysical Fluid Dynamics Laboratory (GFDL) general circulation model. Its horizontal resolution is rhomboidal 30.2 In the vertical direction, there are 10 equally spaced sigma levels. This model is forced by relaxing the temperature field toward an equilibrium temperature profile, Te, with a time scale of 30 days.

The Te profile consists of a base profile, Tbase, and an anomalous profile, ΔTe, which are both symmetric about the equator (see Fig. 2 of SL05), and take the form
i1520-0469-65-3-935-e1
where H and C, respectively, denote additional tropical heating and high-latitude cooling, φ is latitude, and p is pressure. The parameters H and C are introduced to control the strength of the tropical circulation and the meridional width of the extratropical baroclinic zone. SL05 and SL06 found that not only the time-mean jet but also its internal variability are highly sensitive to the values of H and C.

Three sets of experiments are carried out. As described below, each set of experiments consists of six runs, each run with different values of H and C. The runs are 15 000 days in length, and the first 1000 days are discarded. The control set (M0) is performed with a zonally symmetric lower boundary. The other two sets of experiments are performed with a single mountain (M1) and with two mountains (M2), both with the same Te profile3 as the M0 set. The M1 and M2 simulations are performed in order to investigate the effect of zonal asymmetry on the internal variability. For the M1 set, the mountain has a Gaussian shape and is placed at 45°N and 90°E, with a half-width of 45° in longitude and 22.5° in latitude (see the shading later in Fig. 9). For the M2 set, a second mountain of the same size is added at 45°N and 90°W. The mountain height is fixed at 2 km for all mountain runs, except for those runs that examine the sensitivity to mountain height. Therefore, the mountain height is indicated in the text only if it differs from 2 km.

The six model runs in each set (total 18 model runs) correspond to the simulations along the slanting axis “E” in SL06 (see Table 1). The six climate states in each run are referred to as climate state 1 (CS1), climate state 2 (CS2), and so forth. For the M0 set, data from both hemispheres are used, because the flow statistics would be identical if not for sampling error.

a. Empirical orthogonal function analysis

The time scale of the internal variability, ε, is defined by the e-folding time scale of the leading empirical orthogonal function (EOF). The EOF analysis is performed with 〈[u]〉′cos φ, where [u] is the zonal-mean zonal wind, the angle brackets denote a mass-weighted vertical integration, and the prime indicates a deviation from the time mean. The cosine factor is multiplied to weigh the variance of 〈[u]〉′ with area. Temporal filtering is not applied and all EOFs are presented in units of meters per second by regressing 〈[u]〉′ onto the principal component (PC) time series. The positive (negative) phase of the EOF pattern is defined such that the most poleward peak of the pattern is positive (negative).

The internal variability is diagnosed with the vertically integrated zonal-mean zonal momentum equation. In the quasigeostrophic framework, this equation can be written as
i1520-0469-65-3-935-e2
where a is the radius of earth, p0 a reference pressure of 1000 hPa, ps surface pressure, hs topographic height, and R the residual that includes both surface and internal dissipation. The asterisk denotes the deviation from the zonal mean. The two forcing terms on the right-hand side of (2) are the eddy momentum flux convergence and the mountain torque, respectively. The sum of these two terms will be referred to as the total forcing.

The relationship between the forcing terms in (2) and 〈[u]〉EOF1, the zonal-mean zonal-wind anomaly corresponding to EOF1, is examined quantitatively. Following Lorenz and Hartmann (2001), the latitudinal profile of total forcing is projected onto the EOF1 pattern. The resulting time series, to be referred to as F1, is then correlated with PC1. If there is a positive feedback between total forcing and 〈[u]〉EOF1, then at both positive and negative lags (for positive time lags, F1 leads PC1), the lag correlation between F1 and PC1 must be positive, statistically significant, and show particular lag/lead characteristics.4 It is important to emphasize that if all of these requirements between F1 and PC1 are satisfied, then the results strongly suggest that a positive eddy feedback has taken place, but they do not fully prove that a positive feedback has occurred. To prove such a feedback, it would be necessary to show that the zonal-mean-flow anomaly has altered the eddy fluxes in such a manner so as to reinforce the anomaly. This would be very difficult to show.

b. Statistical significance

The statistical significance of all linear regression and correlation analyses are tested with the method utilized by Oort and Yienger (1996). The confidence limits are calculated with Fisher’s Z transformation where the number of degrees of freedom, Ndf, is given by
i1520-0469-65-3-935-eq1
Here, N denotes the sample size of 14 000 days, and rτ(x) the autocorrelation coefficient of variable x at lag τ days. See Oort and Yienger (1996) for details.

c. Wave-breaking latitude

The two types of wave breaking described in the introduction, cyclonic and anticyclonic (e.g., Thorncroft et al. 1993; Lee and Feldstein 1996; Orlanski 2003), are identified by the latitude where the dominant wave breaking occurs: that is, the wave breaking is regarded as cyclonic if it occurs at a relatively high latitude and vice versa.

The wave-breaking latitude (WBL) is calculated with the methodology introduced by Akahori and Yoden (1997), which is based on the fact that wave breaking is accompanied by a negative PV gradient on an isentropic surface. For this calculation, PV on the 330-K isentrope (PV330K) is chosen,5 because the 330-K isentrope lies at about 200 hPa in midlatitudes. The potential vorticity on the 330-K surface can be written as
i1520-0469-65-3-935-eq2
where ζ is the relative vorticity, f the Coriolis parameter, g the gravitational acceleration, and θ the potential temperature. Focusing on midlatitudes, the meridional PV330K gradient, PV330Ky, is calculated only when PV330K is greater than 0.5 PV units (PVU), and less than 6.0 PVU (1 PVU = 10−6 K m2 s−1 kg−1). These reference values correspond approximately to 25° and 65° latitude. The WBL is determined by the following summation, in which the latitude is weighed by negative values of PV330Ky, that is,
i1520-0469-65-3-935-e3
where
i1520-0469-65-3-935-eq3

It should be stated that the values of the WBL are somewhat sensitive to the minimum and maximum values specified for PV330K. For example, if the lower limit of PV330K is decreased, anticyclonic wave breaking is emphasized, since the WBL becomes smaller. However, other calculations, such as the correlation between PC1 and WBL, are insensitive to these limits.

d. Refractive index (n2)

To examine wave propagation characteristics, one-point lag correlations of the 250-hPa eddy streamfunction field and the refractive index, (Andrews et al. 1987), are calculated. Under the quasigeostrophic approximation,
i1520-0469-65-3-935-e4
where
i1520-0469-65-3-935-eq4
All symbols are standard; qy is the meridional gradient of quasigeostrophic potential vorticity, u the zonal wind, c the phase speed, H the scale height, N2 the Brunt–Väisälä frequency, k the zonal wavenumber, β the variation of the Coriolis parameter with latitude, f0 the Coriolis parameter at 45°, and θR a reference potential temperature. The overbar denotes a long-term time mean. In (4), the phase speed c, which is a function of longitude and latitude, is derived from the one-point lag correlations at each grid point as in Chang and Yu (1999, see their Fig. 6). The distance between the maximum correlation at the lag −1 day and that at the lag 1 day is divided by 2 days. The zonal wavenumber is set equal to 5, which is the most energetic wavenumber at 45° latitude. The refractive index was found to be rather insensitive to the choice of wavenumber.

3. Base runs

a. Control runs (M0)

The latitudinal profiles of the time-mean zonal wind for the M0 runs are illustrated in the right panels of Fig. 1 (gray lines). As can be seen, the strength and shape of the jets are highly sensitive to the tropical heating (H) and high-latitude cooling rates (C). When H is large and C is small (CS1 and CS2), a strong single jet emerges at ≈40°. In contrast, if H is small and C is large (CS5 and CS6), an eddy-driven jet separates itself from the subtropical jet, forming a weak double-jet state. These properties are described in full detail in Son and Lee (2005).

The above differences in the time-mean flow lead to a fundamentally different internal variability (Son and Lee 2006). In Fig. 1, the internal variability is summarized by regressing 〈[u]〉′ at all latitudes with 〈[u]〉′ at a base latitude of 30°. For single-jet states (CS1 and CS2), the 〈[u]〉′ variability takes on the form of a standing oscillation with a node at the center of the time-mean jet. Consistent with this regression pattern, EOF1 (black curve in the corresponding panel) represents a latitudinal meandering of the jet and explains a much higher fraction of the total variance than does EOF2 (Table 2). In contrast, for the double-jet states (CS5 and CS6), the regressed 〈[u]〉′ fields propagate poleward with time. For the intermediate CS3 and CS4 M0 runs, both zonal-index and poleward propagation can be seen.

It is noteworthy that the EOF1 pattern relative to the time-mean jet differs between the CS1–CS2 and CS5–CS6 M0 runs. While the nodes of EOF1 in the former runs are located at the jet center (Figs. 1a,b), those in the latter runs occur off the jet center (Figs. 1e,f). This structural difference is consistent with findings of Vallis et al. (2004) who showed that the peak in the EOF1 pattern is closer to the jet center if the jet is broader than the eddy stirring, baroclinic zone.

The PC1 autocorrelation functions for all M0 runs are presented in Fig. 2a. The correlation coefficient decreases slowly in CS1–CS3 M0 runs with a secondary peak near lag 10 days. Although it may be a coincidence, a similar secondary peak is also found in the observations [cf. Fig. 2a with Fig. 7 of Ambaum and Hoskins (2002)]. In contrast, the PC1 autocorrelation functions for the CS5–CS6 M0 runs decay relatively fast, lacking a shoulder, and then oscillate with time, as expected from the quasi-periodic poleward propagation of the zonal-mean flow (Figs. 1e,f). Therefore, the e-folding time scale can be understood to represent the decorrelation time scale in the CS1–CS4 M0 runs and approximately one-quarter of the period of oscillation in the CS5–CS6 M0 runs. Furthermore, even though the variability in the CS5–CS6 M0 runs (see Fig. 2a) is fundamentally different from that of the CS1–CS4 runs, that is, quasi-periodic behavior versus exponential decay, by using the e-folding time scale of the autocorrelation function, we can measure the persistence of the zonal-mean anomalies at a fixed latitude for all six runs. Thus, when interpreting the e-folding time scale, it is important that one also takes into account whether zonal index or poleward propagation is taking place. The e-folding time-scale values, illustrated in Fig. 2b, range from ∼40 days for the CS5–CS6 runs to ∼60 days for the CS1–CS4 M0 runs. These time scales are unrealistically long, and strongly contrast with those in the atmosphere, which vary from 8 to 18 days, as described in the introduction.

Why are the time scales so unrealistically long? Given that the zonal-mean flow is driven by eddy fluxes, the long time scale is likely due to a positive eddy–zonal flow feedback, which is stronger than in the atmosphere. As a measure of the positive feedback, lag correlations between F1, the projection of the eddy momentum flux convergence onto EOF1, and PC1 are calculated. Because the results are qualitatively similar among the CS1, CS2, CS3, and CS4 runs, and also between the CS5 and CS6 runs, results are shown only for CS2 and CS6. As can be seen for the CS2 run (Fig. 3a), F1 is significantly correlated with PC1, with positive correlations occurring at both negative and positive time lags. This meets the necessary requirements for a positive feedback between the eddy momentum fluxes and the zonal-mean zonal flow. Similar results were found by Lorenz and Hartmann (2001, 2003) for the annular modes in both hemispheres, but the implied feedback in their study was much weaker. For the CS2 run, the positive feedback, as measured by statistically significant positive correlations at positive time lags, persists for 61 days. In contrast, for CS6, the period of persistence is much shorter (Fig. 3b). This is clearly because of the poleward propagation, as indicated by the 140-day periodicity in the correlations.

As expected, the correlation analyses suggest that the very long e-folding time scales in the simulations are due to an unrealistically strong positive feedback. Therefore, the question of why the EOF1 persistence time scales are so long can be rephrased as why is the positive feedback so persistent? Because a GFDL aquaplanet general circulation model, which has a full cloud and radiation package, was also found to have a very long, 65-day time scale (not shown), it is unlikely that the very long time scales are an artifact of the simplifications within our model. One possibility is that it is the absence of zonal asymmetries in the model that allows for the long time scales and the very persistent eddy feedback. This perspective is alluded to by the findings of Franzke et al. (2004) and Benedict et al. (2004), who show that the existence of the much shorter time-scale NAO, which is a local representation of the zonal index, is dependent upon the presence of a strong stretching deformation field.

b. Mountain runs (M1 and M2)

To find out whether the time scale would be reduced by introducing zonal asymmetries to the background flow, additional experiments are performed with Gaussian-shaped mountains in midlatitudes. The results from these mountain runs are summarized in Fig. 4. For the CS5 and CS6 runs, the presence of the mountains have very little impact on the time scale. On the other hand, for the CS1–CS4 runs, there is a substantial reduction in the time scale, from about 55 days down to about 30 days (Fig. 4b). For each climate state, the value of ε for M1 and M2 are similar, which suggests that a single Gaussian mountain is sufficient at altering the internal dynamics required for reducing the time scale.

As expected from the foregoing analysis (section 3a), the key internal dynamics behind the sensitivity in the time scale involve eddy–mean flow interaction. For the CS2 runs, where the mountains have a dramatic impact in reducing the time scale, the F1–PC1 correlation (Fig. 5a) suggests that the eddy–zonal-mean-flow feedback is substantially weaker in the presence of mountains (cf. Figs. 3a and 5a). In contrast, for the CS6 runs, where the mountains have a negligible impact on the time scale, there is little change in the F1–PC1 correlation (cf. Figs. 3b and 5d).

The contribution from the eddy forcing and the mountain torque for the F1–PC1 relationship is examined separately in Fig. 5. In general, the eddy forcing is more than three times stronger than mountain torque (not shown). For the CS2 M1 run, while the mountain torque causes a further shortening of the time scale,6 we see that the weak eddy–mean flow interaction in the mountain runs stems primarily from a modification of the eddy fluxes by the mountain. For the CS6 M1 run, the time scale due solely to the eddy fluxes (Fig. 5e) is slightly longer than that in its control counterpart, but the mountain torque again shortens the time scale so that the time scale is very close to that of the control run.

The apparent weakening of the eddy feedback in the CS2 M1 run is not only consistent with the reduction in the time scale (see Fig. 4b), but also with the changes in the characteristics of the internal variability. While the form of the internal variability in the CS6 runs is largely insensitive to the mountains (cf. Figs. 1f and 6b), the internal variability in the CS2 runs is significantly modified by the introduction of the mountains, as poleward propagation appears as a significant part of the zonal-mean-flow variability (cf. Figs. 1b and 6a). This change is also evident in Figs. 4a and 5a, which show a weak indication of oscillatory behavior at a period of about 150 to 200 days. However, the oscillating behavior in the CS2 M1 run is not as pronounced as that in the CS6 runs since both the zonal index and the poleward propagation coexist.

4. Feedback mechanism

This section addresses the question of how the mountains modify the eddy–mean flow interaction, and the associated eddy feedback, so as to cause the changes in the time scale. We begin this query by considering the wave-breaking properties that are expected to accompany an eddy feedback process for the zonal index in the CS2 M0 run. For the positive zonal-index phase, which corresponds to a poleward shift in the latitude of the jet, the anomalous zonal wind takes on a dipole structure, with the zonal-wind anomaly being positive (negative) on the poleward (equatorward) side of the time-mean jet. Such an anomalous zonal-wind structure must be maintained by an anomalous poleward eddy momentum flux, or equivalently, with eddies that have a northeast–southwest tilt. If wave breaking were taking place, then the wave breaking would be anticyclonic and it would occur on the equatorward side of the jet. This particular meridional tilt and breaking would enhance the anticyclonic shear of the background zonal-mean zonal wind, which would in turn favor the same meridional tilt and wave breaking for subsequent eddies. In this manner, anticyclonic breaking waves can maintain the poleward displaced jet. Similarly, for the negative phase of the zonal index, cyclonically breaking waves would maintain the equatorward displaced jet. Therefore, a sequence of breaking waves, via a feedback process, can maintain the latitudinally displaced jet at its new position.

Following Akahori and Yoden (1997), the above relationship is examined with a scatterplot between the values of PC1 and the WBL. The results, shown in Fig. 7a, indicate that the positive (negative) values of PC1 are associated with lower (higher) values of the WBL, and therefore with anticyclonic (cyclonic) wave breaking. In addition, there is a systematic correspondence in time between individual wave-breaking events and PC1 extrema. This can be seen in Fig. 8, which shows that the movement of the WBL to lower (higher) latitudes is synchronized with the amplification of the positive (negative) phase of PC1. Since the previous results (see section 3a, and also Fig. 3a) suggested that an eddy feedback was maintaining the zonal-index anomaly, these results show that if this eddy feedback is taking place, then it is associated with wave breaking, as discussed in the above paragraph.

In contrast to the CS2 M0 run, we saw for the CS6 M0 run that the zonal-mean anomalies undergo poleward propagation (Figs. 1d, 1e and 1f) and that the eddy feedback is much weaker (Fig. 3b). The mechanism that drives this poleward propagation was recently examined by Lee et al. (2007). With the CS6 M0 run, they showed that the poleward propagation can be caused by a combination of equatorward Rossby wave propagation, multiple anticyclonic wave breakings, and radiative relaxation. This preference for anticyclonic wave breaking being associated with the poleward propagation can be seen in the scatterplot for the CS6 M0 run (Fig. 7b), as the WBL distribution is skewed toward low latitudes where anticyclonic wave breaking dominates. This behavior for the wave breaking strongly contrasts that associated with the zonal-index anomalies in the CS2 M0 run, where as we have seen, the wave breaking appears to be maintaining the zonal-wind anomalies in a fixed position via an eddy feedback. As explained by Lee et al. (2007), these differences between the wave breaking in the CS2 and CS6 M0 runs depend upon the PV gradient. When the PV gradient is strong and sharp, as in the CS2 M0 run (see Fig. 9a), the particle displacements are relatively small, the wave breaking is weak, and the eddy momentum fluxes are confined to a narrow range of latitudes, which results in zonal-mean anomalies that do not propagate meridionally. In contrast, when the PV gradient is relatively weak over a large region, as in the CS6 M0 run (Fig. 9d), the particle displacements are large, the wave breaking is violent, the eddy momentum fluxes extend over a broad range of latitudes, and the zonal-mean anomalies can propagate poleward.

Because poleward propagation would disrupt a positive feedback of the sort that we believe is operating in the CS2 M0 run, it appears then that the properties of the PV gradient are the key to answering the question of why the eddy feedback in the CS2 M0 run is so persistent. The CS2 mountain runs reinforce this clue, since the PV gradient is reduced in these runs (Figs. 9b and 9c), the zonal-mean anomalies propagate poleward (Fig. 6a), and the time scale is lowered (Fig. 4), all characteristics of the CS6 runs.

To further address the influence of the PV gradient, we examine a series of one-point correlation maps (Fig. 9) of the 250-hPa streamfunction field. These correlations can provide information on the dominant linear wave propagation characteristics associated with each particular background flow. As can be seen for the CS2 M0 run (Fig. 9a), the wave tilt is very small in midlatitudes, indicating that little meridional propagation is taking place. On the other hand, for the CS2 M1 run (Figs. 9b and 9c), strong equatorward wave propagation is observed, as in Figs. 9d, 9e and 9f for the CS6 runs. An inspection of Figs. 9b and 9c indicates the presence of broad regions where the PV gradient is weaker, as in the CS6 runs. These results show that the introduction of the mountains to the CS2 run alters the PV gradient in a manner that favors a strong meridional component to the wave propagation. These changes in the wave propagation characteristics disrupt the eddy feedback, drive the zonal-mean anomalies poleward, and thus reduce the time scale.

Additional insight into the changes in the wave propagation characteristics can be gained by considering the refractive index, , defined by (4). As is well known from linear Rossby wave theory, Rossby waves are reflected at turning latitudes, where = 0, and propagate toward critical latitudes, where = ∞ (in practice, critical latitudes are identified by very large values of ). For the runs without mountains, the refractive index calculation is based on a Wentzel–Kramers–Brillouin (WKB) assumption that the local meridional scale of the waves is substantially shorter than that of the background flow. When topography is included, the WKB assumption is that this relationship between the scale of the waves and background flow extends to the zonal direction. The applicability of this assumption appears to be weakest for the M1 runs in the zonal direction. Even though the WKB assumption may be not valid in the strictest sense, as we will see, the calculated values of are consistent with the characteristics of wave propagation.

The values of the refractive index for various runs are illustrated in Fig. 10. In the CS2 M0 run, the eddies are bounded by turning latitudes on both the poleward and equatorward sides of the jet (Fig. 10a).7 Consistently, since the presence of a turning latitude allows for the occurrence of both incident reflected waves, the meridional wave propagation is limited. Figure 10b shows that the introduction of a mountain drastically alters , particularly on the equatorward side of the jet, as turning latitudes occurs directly to the south of the mountain, and critical latitudes are present both farther upstream and downstream. The reflection of waves away from the turning latitudes and toward the critical latitudes can be seen in Figs. 9b and 9c, respectively. Similar refractive index (Figs. 10c and 10d) and wave propagation (Figs. 9d and 9e) characteristics can be seen in the CS6 runs. In their study on poleward zonal-mean anomaly propagation, Lee et al. (2007) also found that the poleward propagation is intimately tied to the presence of critical latitudes and very strong wave breaking. These results provide further support to the idea that the introduction of the mountain in the CS2 M1 run has altered the background flow in a manner that resembles that for the CS6 runs. This in turn allow for equatorward wave propagation, poleward zonal-mean anomaly propagation, and a reduction in the time scale.

Unlike for the CS2 runs, as we have seen, the waves in the CS6 runs are only weakly affected by the mountains (Figs. 9e and 9f). The underlying reason for this weak sensitivity to the presence of the mountains appears to be that the surface winds (not shown) are much weaker in the CS6 runs than in the CS2 runs. Therefore, given identical topography, barring any resonance, the resulting stationary wave response would be expected to be weaker in the CS6 runs than in the CS2 runs. This is borne out in the field (cf. Figs. 9d and 9e), where the changes in due to the mountain are not sufficiently strong to substantially alter the wave propagation characteristics.

5. Sensitivity runs

In this section, we examine the sensitivity of the model results to both mountain height and model resolution.

a. Mountain height

The height of the mountain in the CS2 M1 and the CS6 M1 runs is systematically varied from 1 to 5 km (Fig. 11a). For the CS6 runs, the value of ε (the solid line in Fig. 11a) is essentially independent of the mountain height. This result is not surprising, because as discussed earlier, the weak surface winds in the CS6 runs do not allow for a significant response in the topographically forced stationary wave. In contrast, the values of ε in the CS2 runs (the dashed line in Fig. 11a) show a dramatic sensitivity to the mountain height. For mountain heights less than or equal to 4 km, the value of ε is very similar. However, when the mountain height is increased to 5 km, the value of ε increases to 105 days, about 2.5 times that of the 4-km run!

We next address the question of whether the explanation given in the previous section for the CS2 M0 run can account for the even longer time scale of the 5-km CS2 M1 run. It is first seen that the zonal-mean-flow anomaly for the 5-km CS2 M1 run is dominated by zonal-index behavior (Fig. 12a), as in the CS2 M0 run. Figure 13 exhibits the 250-hPa eddy momentum flux convergence for all CS2 runs with differing mountain heights. For mountain heights less than or equal to 4 km, the eddy momentum flux convergence is dominated by the transient eddy contribution. In contrast, for the 5-km mountain run, it is the eddy momentum flux convergence associated with the stationary eddies that is larger. This strong eddy momentum flux convergence by the stationary eddies drives a strong and sharp westerly jet (Fig. 12a), and therefore also a sharp field (Fig. 14a). Furthermore, over a wide range of longitudes, the jet is bounded on both sides by turning latitudes (Fig. 14c), indicating that the waves must be trapped at the majority of longitudes. This wave trapping can be seen in the one-point correlation map illustrated in Fig. 14a. Also, upstream of the mountain, where there are no turning latitudes on the equatorward side of the jet (Fig. 14c), equatorward wave propagation takes place (Fig. 14b). All of these results show that when the mountain height is raised to 5 km, the impact of the stationary eddies is to restore the background flow toward a state that has much more in common with the very persistent CS2 M0 run. One primary difference though, is that the stronger field and the tighter trapping of the eddies, as compared to the CS2 M0 run, appears to be able to account for the even longer time scale in the 5-km CS2 M1 run.

b. Model resolution

According to the picture that wave breaking (and therefore eddy momentum flux convergence) is important for determining the time scale, if the largest wavenumber allowed in the model is insufficient to resolve the relevant details of the wave breaking, then the zonal-mean anomaly may be artificially prolonged. To test this possibility, the zonal resolution of the model is varied. The sensitivity of ε to the model resolution is summarized in Fig. 11b. (It should be stated that the background fields and EOFs in these sensitivity runs are almost indistinguishable from those in the base runs.) It is seen that except for the runs in which poleward propagation dominates, the value of ε is found to be about 25% larger if the maximum resolvable zonal wavenumber is reduced to 15, while retaining the same maximum meridional wavenumber. However, when the maximum zonal wavenumber is increased to 60 for the CS2 run (due to computational burden, this high-resolution integration is performed only for the CS2 and CS6 cases), the value of ε is only slightly reduced. This suggests that the rhomboidal 30 resolution in this study does not cause a large misrepresentation of the time scale, and that the time scales found in our simulations must be very close to the corresponding asymptotic values.

6. Conclusions and discussion

A series of model simulations show that the persistence time scale of the dominant pattern of zonal-mean-flow variability is highly dependent upon the nature of the variability. The time scale is much greater when the variability takes the form of the zonal index than it is for poleward propagation. As shown by SL06, the form of the zonal-mean-flow variability is closely associated with the properties of the time-mean PV gradient, as zonal-index (poleward propagation) behavior was found to occur if the time-mean PV gradient is relatively strong and sharp (weak and broad). In this study, it was found that strong and sharp PV gradients coincided with a positive feedback between the eddies and zonal-mean flow. Further analysis showed that the eddies have little meridional tilt and propagate mostly in the zonal direction. As a result, the eddy momentum flux convergence is confined to midlatitudes, a feature that is amenable to a positive eddy feedback and long time scales. In contrast, when the PV gradient was weak and broad, the eddies exhibited a strong meridional tilt and propagated equatorward. This resulted in an eddy momentum flux convergence that extended over a wide range of latitudes, allowing the zonal-mean anomalies to propagate from the tropics to high latitudes.

In this study, the strong positive feedback was attributed to wave breaking. Other studies of zonal-mean-flow variability have also proposed mechanisms for a positive feedback. For example, Lorenz and Hartmann (2003) indicated that modification of the refractive index can aid the feedback. Gerber and Vallis (2007) discussed the possible role played by the mechanism proposed by Robinson (2006) who showed that the mean meridional circulation driven by the upper-level PV flux can maintain the midlatitude baroclinicity. Our wave-breaking argument is neither inconsistent nor mutually exclusive of these mechanisms. As discussed in section 4, the wave propagation and breaking are closely related with the refractive index of the background flow. In addition, the occurrence of multiple life cycles, as evidenced by the multiple peaks in Fig. 8 near the lag 0 day, suggest that baroclinic eddy activity is often recurrent although it is unclear if Robinson’s mechanism is the cause.

Since the time scale of the zonal index in the control simulations was found to be about 55 days, much longer than the approximately 10-day time scale observed in the atmosphere, idealized mountains were introduced to the model to investigate whether this very long time scale is related to the absence of asymmetries in the background flow. Two model runs were examined, one with strong tropical heating and a narrow extratropical baroclinic zone, and the other with weak tropical heating and a broad extratropical baroclinic zone. These runs were referred to as CS2 M1 and CS6 M1, respectively. In the absence of the mountain, the CS2 run was characterized by zonal-index behavior and the CS6 run by poleward propagation. For the CS6 M1 run, very little change in the zonal-mean anomalies was observed. This was because the surface winds were weak, generating weak stationary wave response. As a result, the topography made no appreciable impact on the time-mean PV gradient, and thus the zonal-mean flow variability. In contrast, for the CS2 M1 run, the addition of the mountain resulted in a weakening of the time-mean PV gradient, poleward propagation, and a reduction in the time scale to about 30 days. These results suggest that the reduction in the time scale in the CS2 M1 run is in part because the topographically excited stationary eddies reduce the PV gradient, which, by allowing for equatorward wave propagation, churn poleward zonal-mean anomaly propagation.

Although we have shown that topography can influence the time scale of the zonal index by modifying the local PV gradient, there are other processes through which topography can also affect the time scale. In two very recent studies, Kornich et al. (2006) and Gerber and Vallis (2007) found that the introduction of topography to a general circulation model reduced the time scale by more than one-half. With regard to a physical mechanism, Gerber (2005) and Gerber and Vallis (2007) suggested that the presence of topography reduces the time scale by disconnecting the eddy activity on one side of the topography from that on the other side. Although the CS2 5-km mountain run, in which time scale is much longer than that in the corresponding no-mountain run, provides a counter example, the mechanism of Gerber (2005) may still explain our finding that the CS2 M1 run has a shorter time scale than that in the CS6 M0 run (Fig. 4b). In addition to the topographically forced stationary waves, processes involving the stratosphere (Baldwin et al. 2003; Norton 2003; Kushner and Polvani 2004) or external forcing that happen to project onto the zonal index (Son and Lee 2006; Ring and Plumb 2007) can also change the time scale.

Because the stationary waves in the Southern Hemisphere (SH) are much weaker than those in the Northern Hemisphere (NH), one might expect that the time scale in the SH would be much greater than that in the NH. However, as Baldwin et al. (2003) indicated, the observed time scales of the annular mode in the two hemispheres are comparable with each other. One possible explanation for this apparent paradox is that the time scale is insensitive to the detailed structure of the topography. To test this possibility, additional simulations are performed. In one set of simulations, with each set comprising the CS2 and CS6 runs, the mountain in the M1 run was placed at 35°N instead of at 45°N. In the second set of simulations, the mountain was placed at 45°N, but the width of the mountain reduced by one-half. The resulting time scales for the CS2 (CS6) runs are 44 (46) days for the former and 40 (38) days for the latter experiments. These time scales are not dramatically different from those in the other single mountain runs. In addition, it was found that the time scales are not substantially different between the one and two mountain runs, and between the 1-, 2-, 3-, and 4-km mountain runs. Thus, we conclude that the time scale is not appreciably sensitive to the detailed structure of the topography, unless the topography is sufficiently high to change the property of the background flow (e.g., CS2 5-km M1 run). It is also important to recall that the topography is not the only source of zonal asymmetry in the time-mean flow. Stationary eddies, which are generated by heating, can also weaken the local PV gradient. Given the presence of these processes, together with our finding that the time scale is relatively insensitive to the detailed structure of the topography, it seems reasonable to conclude that the zonal asymmetries present in the SH are strong enough to shorten the time scale by an amount that is comparable to that in the NH.

The results of this study may also be useful for understanding other modeling results if the sharpness of the PV gradient is the primary factor in determining the characteristics of the zonal-mean-flow variability. One example of the impact of a sharp PV gradient is based on the recent model results of Gerber (2005), who showed that the time scale of the zonal-index increases in response to finer vertical resolution in the upper troposphere. This finding is not inconsistent with our conclusion, because the background PV field is found to be sharper with higher vertical resolution. By the same token, the addition of a feature that acts to diffuse the background PV field can shorten the time scale.

An important question that was not addressed in this study is what processes determine the time scale for the decay of the zonal index. The decay of the zonal index has often been attributed to low-frequency nonlinear eddy fluxes (Feldstein and Lee 1998; Shiogama et al. 2005). Benedict et al. (2004) showed that these low-frequency fluxes correspond to potential temperature mixing that follows wave breaking, implying that the decay time scale is closely linked to eddy mixing. Benedict et al. (2004) also found that prolonged NAO events take place when a sequence of synoptic-scale disturbances occur upstream of the NAO region. To the extent that the zonal index corresponds to a local, zonally oriented dipole pattern (Cash et al. 2002; Gerber and Vallis 2005), the time scale for the decay of the zonal index is also determined by the presence and location of synoptic-scale disturbances relative to the zonal-index anomaly. Furthermore, as described above, the anomaly decay can also arise from the meridional propagation of Rossby waves. Therefore, multiple factors may contribute toward determining the time scale of the zonal-index decay.

Acknowledgments

We thank Dr. Edwin Gerber for helpful discussion. This research was supported by the National Science Foundation under Grants ATM-0324908, ATM-0351044, and ATM-0649512.

REFERENCES

  • Akahori, K., , and S. Yoden, 1997: Zonal flow vacillation and bimodality of baroclinic eddy life cycles in a simple global circulation model. J. Atmos. Sci., 54 , 23492361.

    • Search Google Scholar
    • Export Citation
  • Ambaum, M. H., , and B. J. Hoskins, 2002: The NAO troposphere–stratosphere connection. J. Climate, 15 , 19691978.

  • Andrews, D. G., , J. R. Holton, , and C. B. Leovy, 1987: Middle Atmosphere Dynamics. Academic Press, 489 pp.

  • Baldwin, M. P., , D. B. Stephenson, , D. W. J. Thompson, , T. J. Dunkerton, , A. J. Charlton, , and A. O’Nell, 2003: Stratospheric memory and extended-range weather forecasts. Science, 301 , 636640.

    • Search Google Scholar
    • Export Citation
  • Barnston, A. G., , and R. E. Livezey, 1987: Classification, seasonality, and persistence of low-frequency atmospheric circulation patterns. Mon. Wea. Rev., 115 , 10831126.

    • Search Google Scholar
    • Export Citation
  • Benedict, J. J., , S. Lee, , and S. B. Feldstein, 2004: Synoptic view of the North Atlantic oscillation. J. Atmos. Sci., 61 , 121144.

  • Cash, B. A., , P. J. Kushner, , and G. K. Vallis, 2002: The structure and composition of the annular modes in an aquaplanet general circulation model. J. Atmos. Sci., 59 , 33993414.

    • Search Google Scholar
    • Export Citation
  • Chang, E. K. M., , and D. B. Yu, 1999: Characteristics of wave packets in the upper troposhere. Part I: Northern Hemisphere winter. J. Atmos. Sci., 56 , 17081728.

    • Search Google Scholar
    • Export Citation
  • Feldstein, S. B., 1998: An observational study of the intraseasonal poleward propagation of zonal mean flow anomalies. J. Atmos. Sci., 55 , 25162529.

    • Search Google Scholar
    • Export Citation
  • Feldstein, S. B., 2000a: Is interannual zonal mean flow variability simply climate noise? J. Climate, 13 , 23562362.

  • Feldstein, S. B., 2000b: The timescale, power spectra, and climate noise properties of teleconnection patterns. J. Climate, 13 , 44304440.

    • Search Google Scholar
    • Export Citation
  • Feldstein, S. B., , and S. Lee, 1998: Is the atmospheric zonal index driven by an eddy feedback? J. Atmos. Sci., 55 , 30773086.

  • Franzke, C., , S. Lee, , and S. B. Feldstein, 2004: Is the North Atlantic Oscillation a breaking wave? J. Atmos. Sci., 61 , 145160.

  • Gerber, E., 2005: A dynamical and statistical understanding of the NAO and annular modes. Ph.D. thesis, Princeton University, 242 pp.

  • Gerber, E., , and G. K. Vallis, 2005: A stochastic model for the spatial structure of annular patterns of variability and the North Atlantic Oscillation. J. Climate, 18 , 21022118.

    • Search Google Scholar
    • Export Citation
  • Gerber, E., , and G. K. Vallis, 2007: Eddy–zonal flow interactions and the persistence of the zonal index. J. Atmos. Sci., 64 , 32963311.

    • Search Google Scholar
    • Export Citation
  • Hurrell, J. W., , Y. Kushnir, , M. Visbeck, , and G. Ottersen, 2003: An overview of the North Atlantic Oscillation. The North Atlantic Oscillation: Climate Significance and Environmental Impact, Geophys. Monogr., Vol. 134, Amer. Geophys. Union, 1–35.

    • Search Google Scholar
    • Export Citation
  • Kornich, H., , G. Schmitz, , and E. Becker, 2006: The role of stationary waves in the maintenance of the northern annular mode as deduced from model experiments. J. Atmos. Sci., 63 , 29312947.

    • Search Google Scholar
    • Export Citation
  • Kravtsov, S., , A. W. Robertson, , and M. Ghil, 2006: Multiple regimes and low-frequency oscillations in the Northern Hemisphere’s zonal-mean flow. J. Atmos. Sci., 63 , 840860.

    • Search Google Scholar
    • Export Citation
  • Kushner, P. J., , and L. M. Polvani, 2004: Stratosphere–troposphere coupling in a relatively simple AGCM: The role of eddies. J. Climate, 17 , 629639.

    • Search Google Scholar
    • Export Citation
  • Lee, S., , and S. B. Feldstein, 1996: Two types of wave breaking in an aquaplanet GCM. J. Atmos. Sci., 53 , 842857.

  • Lee, S., , S-W. Son, , K. Grise, , and S. B. Feldstein, 2007: Mechanism of poleward propagation of zonal mean flow anomalies. J. Atmos. Sci., 64 , 849868.

    • Search Google Scholar
    • Export Citation
  • Lorenz, D. J., , and D. L. Hartmann, 2001: Eddy–zonal flow feedback in the Southern Hemisphere. J. Atmos. Sci., 58 , 33123327.

  • Lorenz, D. J., , and D. L. Hartmann, 2003: Eddy–zonal flow feedback in the Northern Hemisphere winter. J. Climate, 16 , 12121227.

  • Norton, W. A., 2003: Sensitivity of Northern Hemisphere surface climate to simulation of the stratospheric polar vortex. Geophys. Res. Lett., 30 .1627, doi:10.1029/2003GL016958.

    • Search Google Scholar
    • Export Citation
  • Oort, A. H., , and J. J. Yienger, 1996: Observed interannual variability in the Hadley circulation and its connection to ENSO. J. Climate, 9 , 27512767.

    • Search Google Scholar
    • Export Citation
  • Orlanski, I., 2003: Bifurcation in eddy life cycles: Implications for storm track variability. J. Atmos. Sci., 60 , 9931023.

  • Riehl, H., , T. C. Yeh, , and N. E. La Seur, 1950: A study of variations of the general circulation. J. Meteor., 7 , 181194.

  • Ring, M. J., , and R. A. Plumb, 2007: Forced annular mode patterns in a simple atmospheric general circulation model. J. Atmos. Sci., 64 , 36113626.

    • Search Google Scholar
    • Export Citation
  • Robinson, W. A., 1991: The dynamics of the zonal index in a simple model of the atmosphere. Tellus, 43A , 295305.

  • Robinson, W. A., 2006: On the self-maintenance of midlatitude jets. J. Atmos. Sci., 63 , 21092122.

  • Shiogama, H., , T. Terao, , H. Kida, , and T. Iwashima, 2005: Roles of low- and high-frequency eddies in the transitional process of the Southern Hemisphere annular mode. J. Climate, 18 , 782794.

    • Search Google Scholar
    • Export Citation
  • Son, S-W., , and S. Lee, 2005: The response of westerly jets to thermal driving in a primitive equation model. J. Atmos. Sci., 62 , 37413757.

    • Search Google Scholar
    • Export Citation
  • Son, S-W., , and S. Lee, 2006: Preferred modes of variability and their relationship with climate change. J. Climate, 19 , 20632075.

  • Thompson, D. W. J., , and J. M. Wallace, 1998: The Arctic Oscillation signature in the wintertime geopotential height and temperature fields. Geophys. Res. Lett., 25 , 12971300.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W. J., , 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
  • Thorncroft, C. D., , B. J. Hoskins, , and M. E. McIntyre, 1993: Two paradigms of baroclinic life-cycle behavior. Quart. J. Roy. Meteor. Soc., 119 , 1755.

    • Search Google Scholar
    • Export Citation
  • Vallis, G. K., , E. P. Gerber, , P. J. Kushner, , and B. A. Cash, 2004: A mechanism and simple dynamical model of the North Atlantic Oscillation and annular modes. J. Atmos. Sci., 61 , 264280.

    • Search Google Scholar
    • Export Citation
  • Wallace, J. M., , and D. S. Gutzler, 1981: Teleconnections in the geopotential height field during the Northern Hemisphere winter. Mon. Wea. Rev., 109 , 784812.

    • Search Google Scholar
    • Export Citation
  • Yu, J-Y., , and D. L. Hartmann, 1993: Zonal-flow vacillation and eddy forcing in a simple GCM of the atmosphere. J. Atmos. Sci., 50 , 32443259.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

(a)–(f) Linear regression of 〈[u]〉′ at all latitudes against 〈[u]〉′ at a base latitude of 30°. All six M0 runs are illustrated. The contour interval is 0.2 m s−1. Solid contours are positive, dashed contours negative, and the zero contour is omitted. The regressed 〈[u]〉′ values, which are statistically significant at the 99% confidence level, are shaded. The right panel of each frame shows the latitudinal profiles of the time-mean 〈[u]〉 (thick gray line), EOF1 (thin solid line), and EOF2 (thin dashed line). The location of polar front or eddy-driven jet is denoted with a straight line.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 2.
Fig. 2.

(a) Autocorrelation coefficients for PC1 in all six M0 runs, and (b) their corresponding e-folding time scale (ε). In (a), the poleward propagation cases, CS5 and CS6, are discriminated from the other cases with a dashed line. In (b), the upper and lower limits of the error bars correspond to the values of ε in each hemisphere, and open circles indicate an average.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 3.
Fig. 3.

Lag-correlation coefficients between F1 and PC1 for the (a) CS2 and (b) CS6 M0 runs. Negative time lag denotes that F1 leads PC1. The values that are statistically significant at the 99% (90%) confidence level are indicated by dark (light) gray. The number within each panel denotes the time scale of the positive feedback, which is defined as the maximum lag day for which the correlation coefficient is statistically significant at the 90% confidence level.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 4.
Fig. 4.

(a) Autocorrelation coefficients for PC1 in all six M1 runs, and (b) the values of the e-folding time scales in all M0 (circle), M1 (triangle), and M2 runs (square).

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 5.
Fig. 5.

Lag-correlation coefficients between F1 and PC1 for (a)–(c) the CS2 and (d)–(f) CS6 M1 runs; (a), (d) total forcing, (b), (e) momentum flux forcing, and (c), (f) mountain torque forcing.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 6.
Fig. 6.

As in Figs. 1b and 1f, but for the M1 runs.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 7.
Fig. 7.

Scatterplots of the WBL and PC1 for the (a) CS2 and (b) CS6 M0 runs. The probability density functions are illustrated at each axis. White crosses in the figure denote the location of jets. Two crosses in (b) correspond to subtropical and eddy-driven jets, respectively. Note that the sign of PC1 in (b) is reversed in order that the sign of the midlatitude EOF1 in the CS6 M0 run matches with that in the CS2 M0 run. While the positive phase of EOF1 in the CS2 M0 run corresponds to positive 〈[u]〉′ at 50° (Fig. 1b), that in the CS6 M0 run corresponds to negative 〈[u]〉′ (Fig. 1f).

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 8.
Fig. 8.

Composite time series of (a) PC1 and (b) WBL in the CS2 M0 run. The value of PC1 is normalized by 1 std dev. The composite fields are constructed with respect to the PC1 extrema. A total of 329 positive and 333 negative cases, in which the local extrema are greater or less than 1 std dev of PC1 time series, are utilized. Positive (negative) cases are indicated with solid (dashed) lines. Both PC1 and WBL are slightly smoothed with 10-day low-pass filter.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 9.
Fig. 9.

One-point correlation maps of 250-hPa eddy streamfunction in the (a) CS2 M0, (b), (c) CS2 M1, (d) CS6 M0, and (e), (f) CS6 M1 runs. The contour interval is 0.2. Solid contours are positive, dashed contours negative, and the zero contour is omitted. The shading and thick cross symbol denote the mountain and the base point for the correlation, respectively. The superimposed thin contours are the time-mean qy at 250 hPa. The contour interval is 2.5 × 10−11 m−1 s−1 and zero values are denoted with a thick gray line.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 10.
Fig. 10.

The values of the refractive index for the (a) CS2 M0, (b) CS2 M1, (c) CS6 M0, and (d) CS6 M1 runs. The values are slightly smoothed with a nine-point smoother. The contour interval is 2 × 10−7 m−2. The superimposed shading denotes the mountain. The insets in (a) and (c) show the refractive index for the zonal-mean flow from equator to 60°N. The values displayed in the insets range from −1 × 10−9 to 1 × 10−9 m−2.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 11.
Fig. 11.

The e-folding time scale as a function of (a) mountain height and (b) model resolution. In (a), the M0 and 2-km M1 runs are marked with a circle and triangle, respectively. In (b), only the zonal resolution is varied, while the maximum meridional wavenumber is fixed at 30. A maximum zonal wavenumber of 15 (60) is denoted by K15 (K60).

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 12.
Fig. 12.

As in Figs. 1b and 1f, but for the 5-km M1 runs.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 13.
Fig. 13.

The contribution by stationary (long-dashed lines) and transient (short-dashed lines) eddies to the total (solid lines) zonal-mean eddy momentum flux convergence at 250 hPa in the (a) CS2 M0 and (b)–(f) the CS2 M1 runs with varying height. The mountain height in each run is denoted at the top of each panel.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Fig. 14.
Fig. 14.

Various quantities from the CS2 5-km M1 run. (a), (b) One-point correlation maps of 250-hPa eddy streamfunction with the base point denoted by a strong cross mark and (c) the refractive index. The shading and contour intervals are the same as in Figs. 9 and 10.

Citation: Journal of the Atmospheric Sciences 65, 3; 10.1175/2007JAS2380.1

Table 1.

Anomalous thermal forcing (climate state) utilized in each experiment set. The values of H and C, respectively, are the additional equatorial heating and polar cooling in ΔTe measured at the model surface. Units are K day−1.

Table 1.
Table 2.

Percentage variance explained by EOF1/EOF2 in (top) M0, M1, and M2 runs, and (bottom) CS2 and CS6 M1 runs with varying mountain height.

Table 2.

1

Following Thorncroft et al. (1993), anticyclonic (cyclonic) wave breaking is characterized by southwest–northeast (southeast–northwest) tilt of the trough/ridge pair.

2

In SL05 and SL06, the zonal resolution is truncated at wavenumber 15, while the meridional resolution is truncated at wavenumber 30. This is the only difference between the base model simulations in this study and those in SL05 and SL06.

3

For the mountain simulations, Te in (1) is interpolated onto sigma coordinates, assuming that the surface pressure is 1000 hPa everywhere except over the topography where surface pressure values are estimated from the hypsometric equation.

4

This relationship can be physically understood from the NAO study of, for example, Benedict et al. (2004). The NAO is first driven by the transient eddy vorticity fluxes, F1. Therefore, in order for the NAO to be driven by eddy fluxes, at negative lags, the correlation between F1 and PC1 must be positive. After the NAO pattern is established, Benedict et al. (2004) find that the NAO is maintained by eastward-propagating synoptic-scale disturbances that are located upstream of the NAO region. As these disturbances reach the NAO region, their spatial structure is modified in a manner that reinforces the NAO pattern. As a result, if there is a positive feedback taking place, for positive lags, the correlation must also be positive. This reversal in the lag–lead relationship implies that there should be a local minimum correlation near the lag 0 day, which is indeed observed in Fig. 5 of Lorenz and Hartmann (2001) and Fig. 3a of this study.

5

Qualitatively, the results are insensitive to the choice of isentrope. Although not shown, almost identical results are found when PV320K is used.

6

Since Fig. 5 shows correlation coefficients, the sum of the correlations in the middle and the bottom panels is not equal to that in the top panel.

7

The turning latitude on the equatorward side of the jet arises from the weak PV gradient in low latitudes, which leads to a small value for the first term on the right-hand side of (4).

Save