Reconciling conflicting interpretations of the forcing of the tropical upwelling in the Brewer- Dobson circulation

Transcription

1 Reconciling conflicting interpretations of the forcing of the tropical upwelling in the Brewer- Dobson circulation Rei Ueyama Earth Science Division, NASA Ames Research Center, Moffett Field, CA Edwin P. Gerber Center for Atmosphere Ocean Science, Courant Institute of Mathematical Sciences, New York University, New York, NY John M. Wallace Department of Atmospheric Sciences, University of Washington, Seattle, WA Dargan M. W. Frierson Department of Atmospheric Sciences, University of Washington, Seattle, WA Submitted to Journal of the Atmospheric Sciences June 0 Corresponding author address: Rei Ueyama, Earth Science Division, NASA Ames Research Center, Mail stop 4-, Moffett Field, CA rei.ueyama@nasa.gov

2 Abstract The forcing of tropical upwelling in the Brewer-Dobson circulation (BDC) on intraseasonal time scales is investigated in integrations of an idealized general circulation model, ECMWF 40-year Reanalysis (ERA-40), and lower-stratospheric temperature measurements from the (Advanced) Microwave Sounding Unit, with a focus on the extended boreal winter season. Enhanced poleward eddy heat fluxes in the high latitudes (4-90 N) at the 00 hpa level are associated with anomalous tropical cooling and anomalous warming on the poleward side of the polar night jet at the hpa level and above. In both the model and the observations, planetary waves entering the stratosphere at high latitudes propagate equatorward to the subtropics and outer tropics at levels above hpa over a ~0 day period, exerting a forcing at latitudes low enough to effectively modulate the tropical upwelling in the deep branch of the BDC. The influence of planetary waves originating in high latitudes upon the strength of the tropical upwelling 4 diminishes rapidly with depth below hpa. Intraseasonal variations in tropical temperature tendency at 00 hpa are not strongly correlated with those at hpa and above or with variations in high-latitude planetary wave activity, suggesting that variations in tropical upwelling below hpa could be primarily controlled by variations in the strength of the shallow branch of the BDC.

3 3 9. Introduction The Lagrangian mean meridional circulation that ventilates the stratosphere, commonly referred to as the Brewer-Dobson circulation (BDC), is marked by lowlatitude ascent and high-latitude descent (Brewer 949; Dobson 96; Andrews et al. 987). It is mechanically forced and operates in accordance with the downward control principle stated in Haynes and McIntyre (987) and articulated in detail by Haynes et al. (99) and Holton et al. (99). In the ascending branch of the BDC, adiabatic cooling maintains lower-stratospheric temperatures well below radiative equilibrium values (Andrews et al. 987; Gettelman et al. 004; Corti et al. 00). The BDC is chiefly driven by the dissipation of vertically-propagating Rossby and gravity waves originating from the extratropical winter hemisphere (Garcia 987; Holton 990; Haynes et al. 99; Iwasaki 99). In all seasons, synoptic-scale baroclinic waves that originate in the midlatitudes and break just above the subtropical tropospheric jets also exert a westward torque on the subtropical lower stratosphere, driving poleward mass flux in the lower stratosphere in both hemispheres (Held and Hoskins 98; Plumb 00; Shepherd 007). Thus the BDC can be viewed as consisting of two distinct branches (e.g., as depicted in Fig. of Plumb 00 and Fig. 3.4 of Vallis 006): a deep (or upper) branch driven by high-latitude wintertime planetary-scale waves, and a shallow (or lower) branch driven by midlatitude baroclinic synoptic-scale waves. Based on the analyses of the residual circulation trajectories, Birner and Bönisch (0) showed that the two branches of the stratospheric residual circulation are distinguishable in terms of their aspect ratio, the latitude of stratospheric entry of tropical tropospheric air, transit

4 times of air traveling from the tropical tropopause to the lower stratosphere, and the integrated mass flux along the trajectories. The shallow branch of the BDC, which coincides with the uppermost layer of the tropospheric Hadley cell, may be interpreted as the wave-driven portion of the Hadley cell. The strength of the upwelling in the tropical lower stratosphere varies in response to sporadic bursts of planetary wave activity entering the poleward stratosphere from below, as evidenced by the empirical studies of Iwasaki (99), Yulaeva et al. (994), Randel et al. (00a,b), Salby and Callaghan (00), Dhomse et al. (008) and Ueyama and Wallace (00). The breaking of these waves near the polar night jet in the stratosphere evidently drives the deep branch of the BDC. Downward control diagnostics by Rosenlof (99), Plumb and Eluszkiewicz (999), Semeniuk and Shepherd (00), Scott (00), Boehm and Lee (003), Zhou et al. (006), Kerr-Munslow and Norton (006), Geller et al. (008), Deckert and Dameris (008), Rosenlof and Reid (008), Ryu and Lee (00), Chen and Sun (0), Garny et al. (0) and Zhou et al. (0) have shown unequivocally that extratropical wavebreaking does not generate mean meridional circulations that extend across the tropics. To force a robust tropical response requires wave breaking extending into the outer tropics or subtropics. It has yet to be clearly demonstrated that high-latitude planetary waves are capable of penetrating far enough out to the subtropics and tropics to effectively force tropical upwelling in the deep branch of the BDC. The absence of a clear connection between the breaking of high-latitude planetary waves and the forcing of tropical upwelling as manifested in the downward control diagnostics has led to suggestions that lower-stratospheric temperature is modulated

5 primarily by the eddy forcing of the shallow branch of the BDC. Kerr-Munslow and Norton (006) and Norton (006) argued that the annual cycle in tropical cold point temperature is driven by seasonal variations in the strength of the equatorial planetary waves, which break in the outer tropics, inducing equatorial upwelling at the cold point tropopause around ~90 hpa level. Boehm and Lee (003), Randel et al. (008), Deckert and Dameris (008), Rosenlof and Reid (008), Ryu and Lee (00) and Garny et al. (0) have similarly proposed the link between tropical upwelling and convectivelydriven equatorial planetary waves. Studies of Rosenlof (99), Plumb and Eluszkiewicz (999), Semeniuk and Shepherd (00), Scott (00), Zhou et al. (006), Randel et al. (008), Geller et al. (008), Chen and Sun (0), Garny et al. (0) and Zhou et al. (0) have emphasized the breaking of midlatitude, synoptic-scale baroclinic waves in forcing the shallow branch of the BDC. To reach closure on the relative importance of high-latitude planetary waves, the midlatitude synoptic-scale baroclinic waves, and equatorial planetary waves in forcing seasonal and nonseasonal variability in the strength of the BDC, it will be necessary to determine the extent to which the breaking of high-latitude planetary waves extends into the subtropics and outer tropics. A major obstacle is the limited ability of the reanalysis datasets to resolve the temporal variations in tropical upwelling and the wave breaking events that give rise to them. In this study, we strive to overcome these limitations by analyzing intraseasonal (i.e., day-to-day) variations in the time rate of change of zonallyaveraged temperature ( [T]/ t), a reliable proxy for the rate of ascent in the tropical lower stratosphere. As a guide in interpreting our observational results, we rely on an extended integration of an idealized general circulation model that produces a realistic simulation

6 of the high-latitude planetary waves and their role in forcing the ascending branch of the BDC. In carrying out this analysis, we will distinguish between the forcing of the tropical upwelling at ~ hpa and above, which mainly involves the deep branch of the BDC, and the upwelling in the vicinity of the tropical cold point tropopause near 00 hpa, which is more strongly influenced by the shallow branch of the BDC. Section provides the details of the model and observational data sets used, and the methods that are employed. Section 3 describes the structure of the tropical upwelling in the BDC as simulated in the model and in observations. The role of highlatitude wintertime planetary waves in forcing the variability in the tropical upwelling is explored in Section 4, followed by a discussion of the contrasting variability in tropical upwelling at the and 00 hpa levels in Section. Our results are summarized and interpreted in the context of previous work in Section Data and methods This study compares the results of an idealized general circulation model and two sets of observation-based data. We use the pseudo-spectral, dry dynamical core model described in Held and Suarez (994) and modified to capture the stratospheric circulation as in Polvani and Kushner (00). The integrations are run with a vortex lapse rate ϒ = 4 K km - at triangular truncation 4 (T4; Gaussian grid of ~.8 ) resolution and 40 vertical σ (= pressure / surface pressure) levels that are spaced approximately evenly in log pressure height (i.e., 0.0,... 0, 6, 80, 00,, hpa). A large-scale surface topography of wavenumber k = and amplitude h 0 = 3 km is added in the winter hemisphere between and 6 N, centered at 4 N. These parameter settings have been

7 shown to produce realistic troposphere-stratosphere coupling (Gerber and Polvani 009) and a realistic BDC (Gerber 0), and the reader is referred to these papers for a detailed description. We analyze a 0,000 day, perpetual January integration after a 300- day spin-up period. The global European Centre for Medium-Range Weather Forecasts 40-year Reanalysis (ERA-40: Uppala et al. 00) provided by the Research Data Archive at the National Center for Atmospheric Research (data set number ds7. at are analyzed for the 3-year period from Jan 979 to 3 Dec 00. The data are gridded at approximately. latitude by. longitude from 89. N to 89. S and 3 pressure levels from to 000 hpa (e.g.,,, 3,... 0,, 00, 0, 00, hpa). Except where specifically indicated, the study focuses on the nonseasonal variability defined as departures from the seasonally-varying climatological mean computed from climatological daily means (i.e., Jan , Jan ,..., 3 Dec ) from November through March. February 9 data of the leap years have been omitted. The daily brightness temperature (T) fields derived from the lower-stratospheric channel of Version 3. Microwave Sounding Unit / Advanced Microwave Sounding Unit carried aboard National Oceanic and Atmospheric Administration satellites (hereafter, MSU-4) are also analyzed for the same period of record, November through March The data are gridded at. x. resolution and extend to 8. N/S. The weighting function of the lower-stratospheric channel is concentrated mainly in the to 9 km (30 to 0 hpa) layer, as detailed in the Remote Sensing Systems Web site (see

8 Zonal averaging is denoted by square brackets [ ] and primes denote departures from the zonal mean. The poleward eddy heat flux term is computed by averaging the daily products of the meridional wind v' and temperature T' anomalies around latitude circles and subtracting the climatological daily means at each latitude and pressure level (i.e., [v't'] minus the climatological daily mean value of [v't']). Time derivative fields of temperature [T]/ t are estimated as centered differences. For example, what we refer to as a -day time rate of change is computed by subtracting the value for the previous day from the value for the following day in the time series. The residual circulation is estimated using the transformed Eulerian-mean (TEM) equations (Edmon et al. 980, Andrews et al. 987) based on eddy covariance statistics, where asterisks denote TEM variables. The TEM streamfunction ψ* in pressure coordinates is defined as 44 4 [ ] [ ] ψ * =ψ πag cosϕ v θ p θ where a is the radius of the Earth, g is gravity, φ is latitude, and θ is potential () temperature. The conventional streamfunction ψ is computed from a downward vertical integration of [v], using ψ = 0 as the boundary condition at the top of the domain where the mass circulation approaches zero. The Empirical Orthogonal Function (EOF) patterns of ψ* displayed in this paper are obtained by regressing the ψ* field upon the standardized expansion coefficient time series computed from an EOF analysis of ψ* at specified pressure levels. The amplitudes of the patterns represent typical amplitudes of the anomalies associated with fluctuations in the respective EOF modes. Tests for statistical significance of the correlation and

9 9 4 regression fields are based on the student-t test with the number of temporal degrees of freedom (or effective sample size ) estimated using Eqn. 3 in Bretherton et al. (999) Structure of the tropical upwelling in the BDC The time mean structure of the TEM streamfunction ψ* field in the model simulation is shown in Fig. a and compared to the climatological-mean boreal winter (Dec, Jan, Feb) structure of ψ* in the ERA-40 reanalysis in Fig. b. The tropospheric contours highlighted with colored shading depict the tropical Hadley cells and their eddydriven extension to higher latitudes. In this perpetual January integration of the model, the simulated Southern Hemisphere (SH) circulation cell is substantially stronger than observed because of the unrealistically strong baroclinic eddies in the model s summer hemisphere. The wave-driven BDC in the stratosphere is indicated by the upper set of contours that appear above the colored shading. As in the troposphere, the SH (summer) stratospheric circulation is much too strong in the model: the BDC cells in the two hemispheres are of comparable strength in the simulation (Fig. a), whereas the Northern Hemisphere (NH) cell is clearly dominant in ERA-40 (Fig. b) and stronger than its SH counterpart in boreal summer (Jun, Jul, Aug), shown in Fig. c. The hemisphericallysymmetric pattern of ψ* in the simulation resembles the climatological annual mean structure of ψ* in the reanalysis (not shown). Since the model does not simulate the effects of latent heating on tropical convection, the simulated tropical cold point tropopause at ~00 hpa is lower than observed. The greater strength of the circulation in the NH winter stratosphere (Fig. b) than in the SH winter stratosphere (Fig. c) gives rise to the observed Jan-Feb peak in tropical

10 lower-stratospheric upwelling and the associated temperature minimum first noted by Reed and Vlcek (969). In ERA-40, the tropical upwelling region widens from ~ S- N in the troposphere to ~30 S-30 N in the lower stratosphere, illustrating the greater meridional width and the stronger equatorial symmetry of the upwelling branch of the BDC as compared with that of the Hadley cell. The nonlinearly-spaced contours depicting the stratospheric BDC are indicative of a change in the character of the circulation around the hpa level in ERA-40 (Fig. b). The steep vertical gradient of ψ* below hpa in the tropics is indicative of a strong circulation that is concentrated near or just above the cold point. An estimate of the upward mass flux based on ψ* in Fig. b indicates that only ~/4 of the mass of air that upwells through the 00 hpa surface in the S- N latitudinal belt reaches the hpa level in ERA-40, which is consistent with the vertical mass flux estimates of Corti et al. (00) and Fu et al. (007) derived from radiative transfer calculations. Above ~ hpa, the contours are more evenly spaced and the ψ* values decrease more gradually with height, indicating a weaker but deeper equator-to-pole circulation. This change in the vertical gradient of the streamfunction contours in ERA-40 is suggestive of a distinction between the stronger, shallow branch and the slower, deep branch of the BDC. An analysis of the mean age of air in the model by Gerber (0) confirms the existence of a qualitative change in the character of the circulation, which occurs at ~00 hpa in the model. He found that the rapid decay of the strength of the TEM circulation with height in the lower stratosphere results in a sharp gradient in the mean age of air, while much older air with a more uniform age prevails at higher levels where the strength of the circulation decays more slowly with height. These results are also consistent with

11 the vertical structure of the amplitudes of the annual cycle in tropical lower-stratospheric temperature, shown in Fig. of Reed and Vlcek (969). The amplitude peaks at ~80 hpa, presumably due to the strong tropical upwelling of the deep branch of the BDC during boreal winter in conjunction with the long radiative relaxation time scale at that level. The amplitude of the annual cycle diminishes rapidly with depth below ~80 hpa, reflecting the weaker seasonality in the strength of the shallow branch and the shorter radiative relaxation time scale at 00 hpa relative to 80 hpa. The intraseasonal variability of the deep and shallow branches of the BDC is investigated in Section. The dominant meridional structure of the intraseasonal variability of the observed and simulated BDC is clearly apparent in the time-latitude sections shown in Fig.. It resembles the time-mean structure of ψ* (Fig. ) with broad equatorially-symmetric perturbations encompassing the entire tropical belt out to ~30 latitude. Figure a shows TEM pressure velocity ω* anomalies at 6 hpa during a selected 36-day period in the 0,000 day integration of the model. Tropical ω* anomalies are usually flanked by anomalies of opposing sign in the winter hemisphere, in this case the NH, consistent with the conservation of mass. The same dynamically-induced perturbations are even more clearly evident in the time-latitude section of the -day time rate of change of temperature [T]/ t at 6 hpa for the same period of record, shown in Fig. b. Negative ω* anomalies, indicative of enhanced upwelling, coincide with negative [T]/ t anomalies, indicative of anomalous cooling, and vice versa. The strong positive correlation between tropical mean ω* and [T]/ t at the 6 hpa level follows directly from the TEM thermodynamic equation

12 [ ] t T + [ T ] p + R [ T ] c p p ω * = ( [ T ] [ T e ]) /τ () where R is the gas constant and c p the heat capacity of air at constant pressure. The zonal mean diabatic heating in the model is calculated by Newtonian relaxation to a radiative equilibrium profile T e with radiative time scale τ = 40 day - in the free troposphere and stratosphere. Newtonian relaxation is a useful approximation of the diabatic heating in the real atmosphere (e.g., Andrews et al. 987) with τ 30 day - in the lower stratosphere (Newman and Rosenfield 997). For time scales longer than τ, the dominant balance is between the latter two terms in () such that upwelling and temperature vary with opposite phase. On time scales much shorter than τ, however, the terms on the left-hand side of () tend to balance, and upwelling varies in phase with a tendency toward cooling. Planetary-scale fluctuations in [T]/ t, with broad and equatorially-symmetric perturbations in the tropics flanked by perturbations of opposing sign, are also apparent in the observations, as evidenced by the section of [T]/ t based on MSU-4 data from Oct 994 through Apr 99, shown in Fig. c. As in the simulation, the successive bands of anomalous cooling and warming are remarkably symmetric across the tropics out to ~30 latitude, and are flanked by strong anomalies of the opposing sign in high latitudes of the winter hemisphere. Intraseasonal fluctuations in ω* at hpa in ERA-40 (not shown) are dominated by high-frequency variability and are not as well correlated with variations in [T]/ t at the same level as in the model: r = 0. in ERA-40 vs in the model (see also Table ). Since we are not convinced that the data assimilated in ERA-40 yields reliable vertical velocities, we will henceforth use [T]/ t as a proxy for vertical velocities in the observed BDC.

13 Role of high-latitude wintertime planetary waves As an indicator of the high-latitude planetary wave forcing of the stratosphere, we superimpose the 00 hpa level poleward eddy heat flux [v't'] anomalies greater than m s - K on the time-latitude sections in Fig.. Eddy heat fluxes are associated with the vertical propagation of Rossby wave activity, as quantified in the Eliassen-Palm flux (Andrews and McIntyre 976; Edmon et al. 980). Episodes of strong poleward [v't'], centered around 60 latitude, are associated with strong tropical upwelling / cooling in the model (e.g., day ~ in Figs. a and b) and in the observations (e.g., mid-jan 9 in Fig. c). Similarly, anomalously weak fluxes tend to occur in concert with weak tropical upwelling / warming (e.g., day ~60 in Figs. a and b, Feb 9 in Fig. c). The highlatitude 00 hpa heat fluxes are primarily associated with planetary-scale waves; Polvani and Waugh (004) show that such bursts of upward wave dispersion often precede stratospheric sudden warming events. The strong correspondence between enhanced high-latitude heat fluxes and tropical upwelling and cooling in Fig. suggests that these wave breaking events force tropical upwelling in both the simulation and in the observations a. Forcing of tropical upwelling The impact of the high-latitude planetary wave forcing on the lower stratosphere is summarized in Fig. 3, which shows meridional profiles of lower-stratospheric ω* and [T]/ t constructed by regressing these fields upon an index of anomalous high-latitude (4-90 N) 00 hpa [v't'] at various lags. Both regression coefficients and correlation

14 coefficients are shown, and the fields correspond with those in the three panels of Fig.. The response of the lower-stratospheric [T]/ t profiles to high-latitude wave forcing is remarkably similar in the model simulation and the MSU-4 observations (Figs. 3b,c). Both show nearly instantaneous warming of the polar cap region and cooling of the tropics, consistent with an enhancement of the BDC, as confirmed by the ω* section of the model shown in Fig. 3a. The polar warming, which implies a weakening or collapse of the stratospheric polar night jet, lasts for ~ days and is followed by an interval of cooling associated with the recovery of the polar vortex. The equatorward propagation of the polar [T]/ t anomalies in Fig. 3 is also apparent in the time-latitude sections in Fig.. The widespread tropical cooling in response to high-latitude planetary wave forcing takes place over the course of ~0 days, centered around the time of the peak forcing. In the observations (Fig. 3c), the maximum tropical cooling occurs ~ days after the peak forcing, but no such delay is evident in the simulation at the 6 hpa level (Fig. 3b). The results based on correlation analysis, indicated by the colored shading in Fig. 3, show enhanced cooling on the equator and at ~0 S and ~0 N, forming a three-prong feature that is more prominent in the simulation than in the observations. Upon closer inspection, strong poleward [v't'] at 00 hpa over 4-90 N is also accompanied by a relatively weak and short-lived local cooling at ~30-60 N a few days prior to the time of the peak forcing, indicative of upwelling that appears first in midlatitudes and quickly spreads into and across the tropics. Figure 4 shows a sequence of meridional cross-sections depicting the evolution of the temperature response to the high-latitude wave forcing, as represented by the same

15 high-latitude 00 hpa [v't'] index used in constructing Fig. 3. Results for both the simulation and the observations are shown. The [T]/ t field is significantly correlated with the high-latitude eddy heat flux at lags ranging from 4 to +4 days, reflecting the persistence of the eddy heat flux anomalies. Hence, the ~0-day life cycle of the tropical upwelling response inferred from Fig. 3 appears to be set by the characteristic time scale of the modulations in the intensity of the planetary Rossby waves. Tropical cooling and polar warming first become evident ~4 days prior to the peak eddy forcing. The correlations become stronger in subsequent days with very little modulation of their spatial structures. The polar warming is strongest at the time of peak forcing and weakens thereafter. In ERA-40 (Fig. 4b), the warming of the polar cap region is flanked by cooling at ~40-60 N in the ~00-00 hpa layer. This lobe of localized cooling is connected to a broader tropical cooling above ~ hpa level that spreads across the equator into the SH. Tropical temperatures below hpa are minimally affected, a topic that we explore further in Section. The cooling of the entire tropical stratosphere peaks ~ days after the time of peak forcing in both the simulation and in ERA-40. The broad, equatorially-symmetric cooling signature persists for more than 4 days, with somewhat stronger correlations between tropical [T]/ t and high-latitude [v't'] in the model than in ERA-40. To relate the upward propagation of high-latitude planetary waves through the 00 hpa surface to the mechanical forcing of the stratosphere, we show in Fig. series of meridional cross-sections generated by correlating the divergence of the Eliassen-Palm (EP) flux upon the same high-latitude [v't'] index as was used in constructing Figs. 3 and 4. The evolution of the EP flux divergence may be interpreted as tracing out the region

16 of wave breaking. The sequence begins with wave forcing concentrated in the extratropical stratosphere poleward of 4 N. In the simulation (Fig. a), there is a hint of a secondary center in the subtropics that becomes stronger and more prominent in subsequent panels. By -4 days after the peak forcing, nearly all of the wave forcing in the model has shifted to the tropical and subtropical latitudes. The wave forcing in ERA- 40 is not as strong as in the model, as evidenced by the weaker correlations in Fig. b compared to those in Fig. a, but the evolution of the EP flux divergence field is similar. At the time of peak high-latitude wave forcing, EP flux divergence is maximum in the region poleward of 4 N above the 00 hpa level. The region of wave forcing shifts equatorward with time so that by + days, most of the wave breaking occurs in the subtropical stratosphere. By +4 days, all remnants of the high-latitude wave breaking have completely disappeared, but wave breaking in the subtropics is still clearly evident. The sequence of patterns in Fig. suggests that planetary wave activity entering the stratosphere poleward of ~4 N is deflected equatorward. The rather abrupt shift in the wave-breaking region from the high latitudes to the subtropics around + days in Fig. corresponds to the lagged response of the tropical stratospheric cooling to high-latitude wave forcing in Figs. 3 and 4. These results suggest that high-latitude planetary waves are capable of forcing upwelling in the deep tropics. On shorter time scales relative to the thermal damping time scale of the stratosphere, high-latitude wave breaking alone can potentially modulate upwelling in the tropics, as suggested by sideways control (Holton et al. 99). On longer time scales, when downward control becomes applicable, the wave forcing must penetrate to the low latitudes, as they appear to do so in both the model and in ERA-40. In either case, the spreading of the wave breaking across much of

17 the hemisphere on a time scale of days, however, suggests that a distinction between high and low latitude wave breaking associated with real planetary waves may be somewhat artificial. Realistic wave events are associated with a very broad pattern of wave breaking, as shown in Fig.. While the overall structure and evolution of the EP flux divergence field in ERA- 40 resemble those in the model, the variability of the low-latitude EP fluxes is significantly less in ERA-40. Not only is the subtropical disturbance much weaker, but its equatorward extent does not reach beyond ~0 N. In comparison, the simulated waves appear to propagate to within a few degrees of the equator. It is unclear whether the model should be believed at this level of detail, and quantitative comparisons should be made with more sophisticated models b. Vertical structure of the tropical upwelling response The high-latitude wintertime planetary waves perturb tropical temperatures by way of an equator-to-pole residual circulation that extends deep into the stratosphere, as depicted in Fig. 6. The ψ* fields in the model (left) and observations (right) are correlated with and regressed upon the daily time series of (in the top panels) tropical (30 S-30 N) lower-stratospheric [T]/ t and (in the lower panels) the high-latitude 00 hpa [v't'] index used in Figs. 3-. The TEM circulation observed in association with tropical [T]/ t (Figs. 6a,b) exhibits a deep vertical structure, with upwelling spreading across the equator into the SH, particularly in the model. The overall shape of these patterns resembles the deep, upper branch of the climatological-mean boreal winter BDC in Fig. b. The striking similarity between the circulation cells observed in association

18 with high-latitude [v't'] (Figs. 6c,d) and those associated with tropical temperature tendencies (Figs. 6a,b), aside from the reversal in the signs, indicates that the highlatitude planetary wave activity and tropical lower-stratospheric temperature perturbations are tightly coupled through variations in the strength of the deep branch of the BDC. The 0-day Lanczos lowpass filtered daily time series of tropical [T]/ t at 6 hpa and high-latitude [v't'] at the 00 hpa level are correlated at a level of 0.7 in the simulation. The corresponding correlations based on MSU-4 and ERA-40 [T]/ t are slightly weaker ( 0.63 and 0., respectively; Table ), but still highly significant. We conclude from these results that planetary waves emanating from high latitudes propagate into and break in the tropical to subtropical stratosphere, inducing a residual circulation with downwelling / warming in the polar cap region and upwelling / cooling in the tropical stratosphere above ~ hpa, as depicted in the regression crosssections shown in Fig. 7. Temperature changes in the upper stratosphere dominate the variability in the model simulation, whereas the reanalysis also shows a narrow band of temperature tendencies associated with the strong local residual circulation poleward of ~40 N extending down to ~00 hpa. Nonetheless, the general features in the simulation and ERA-40 are similar, indicating that the sequence of events involving the high-latitude planetary wave forcing of the tropical upwelling is common to both Forcing of the variability below hpa The results presented in the previous sections suggest that planetary waves that propagate upward and equatorward from the high latitudes and break in the tropical and

19 subtropical stratosphere are the dominant driver of upwelling in the tropical stratosphere at and above the hpa level. In this section, we expand the domain of the investigation to include the -00 hpa layer. The analyses in this section are based on ERA-40 and MSU-4 data only, given the bias in the model s tropical tropopause height. Figure 8 shows correlations and regressions of the [T]/ t field with the daily time series of tropical (30 S-30 N) [T]/ t at various levels. In Fig. 8a, the tropical [T]/ t time series is derived from MSU-4 temperatures. The ERA-40 [T]/ t field is significantly correlated with MSU-4 tropical [T]/ t time series throughout the 0- hpa layer, with a maximum correlation of 0.64 at the 0 hpa level. Hence, it is evident that MSU-4 samples temperatures over a deep layer in the stratosphere centered at ~0 hpa in ERA-40 and that the intraseasonal variability in [T]/ t is relatively uniform within the 0- hpa layer. Figures 8b,c show correlation and regression cross-sections of [T]/ t upon tropical [T]/ t time series at the 0 and 00 hpa levels in the ERA-40 data. By construction, intraseasonal variations in [T]/ t are highly correlated at the level of the tropical time series. Tropical [T]/ t variations at the 0 hpa level are strongly correlated with tropical [T]/ t at levels ranging from 0 to hpa (Fig. 8b), as evidenced by the deep vertical extent of the shaded region. However, the correlations drop off rapidly with depth below hpa. A similar result is obtained when the [T]/ t field is correlated with tropical [T]/ t time series at the hpa level (not shown). In contrast, the variability in tropical [T]/ t at the 00 hpa level (Fig. 8c) is localized and only weakly correlated with the variability at neighboring levels. Time series of tropical [T]/ t at the and 00 hpa

20 levels are weakly correlated (r = 0.36 for 0-day Lanczos lowpass filtered daily data; Table ). The breaking of planetary waves originating in high latitudes explains much of the variability in tropical upwelling at ~ hpa and above, but it is by no means certain that it is of comparable importance in accounting for variations in the temperature of the tropical cold point tropopause, which is situated closer to the 00 hpa level. The correlation between high-latitude 00 hpa [v't'] and the tropical temperature tendency, shown in Fig. 4b, drops off with depth from 0. at hpa to 0.7 at 00 hpa (Table ). Variations in MSU-4 temperatures are more strongly correlated with ERA-40 temperatures at hpa than at 00 hpa (r = 0.8 vs. 0.3; Table ) and thus reflect the variability of the deep branch of the BDC, which is forced mainly by planetary waves originating in high latitudes. The structure of the TEM circulation that influences the variability in tropical upwelling at and 00 hpa levels is investigated further by performing a onedimensional EOF analysis of 0-day Lanczos lowpass filtered ψ* at specified pressure levels. The pattern obtained by regressing the ψ* field upon the leading standardized expansion coefficient or principal component (PC) time series derived from EOF analysis of ψ* at the hpa level is shown in Fig. 9a. The TEM circulation pattern resembles the deep, equator-to-pole BDC associated with variations in tropical [T]/ t and high-latitude [v't'] as seen in Figs. 6b,d and 7b. This mode explains 44% of the intraseasonal variability (on time scales 0 days and longer) in ψ* at hpa. The leading EOF of ψ* at the 00 hpa level, shown in Fig. 9b, consists of a shallower equator-to-pole circulation that is mainly confined in the lower stratosphere. A deeper mode that

21 resembles the leading EOF of ψ* at the hpa level (Fig. 9a) appears as the third mode in the 00 hpa expansion shown in Fig. 9c, but it accounts for only % of the variance. The results of the EOF analysis support our conclusion that the influence of the variations in the deep branch of the BDC, driven by the breaking of planetary waves originating in high latitudes, declines rapidly with depth below hpa. The three-fold reduction in the variability of ψ* from to 00 hpa associated with the deep mode (EOF at hpa, EOF 3 at 00 hpa) is fairly consistent with the sharp gradient in the mass transport between and 00 hpa levels: recall that only one quarter of the mass that upwells through the 00 hpa surface reaches the hpa level. Hence wave breaking above hpa, the primary driver of the deep circulation, can only have a limited impact at 00 hpa, in comparison to the wave forcing in the lower stratosphere between and 00 hpa levels Summary of findings and discussion In the foregoing sections, we have investigated the forcing of the tropical upwelling in the Brewer-Dobson circulation (BDC) based on the analysis of perpetual January integrations of an idealized general circulation model and November through March data of the ERA-40 reanalysis and MSU-4 measurements of lower-stratospheric temperature. Our main findings are as follows: ) The idealized model captures the observed structure and intraseasonal variability of the strength of the BDC reasonably well, and the model results show that variations in

22 tropical upwelling are well represented by the time series of the -day time rate of change of temperature ( [T]/ t). ) Enhanced poleward eddy heat fluxes [v't'] in the high latitudes (4-90 N) at the 00 hpa level are associated with anomalous tropical cooling (i.e., negative [T]/ t) and anomalous warming (i.e., positive [T]/ t) on the poleward side of the polar night jet (Figs. -4), suggestive of a strengthening of the BDC driven by the breaking of highlatitude wintertime planetary waves. The nearly instantaneous cooling of the tropics above the hpa level persists over a 0 day interval centered around the time of peak wave forcing (Figs. 3, 4). 3) Lag-lead correlations of the EP flux divergence field with high-latitude wave forcing, as represented by an index of the 00 hpa level [v't'] poleward of 4 N, illustrate the equatorward propagation of waves from the high latitudes to the subtropical and tropical stratosphere over a ~0 day period (Fig. ). 4) The response of the transformed Eulerian-mean (TEM) streamfunction ψ* field to the high-latitude [v't'] index shows a pattern that is virtually identical to those obtained by regressing the ψ* field upon the time series of tropical [T]/ t at the hpa level (Figs. 6, 7) and upon the PC time series of the leading EOF of ψ* at the hpa level (Fig. 9a). ) The influence of planetary waves originating in high latitudes upon the strength of the tropical upwelling diminishes rapidly with depth below hpa. Intraseasonal variations in tropical [T]/ t at 00 hpa are not strongly correlated with those at hpa and above (Fig. 8c, Table ) or with the high-latitude [v't'] index (Table ). Hence it is quite possible that variations in upwelling at the tropical cold point near

23 the 00 hpa level could be primarily controlled by variations in the intensity of the shallow branch of the BDC driven by the breaking of midlatitude baroclinic synopticscale waves and/or equatorial planetary waves. This study reconciles the apparent contradiction stated in the Introduction between the numerous empirical studies that have demonstrated the existence of highly significant correlations between high-latitude planetary wave activity and tropical upwelling variations and the even more numerous studies based on downward control diagnostics which show that wave breaking at extratropical latitudes is ineffective in forcing variations in tropical upwelling. There is no real contradiction: we have shown that in both the simulation and in the observations, high-latitude planetary waves propagate equatorward at levels above hpa and break at latitudes low enough to effectively force the deep branch of the BDC. The apparent contradiction stems from the prior lack of observational evidence of the equatorward propagation of the high-latitude planetary waves, which has led many authors to emphasize the role of low-latitude wave sources. There are several factors that have obscured the equatorward propagation of the high-latitude planetary waves in prior observational studies: In the analysis of Ueyama and Wallace (00), the eddy fluxes of heat and zonal momentum were based on -day mean data. Zhou et al. (0) have shown that lowlatitude fluxes computed on the basis of 4 times daily ERA-40 data are substantially larger. Zhou et al. (0) interpreted their larger fluxes as being due to the contribution of synoptic-scale waves, but it is possible that breaking planetary waves

24 emanating from high latitudes could include features that would not be resolvable in pentad-mean data. Even in the 4 times daily fields derived from ERA-40 reanalysis, features associated with the breaking of planetary waves in low latitudes may be obscured by the data assimilation cycle, in which dynamical consistency is continually being sacrificed for the sake of updating the analysis with new observations. In this respect, the results based on numerical simulations presented in Fig. of this study may well provide a more reliable indication of the amplitude of the low-latitude eddy forcing than the results based on observations. Improved quality of observations and more comprehensive simulations should help further resolve this issue in the future. At stratospheric levels, the vertical velocity fields in the reanalyses are subject to large uncertainties. Our use of the -day temperature tendency as a proxy for vertical velocity in this study provides a more reliable estimate of the intraseasonal variations in tropical upwelling than has been available in prior observational studies. The impact of the wave forcing on the tropical upwelling is inversely proportional to the Coriolis parameter. Hence, the smallness of the low-latitude fluxes belies their dynamical significance. Our use of correlation coefficients in this study places greater emphasis on the low-latitude fluxes than the use of regression coefficients in Ueyama and Wallace (00). We have also shown that the relative importance of the high-latitude planetary waves in inducing variability of tropical upwelling in the BDC depends on the level at which the upwelling is evaluated, in agreement with the two branch model of the BDC (Held and Hoskins 98; Plumb 00; Vallis 006; Shepherd 007; Birner and Bönisch

25 ). The column-averaged tropical temperatures between 0 and 00 hpa analyzed by Salby and Callaghan (00) and the zonally-averaged temperature profiles from MSU-4 analyzed in Ueyama and Wallace (00) represent the vertical velocities within the deep branch of the BDC. It is not surprising, then, that these studies found that variations in tropical upwelling are more strongly correlated with high-latitude planetary wave activity than with wave activity originating at lower latitudes. In contrast, an appreciable contribution by low-latitude wave forcing was found in studies that examined the residual mean vertical velocities near the 00 hpa level, which are more strongly influenced by the shallow branch of the BDC; e.g., at 3-7 km (~90-80 hpa) in Boehm and Lee (003), 90 hpa in Kerr-Munslow and Norton (006), 00 hpa in Randel et al. (008), and 00 hpa in Garny et al. (0). The extent to which variations in tropical upwelling at the tropical cold point tropopause level ~00 hpa may be related to temporal variations in the intensity of baroclinic wave activity or to variations in the strength or structure of the equatorial planetary waves via the shallow branch of the BDC remains to be determined Acknowledgments. MSU/AMSU data are produced by Remote Sensing Systems and sponsored by the NOAA Climate and Global Change Program. This work was supported by the National Science Foundation through the Climate Dynamics Program Office, Grant 989 and the division of Atmospheric and Geospace Sciences, Grant R. Ueyama also

26 received partial financial support from Q. Fu supported by NASA Grant NNX09AH73G and NOAA Grant NA08OAR4307.

27 References Andrews, D. G. and M. E. McIntyre, 976: Planetary waves in horizontal and vertical shear: The generalized Eliassen-Palm relation and the mean zonal acceleration. J. Atmos. Sci., 33, , J. R. Holton, and C. B. Leovy, 987: Middle Atmosphere Dynamics. Academic Press, 489 pp. Birner, T., and H. Bönisch, 0: Residual circulation trajectories and transit times into the extratropical lowermost stratosphere. Atmos. Chem. Phys.,, Boehm, M. T., and S. Lee, 003: The implications of tropical Rossby waves for tropical tropopause cirrus formation and for the equatorial upwelling of the Brewer- Dobson circulation. J. Atmos. Sci., 60, Bretherton, C. S., M. Widmann, V. P. Dymnikov, J. M. Wallace, and I. Bladé, 999: The effective number of spatial degrees of freedom of a time-varying field. J. Climate,, Brewer, A. W., 949: Evidence for a world circulation provided by the measurements of helium and water vapor distribution in the stratosphere. Quart. J. Roy. Meteor. Soc., 7, Chen, G., and L. Sun, 0: Mechanisms of the tropical upwelling branch of the Brewer- Dobson circulation: The role of extratropical waves. J. Atmos. Sci., 68,

28 Corti, T., B. P. Luo, and T. Peter, 00: Mean radiative energy balance and vertical mass fluxes in the equatorial upper troposphere and lower stratosphere. Geophys. Res. Lett., 3, L0680, doi:0.09/004gl0889. Deckert, R., and M. Dameris, 008: Higher tropical SSTs strengthen the tropical upwelling via deep convection. Geophys. Res. Lett., 3, L083, doi: 0.09/008GL Dhomse, S., M. Weber, and J. Burrows, 008: The relationship between tropospheric wave forcing and tropical lower stratospheric water vapor. Atmos. Chem. Phys., 8, Dobson, G. M. B., 96: Origin and distribution of polyatomic molecules in the atmosphere. Proc. Roy. Soc. London, 36A, Edmon, Jr., H. J., B. J. Hoskins, and M. E. McIntyre, 980: Eliassen-Palm cross sections for the troposphere. J. Atmos. Sci., 37, Fu, Q., Y. Hu, and Q. Yang, 007: Identifying the top of the tropical tropopause layer from vertical mass flux analysis and CALIPSO lidar cloud observations. Geophys. Res. Lett., 34, L483, doi:0.09/007gl Fueglistaler, S., A. E. Dessler, T. J. Dunkerton, I. Folkins, Q. Fu, and P. W. Mote, 009: Tropical tropopause layer. Rev. Geophys., 47, RG004. Garcia, R. R., 987: On the mean meridional circulation of the middle atmosphere. J. Atmos. Sci., 44, Garny, H., M. Dameris, W. Randel, G. E. Bodeker, and R. Deckert, 0: Dynamically forced increase of tropical upwelling in the lower stratosphere. J. Atmos. Sci., 68, 4-33.

29 Geller, M. A., T. Zhou, and K. Hamilton, 008: Morphology of tropical upwelling in the lower stratosphere. J. Atmos. Sci., 6, Gerber, E. P., and L. M. Polvani, 009: Stratosphere-troposphere coupling in a relatively simple AGCM: The importance of stratospheric variability. J. Climate,, , 0: Stratospheric versus tropospheric control of the strength and structure of the Brewer-Dobson circulation. J. Atmos. Sci., in press. Gettelman, A., P. M. de F. Forster, M. Fujiwara, Q. Fu, H. Vömel, L. K. Gohar, C. Johanson, and M. Ammerman, 004: Radiation balance of the tropical tropopause layer. J. Geophys. Res., 09, D03, doi:0.09/003jd Haynes, P. H., and M. E. McIntyre, 987: On the evolution of vorticity and potential vorticity in the presence of diabatic heating and frictional and other forces. J. Atmos. Sci., 44, , C. J. Marks, M. E. McIntyre, T. G. Shepherd, and K. P. Shine, 99: On the downward control of extratropical diabatic circulations by eddy-induced mean zonal forces. J. Atmos. Sci., 48, Held, I. M., and B. J. Hoskins, 98: Large-scale eddies and the general circulation of the troposphere. Advances in Geophys., 8A, 3-3., and M. J. Suarez, 994: A proposal for the intercomparison of the dynamical cores of atmospheric general circulation models. Bull. Am. Meteor. Soc., 7, Holton, J. R., 990: On the global exchange of mass between the stratosphere and troposphere. J. Atmos. Sci., 47,

30 , P. H. Haynes, M. E. McIntyre, A. R. Douglass, R. B. Rood, and L. Pfister, 99: Stratosphere-troposphere exchange. Rev. Geophys., 33, Iwasaki, T. 99: General circulation diagnosis in the pressure-isentrope hybrid vertical coordinate. J. Meteor. Soc. Japan,, Kerr-Munslow, A. M., and W. A. Norton, 006: Tropical wave driving of the annual cycle in tropical tropopause temperatures. Part I: ECMWF analyses. J. Atmos. Sci., 63, Mote, P. W., and Coauthors, 996: An atmospheric tape recorder: The imprint of tropical tropopause temperatures on stratospheric water vapor. J. Geophys. Res., 0, Newman, P. A., and J. E. Rosenfield, 997: Stratospheric thermal damping times. Geophys. Res. Lett., 4, Norton, W. A., 006: Tropical wave driving of the annual cycle in tropical tropopause temperatures. Part II: Model results. J. Atmos. Sci., 63, Plumb, R. A., and Eluszkiewicz, 999: The Brewer-Dobson circulation: Dynamics of the tropical upwelling. J. Atmos. Sci., 6, , 00: Stratospheric transport. J. Meteor. Soc. Japan, 80, Polvani, L. M., and P. J. Kushner, 00: Tropospheric response to stratospheric perturbations in a relatively simple general circulation model. Geophys. Res. Lett., 9(7), 0.09/00GL0484., and D.W. Waugh, 004: Upward wave activity flux as precursor to extreme stratospheric events and subsequent anomalous surface weather regimes, J. Climate, 7,