• Abreu, V. J., P. B. Hays, D. A. Ortland, W. R. Skinner, and J.-H. Yee, 1989: Absorption and emission line shapes in the O2 atmospheric bands: Theoretical model and limb viewing simulations. Appl. Opt.,28, 2128–2137.

  • Bacmeister, J. T., 1993: Mountain-wave drag in the stratosphere and mesosphere inferred from observed winds and a simple mountain-wave parameterization scheme. J. Atmos. Sci.,50, 377–399.

  • Burrage, M. D., and Coauthors, 1993: Comparison of HRDI wind measurements with radar and rocket observations. Geophys. Res. Lett.,20, 1259–1262.

  • ——, D. L. Wu, W. R. Skinner, D. A. Ortland, and P. B. Hays, 1995: Latitude and seasonal dependence of the semidiurnal tide observed by the High Resolution Doppler Imager. J. Geophys. Res.,100, 11 313–11 321.

  • ——, and Coauthors, 1996: Validation of mesosphere and lower thermosphere winds from the High Resolution Doppler Imager on UARS. J. Geophys. Res.,101, 10 365–10 392.

  • Dunkerton, T. J., and N. Butchart, 1984: Propagation and selective transmission of internal gravity waves in a sudden warming. J. Atmos. Sci.,41, 1443–1460.

  • Fritts, D. C., 1984: Gravity wave saturation in the middle atmosphere: A review of theory and observations. Rev. Geophys. Space Phys.,22, 275–308.

  • Gelman, M. E., A. J. Miller, K. W. Johnson, and R. M. Nagatani, 1986: Detection of long term trends in global stratospheric temperature from NMC analyses derived from NOAA satellite data. Adv. Space Res.,6(10), 17–26.

  • Hays, P. B., V. J. Abreu, M. E. Dobbs, D. A. Gell, H. J. Grassl, and W. R. Skinner, 1993: The High Resolution Doppler Imager on the Upper Atmosphere Research Satellite. J. Geophys. Res.,98, 10 713–10 723.

  • ——, D. L. Wu, and the HRDI Science Team, 1994:Observations of the diurnal tide from space. J. Atmos. Sci.,51, 3077–3093.

  • Holton, J. R., 1982: The role of gravity wave induced drag and diffusion in the momentum budget of the mesosphere. J. Atmos. Sci.,39, 791–799.

  • ——, 1984: The generation of mesospheric planetary waves by zonally asymmetric gravity wave breaking. J. Atmos. Sci.,41, 3427–3430.

  • Lindzen, R. S., 1981: Turbulence and stress owing to gravity wave and tidal breakdown. J. Geophys. Res.,86, 9707–9714.

  • ——, and H.-L. Kuo, 1969: A reliable method for the numerical integration of a large class of ordinary and partial differential equations. Mon. Wea. Rev.,97, 732–734.

  • Matsuno, T., 1970: Vertical propagation of stationary planetary waves in the winter Northern Hemisphere. J. Atmos. Sci.,27, 871–883.

  • McLandress, C., and N. A. McFarlane, 1993: Interactions between orographic gravity wave drag and forced planetary waves in the winter Northern Hemisphere middle atmosphere. J. Atmos. Sci.,50, 1966–1990.

  • Miyahara, S., 1985: Suppression of stationary planetary waves by internal gravity waves in the mesosphere. J. Atmos. Sci.,42, 100–107.

  • Morton, Y. T., and Coauthors, 1993: Global mesospheric tidal winds observed by the High Resolution Doppler Imager on board the Upper Atmosphere Research Satellite. Geophys. Res. Lett.,20, 1263–1266.

  • Ortland, D. A., P. B. Hays, W. R. Skinner, M. D. Burrage, A. R. Marshall, and D. A. Gell, 1995: A sequential estimation technique for recovering atmospheric data from orbiting satellites. The Upper Mesosphere and Lower Thermosphere, Geophys. Monogr., Ser. No. 87, R. M. Johnson and T. L. Killeen, Eds., Amer. Geophys. Union.

  • Randel, W. J., 1987: The evaluation of winds from geopotential height data in the stratosphere. J. Atmos. Sci.,44, 3097–3120.

  • Schoeberl, M. R., and D. F. Strobel, 1984: Nonzonal gravity wave breaking in the winter mesosphere. Dynamics of the Middle Atmosphere, J. R. Holton and T. Matsuno, Eds., Terra Scientific, 45–64.

  • Smith, A. K., 1996: Longitudinal variations in mesospheric winds: Evidence for gravity wave filtering by planetary waves. J. Atmos. Sci.,53, 1156–1173.

  • View in gallery

    Correlation of HRDI zonal (left) and meridional (right) winds with the vertical average of NMC stratospheric winds for the same latitude, longitude, and date for August 1992, 1993, and 1994. The shaded areas indicate where the correlation is significant at the 95% level.

  • View in gallery

    As in Fig. 1 but for September 1992, 1993, and 1994.

  • View in gallery

    Average zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) from NMC and HRDI at 40°S for August 1992, 1993, and 1994.

  • View in gallery

    Longitude × altitude cross sections of perturbation zonal winds from NMC and HRDI averaged for the month of February 1994 (40°–50°N) and August 1994 (30°–40°S).

  • View in gallery

    Correlation of the longitudinally asymmetriczonal and meridional HRDI winds at the same locations and times for August of 1992–94.

  • View in gallery

    Longitude × latitude cross sections of wind vectors computed from NMC at 1 mb on 24 February 1993 and 24 August 1994.

  • View in gallery

    Average of August zonal wind speeds from NMC (below 48 km) and HRDI (above 60 km) for all available data from August 1992–94. On the right is the complete model zonal wind field obtained by interpolation/extrapolation.

  • View in gallery

    As in Fig. 7 but for February.

  • View in gallery

    Rayleigh friction coefficient (solid line), Newtonian cooling coefficient (short dashed line), and timescale for internal forcing (long dashed line) used in the linear wave model.

  • View in gallery

    Latitude × altitude cross section of the geopotential wave amplitude (in meters) calculated by the linear wave model for the propagating wave case (left panel) and the internal forcing case (right panel), using the August mean winds.

  • View in gallery

    Zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) from the linear wave model for August at 42°S for the propagating wave case (upper panels) and internal forcing case (lower panels).

  • View in gallery

    As in Fig. 10 but using the February mean winds.

  • View in gallery

    As in Fig. 11 but for February at 54°N.

  • View in gallery

    Average zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) at 52°N from NMC and HRDI for February 1993 and 1994.

  • View in gallery

    Average zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) at 40°N from NMC and HRDI for December 1992, 1993, and 1994.

  • View in gallery

    Average zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) at 40°S from NMC and HRDI for October 1992.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 123 123 4
PDF Downloads 16 16 4

Stationary Planetary Waves in Upper Mesospheric Winds

View More View Less
  • 1 Atmospheric Chemistry Division, National Center for Atmospheric Research, Boulder, Colorado
© Get Permissions
Full access

Abstract

Quasi-stationary planetary-scale longitudinal variations are found in the upper mesospheric winds measured during winter by the HRDI satellite instrument. These are negatively correlated with eddy winds in the stratosphere. Two different mechanisms are proposed to explain the mesospheric perturbation winds and their anticorrelation with stratospheric winds: 1) Planetary waves propagate through the stratosphere and into the mesosphere, with a phase shift of one-half cycle and 2) mesospheric perturbations are forced in situ by gravity waves whose spectrum has longitudinal asymmetries due to filtering by planetary waves in the stratosphere. The first mechanism is more consistent with the observations during Southern Hemisphere late winter (August) and also may explain the observations during Northern Hemisphere early winter (December). The second mechanism gives a more consistent explanation for the Northern Hemisphere late winter observations, as previously shown by the author. The hemispheric differences are reproduced in two linear model calculations in which the only difference is the background zonal mean wind.

Corresponding author address: Dr. Anne K. Smith, NCAR/ACD, P.O. Box 3000, Boulder, CO 80307-3000.

Email: aksmith@ncar.ucar.edu

Abstract

Quasi-stationary planetary-scale longitudinal variations are found in the upper mesospheric winds measured during winter by the HRDI satellite instrument. These are negatively correlated with eddy winds in the stratosphere. Two different mechanisms are proposed to explain the mesospheric perturbation winds and their anticorrelation with stratospheric winds: 1) Planetary waves propagate through the stratosphere and into the mesosphere, with a phase shift of one-half cycle and 2) mesospheric perturbations are forced in situ by gravity waves whose spectrum has longitudinal asymmetries due to filtering by planetary waves in the stratosphere. The first mechanism is more consistent with the observations during Southern Hemisphere late winter (August) and also may explain the observations during Northern Hemisphere early winter (December). The second mechanism gives a more consistent explanation for the Northern Hemisphere late winter observations, as previously shown by the author. The hemispheric differences are reproduced in two linear model calculations in which the only difference is the background zonal mean wind.

Corresponding author address: Dr. Anne K. Smith, NCAR/ACD, P.O. Box 3000, Boulder, CO 80307-3000.

Email: aksmith@ncar.ucar.edu

1. Introduction

A recent study by Smith (1996) investigated the large-scale zonal asymmetries in HRDI mesospheric winds during Northern Hemisphere late winter. The results indicate that there is a large and persistent anticorrelation between the mesospheric eddy winds and the stratospheric winds below.

Several possible explanations for this correlation were proposed and investigated. The planetary-scale disturbances in the mesosphere were assumed to originate ultimately from one or both of two broad categories of forcing. The first is a planetary Rossby wave that propagates from below and therefore has a related structure in the stratosphere and mesosphere. The other is a planetary-scale variation in localized forcing of the mesospheric flow by gravity waves. There are several factors that could cause gravity wave driving to have large-scale asymmetries: gravity wave sources in the troposphere may be linked to fixed geographical features such as the continent/ocean distribution or position of mountain ranges (Holton 1984; Schoeberl and Strobel 1984; Bacmeister 1993); gravity waves can undergo filtering by stratospheric planetary waves that can affect their momentum deposition in the mesosphere (Dunkerton and Butchart l984; McLandress and McFarlane 1993); and the momentum forcing by gravity waves may depend nonlinearly on the background flow at the breaking level (Lindzen 1981; Fritts 1984).

By examining the HRDI mesospheric winds and their relationship to stratospheric winds derived from conventional NMC (National Meteorological Center, now the National Centers for Environmental Prediction) observations, Smith (1996) was able to narrow down the list of possible causes. Geographical variation in thetropospheric sources of gravity waves was found not to be a major contributor to the observed anticorrelations. Geographical variability could give a correlation between mesospheric wind variations and stratospheric planetary waves if the planetary waves also have a fixed relationship to topography. The data indicated that, during February, there were significant differences in the stratospheric wave patterns for the years 1992–94 and that the mesospheric eddy variations followed the stratospheric winds rather than having a geographically fixed pattern. The nonlinear impact of background wind variations was also eliminated as a cause of the anticorrelation because it cannot by itself produce planetary-scale disturbances. However, this effect needs to be accounted for when considering how the mesospheric flow responds to longitudinal asymmetries in gravity wave momentum deposition.

The two explanations that have not been eliminated both depend on planetary waves in the stratosphere. A wave that propagates from the stratosphere to the mesosphere will have its strongest negative correlation in eddy winds over a vertical depth equivalent to half of the vertical wavelength. Because the large amplitude quasi-stationary Rossby waves in the winter stratosphere can have very large vertical wavelengths, the background winds and the wave structure can change substantially over a wavelength. It is therefore difficult to tell what are the features that can be used to identify the stratospheric and mesospheric disturbances unambiguously as a coherent propagating wave.

Smith (1996) looked at the structure of the dominant planetary wavenumber over the middle atmosphere in the NH February data. Although the zonal wavenumber was in all cases the same in the stratosphere as in the mesosphere, there were other features that did not fit with this being a single planetary wave. The amplitude of meridional wind had a deep minimum in the lower or middle mesosphere (65–70 km). By the geostrophic relation, the meridional wind amplitude is exactly proportional to the geopotential amplitude. This minimum, which occurs below the level that momentum deposition and turbulence generation by gravity waves is believed to be maximum, is difficult to explain in terms of a vertically propagating Rossby wave. The meridional wind structure is therefore not supportive of the interpretation as a coherent Rossby wave through a deep region of the middle atmosphere.

The interpretation is complicated by the structure of the eddy zonal wind, which during two out of three of the Februarys investigated does not show a comparable minimum in the lower mesosphere. The relationship between zonal wind amplitude and geopotential amplitude is not as simple as that for meridional wind but, nevertheless, the absence of a minimum in zonal wind amplitude while one exists for meridional wind amplitude confuses the interpretation.

In addition to the meridional wind amplitude structure, another factor argues against the planetary wave in the upper mesosphere being a continuation of that in the stratosphere. Planetary waves in the winter hemisphere tend to have meridional Eliassen–Palm (EP) flux in the equatorward direction or, equivalently, poleward momentum flux. The momentum flux in the stratosphere was indeed in the poleward direction, as determined by a positive correlation of zonal and meridional wind. The correlation of zonal and meridional eddy winds was also calculated for the mesosphere. There was no consistency in the sign or magnitude in latitude or altitude or between different periods. This, again, is not conclusive but argues against there being a Rossby wave in the upper mesosphere that has propagated upward from the stratosphere.

The explanation accepted in Smith (1996) as the most likely was the in situ forcing of mesospheric asymmetries due to filtering of gravity waves in the stratosphere. If the gravity wave spectral distribution leaving thetroposphere is statistically (over a month) uniform in longitude, that leaving the upper stratosphere will nevertheless be distorted. The global-scale distortion associated with filtering easterly gravity waves in the summer hemisphere and westerly gravity waves in the winter hemisphere is generally accepted as accounting for the closure of the jets in the mesosphere, where many of the gravity waves dissipate (Holton 1982). The modification of the spectrum by the planetary waves in the stratosphere should also be substantial since zonal asymmetries can reach tens of meters per second. Unlike the seasonal differences, longitudinal asymmetries can also be substantial for the meridional winds associated with planetary waves, and they too show a negative correlation between the stratosphere and mesosphere. When the gravity waves break or dissipate in the mesosphere, they will induce a longitudinally asymmetric momentum source that can generate wave patterns whose winds are out of phase with those in the stratosphere (Miyahara 1985).

The present paper extends the analysis of Smith with additional observations. Southern Hemisphere (SH) winter observations are used as a comparison to the NH observations described in Smith (1996). These have significant differences from the NH and therefore cannot be regarded as a simple confirmation of the conclusions based on the NH observations. In general, the SH observations are more consistent with a vertically propagating Rossby wave than the NH results are. A linear wave model is then used to reproduce the observations by adding an internal forcing that is meant to represent the in situ planetary wave generation and its relationship to the propagating planetary wave. The model indicates that the hemispheric differences are consistent with the background mean zonal wind structures, which cause significant differences in the altitude to which planetary waves can propagate.

2. Data and processing

Horizontal winds in the mesosphere were measured by the High Resolution Doppler Imager (HRDI) on the Upper Atmosphere Research Satellite (UARS). The instrument and data inversion procedure are described by Abreu et al. (1989), Hays et al. (1993), and Ortland et al. (1995). The HRDI telescope vertically scans the atmospheric limb at azimuthal angles of 45° and 135° with respect to the satellite’s orbital track. The velocity component along the line of sight is determined by the Doppler shift of one of three O2 emission lines. Evaluation of the wind speed requires accurate removal of the contributions from the satellite motion and the rotation of the earth. Comparisons with wind data from rockets and radar (Burrage et al. 1993, 1996) indicate that the basic structure agrees well with the other measurements and that mean differences in wind speed are in the range of 5–10 m s−1. Part of the difference is due to systematic bias; HRDI winds have on average stronger speeds than those measured by MF radars. When the wind directions (regardless of speed) are compared, the agreement between HRDI and MF radars is good (Burrage et al. 1996).

The observation pattern of HRDI is variable and has undergone several changes during the mission. Until March 1994, alternate days were devoted to observing the stratosphere and the mesosphere/lower thermosphere; after this time a different measurement plan was adopted that sampled the stratosphere and mesosphere during alternate orbits and viewed out of opposite sides of the spacecraft on alternate days. Under the earlier measurement pattern, mesospheric data had higher longitudinal resolution, but there was a gap of a day or more between any two days of data. During later periods the zonal resolution waspoorer, but the gaps were reduced. In addition, several special measurement campaigns were performed based on scientific considerations, and several gaps occurred due to instrument or spacecraft problems. In March 1995, the UARS solar array drive failed and data coverage became much more limited.

The limitations to the coverage that affect the present study are 1) the mesosphere is normally sampled only on alternate days or alternate orbits; 2) HRDI can measure winds only during daylight, except at a single altitude (∼95 km); and 3) for observations from one side of the spacecraft, the latitudinal coverage is limited to about 40° in one hemisphere and 72° in the other. The orbital pattern and daylight constraint limit the ability of HRDI to observe the high latitudes of the winter hemisphere. This affects the ability to investigate the impact of stratospheric planetary waves on the mesosphere since these waves are large primarily during winter. There are, however, sufficient observations during late winter (NH in February–March, SH during August–September) to provide meaningful information.

The HRDI mesospheric winds contain large contributions from the diurnal and semidiurnal tides (Morton et al. 1993; Hays et al. 1994; Burrage et al. 1995). For this study, we need to separate the tides from other perturbations to the zonal mean. In the higher latitudes (poleward of 40°) where HRDI can view from only one side of the spacecraft, all data at a given latitude and day are taken at approximately the same local time. Since data at all longitudes (a single local time) contain the same contribution from the migrating tides, the longitudinal average will include the sum of winds due to tides as well as the true zonal mean. Therefore, removal of the longitudinal mean will remove most of the migrating tidal contribution, as well as removing the actual zonal mean wind. The removal of migrating tides may be incomplete because of the small number of profiles at the higher latitudes (sometimes as few as 6, always less than 17, per latitude per day). In addition, wind variations due to nonmigrating tides will not be removed and may account for some of the variability of the HRDI winds.

This study uses the Level 3at HRDI data. The winds in this data product have been interpolated to regular time intervals and are at fixed altitude intervals (3-km intervals from 60 to 114 km, with an additional level at 55 km). Longitude positions follow the tangent points and are not at fixed intervals. All the data from a single day are considered together. For computation and removal of the longitudinal mean, data from each day and altitude are collected into latitude bins with 4° width. Data from any bin with less than 6 points is discarded since a meaningful longitudinal average cannot be obtained.

It was necessary to have simultaneous measurements of stratospheric and mesospheric winds. Since HRDI does not normally view the stratosphere and mesosphere on the same orbits or days, and since the orbit track is not repeated from one day to the next, the HRDI stratospheric wind data do not coincide either in time or in horizontal location with the mesospheric wind data. Therefore, the HRDI stratospheric winds were not used. The stratospheric observations are from the NMC analyses (Gelman et al. 1986). Horizontal winds were computed from the balance wind formula of Randel (1987), using data and algorithms provided by W. Randel. Data are interpolated in latitude and longitude to lie directly under HRDI wind profiles. The analysis interpolates in the horizontal direction to obtain profiles that coincide horizontally with a HRDI mesospheric profile for the same day and uses only these in the analysis.

3. Planetary-scale variations in the winter mesosphere

This section will present HRDIresults for the middle and high latitudes during winter, with emphasis on the differences between the Northern and Southern Hemispheres. Southern Hemisphere data for the month of August are examined in detail and can be compared with NH data for February presented in Smith (1996) and with additional NH analyses in the present paper.

The first step is to determine whether a similar anticorrelation between stratospheric and mesospheric winds to that found in the NH also occurred in the SH. In these calculations, all HRDI data for a given month are separated into latitude and height bins with resolutions of 4° and 3 km. If there are at least six profiles for a given day in the bin, the longitudinal average is calculated and removed, leaving a set of perturbation winds; otherwise, the data for that bin is discarded. Each HRDI data point is matched with the vertical average of the stratospheric wind for that day, longitude and latitude. The vertical average is defined as the average of the maximum and minimum winds over the pressure interval from 100 to 1 mb. The correlations between the HRDI zonal and meridional winds and the corresponding NMC vertical average winds are computed. The February zonal wind correlations (Smith 1996) showed a large region of negative correlation from 30° to 60°, peaking at about 78–81 km with a (negative) correlation coefficient of greater than .5 in magnitude. Meridional wind correlations were appreciably weaker but nevertheless indicated a substantial region of negative correlation. The zonal wind also showed a region of positive correlation above 90 km for some periods.

Figure 1 shows the comparable results for August in the SH. The peak magnitude of the negative correlation is even greater than that seen for February in the NH, but the latitudinal extent is smaller. In particular, large and statistically significant zonal wind correlation values in August are found primarily in latitudes lower than 50°. This strong decrease in the magnitude of the negative correlation at high latitudes is a consistent characteristic for all three years, although data for August 1992 does not extend to sufficiently high latitudes to confirm the decrease. During September 1992 and 1993 (Fig. 2), when there were high-latitude observations available, the latitudinal extent of the negative correlation center is similar to that during August.

Figure 3 shows the amplitude and phase of the monthly average wave 1 in zonal and meridional wind for the three Augusts. To calculate this wave structure, HRDI perturbation wind data were sorted into 16 longitude bins and averaged for the month; the wave amplitude and phase were determined by Fourier analysis. The results are shown at 40°S, which is near the mesospheric maximum in wave amplitude. Although there are amplitude minima between the upper stratosphere and the upper mesosphere, particularly during 1993, the amplitude does not diminish to zero. There are, however, minima in the zonal wind amplitude in the upper mesosphere (85–90 km) during all three years. A westward vertical phase tilt with height extends to the upper mesosphere. Meridional wave amplitudes are smaller but, during 1993 and 1994, do not approach zero in the lower mesosphere. Minima do occur but they are less pronounced than those reported previously (Smith 1996) for February. Phase variations with altitude are moderate in the mesosphere, except during 1992; this indicates that contamination by incomplete removal of the diurnal tide, which has a vertical wavelength of about 20 km (Hays et al. 1994), is probably not a main contributor.

In the NH winter observations (Smith 1996), a distinct minimum in wave amplitude in the lower mesosphere, particularly in the meridionalwind, was used as one piece of evidence supporting the conclusion that the mesospheric planetary-scale perturbation was unlikely to be an upward extension of the stratospheric planetary wave. Since the minimum is reduced or absent for the August SH observations, we cannot conclude that the stratospheric planetary wave disappears in the lower mesosphere without additional information.

Figure 4 shows the perturbation zonal wind structure as a function of longitude and altitude for late winter: February 1994 in the NH and August 1994 in the SH. For the February case, wavenumber 2 is dominant in the stratosphere but is small in the lower levels of HRDI mesospheric data (60–70 km). Regions of stronger perturbation wind occur above 70 km that are out of phase with the large perturbation winds in the middle to upper stratosphere. During August, the regions of maximum and minimum wind form a coherent pattern extending from the stratosphere to about 85 km, near the mesopause. There is a sharp transition to steeper vertical phase tilt in the mesosphere but no break in the amplitude.

Figure 4 is an example of the hemispheric differences found in the wave structure in the mesosphere during late winter. In the NH, there is not a clear indication of a coherent propagating wave extending into the upper mesosphere. In the SH, the wave pattern shows a significant amount of coherence from the stratosphere to the mesosphere but also has some significant differences with altitude, notably a shift toward a shorter vertical wavelength. The August 1994 observations give the impression that a propagating wave extends well into the mesosphere, while those of February do not suggest a continuous link between the wavenumber 2 perturbation in the upper mesosphere and the out of phase stratospheric wave below.

The correlation between u′ and υ′ gives an indication of the meridional component of the EP flux (Fyuυ). For the SH, a negative correlation indicates poleward momentum flux or equatorward EP flux. Figure 5 shows the correlation between u′ and υ′ in the SH for the three August periods. For all three, there is a region of negative correlation around 30°–40°S in the upper mesosphere. This is another similarity between the global-scale variation in the SH and a propagating Rossby wave.

The latitude extent of the statistically significant negative correlation (Figs. 1 and 2) is consistent with the gravity wave filtering hypothesis, as can be seen from a comparison of the stratospheric flow in the NH winter and the SH winter. Horizontal wind vectors at 1 mb for two representative days are illustrated in Fig. 6. In the NH, wind perturbations at all latitudes show variations in which the zonal wind switches sign and becomes easterly at some longitudes. In the SH, the wind variations are smaller and, at latitudes greater than 50°, never result in an easterly or extremely weak westerly zonal wind speed. Winds in the lower stratosphere are even more zonally symmetric.

The filtering will be most effective if the wind variations are substantial in the range of speeds corresponding to the most energetic part of the gravity wave spectrum. Topographically forced gravity waves will have a range of phase speeds that would normally be expected to be centered near weak or zero speeds with respect to the earth’s surface. Wind variations in the range of large to moderate westerlies will filter only those waves with westerly phase speeds, which are at one end of the spectrum and carry only part of the total energy. On the other hand, if the winds in the stratosphere range fromwesterly to weak easterly, the filtering affects a more energetic part of the spectrum and therefore introduces much larger asymmetries in the energy input to the mesosphere. By this argument, correlations between stratospheric and mesospheric winds due to gravity wave filtering depend not only on the absolute planetary wave amplitude, but also on the amplitude relative to the zonal mean wind.

For the NH results in Smith (1996), the fact that the wave structure in the stratosphere differed during the three years examined provided diagnostic information. For example, February 1994 was dominated by wave 2, while during 1992 and 1993 wave 1 was dominant. The match between the stratospheric wavenumber and that in the mesosphere was evidence that the primary source of the mesospheric asymmetries was not geographically fixed. The same type of information is not available in the SH. Wave 1 with a similar phase was dominant in each of the three years investigated and there was therefore no information that could be used to determine the relative magnitude of any fixed perturbation that might be present.

Several significant interhemispheric differences emerge from these data: 1) the perturbation zonal wind structure in the SH indicates coherence between the stratosphere and mesosphere but that in the NH does not support this; 2) the amplitude minimum in meridional wind that was seen in NH February, particularly in the meridional wind, is absent or reduced in the SH August results; and 3) the correlation between u′ and υ′ indicates an equatorward EP flux in the upper mesosphere in the SH, whereas this did not appear in the NH.

Overall, the August SH data do not support the NH evidence used to argue against a vertically propagating Rossby wave as the cause of the stratospheric-mesospheric wind anticorrelation. Section 4 presents the results of a model used to provide another approach for trying to shed light on the planetary-scale structure in the upper mesosphere.

4. Wave model using observed winds

Although it is only an approximation of what is happening in the real atmosphere, a stationary wave model is a useful tool because it allows one to see what type of dynamical structure is consistent with a particular background wind structure. The model used here is a linearized quasigeostrophic wave model of the middle atmosphere. The model is based on the equations of Matsuno (1970) and is solved using the method of Lindzen and Kuo (1969). It extends from the tropopause to 120 km (near the upper limit of HRDI wind data). Two model cases are described; in the first, the wave can be dissipated within the model domain but has as its only source of forcing a geopotential perturbation specified at the lower boundary. In the second, an additional forcing term is introduced that is meant to represent the variation in the momentum drag caused by filtering of gravity waves by planetary waves in the stratosphere.

The model has only a few variable input fields, but these are important for the results. The primary ones are: 1) zonal mean wind; 2) Rayleigh friction and Newtonian cooling, which are linear damping that represent dissipation by gravity waves and diabatic processes; and 3) wave forcing at the lower boundary. For the present study the following values are used. For the zonal mean wind, the HRDI data for all days observed during a particular month over all three years are averaged and combined into a grid. All NMC data for the three years are used, regardless of whether HRDI observations were available at that latitude or date. Interpolation is used to join the two datasets and extrapolation (with zero velocity at the poles) is used to extend the HRDI winds in latitude. It is hoped that by using data from all three years, tidal components to the zonal wind will have a large degree of cancellation and can beminimized. Figure 7 shows the averaged observations and the interpolated/extrapolated model wind for August. Figure 8 shows the zonal winds for February. The lack of symmetry in the winter circulation between the two hemispheres is particularly notable in the upper stratosphere and lower mesosphere. Average winter hemisphere winds are stronger during August than during February, both in the upper stratosphere and in the mesosphere. Despite the differences in speeds, the overall locations of zero wind lines separating easterly and westerly winds have a fair degree of symmetry during these two periods, although there are differences in high latitudes above 95 km.

The Rayleigh friction and Newtonian cooling coefficients, which depend only on height, are shown in Fig. 9. Rayleigh friction damping is weak through the stratosphere and mesosphere but begins to be large around 90-km altitude. The plot also shows the timescale for internal forcing, which is discussed below.

The wave structure at the lower boundary is represented as a stationary wavenumber 1 that is sinusoidal in latitude with a width of 60°, centered at 60° latitude of the winter hemisphere. The lower boundary is at 15 km.

The two cases in the model will be referred to as the propagating wave case and the internal forcing case. In the propagating wave case, the model contains only the elements that have been described above. In the internal forcing case, an additional contribution has been added to the model formulation. The vorticity equation used to solve the model in this case is given by (Matsuno 1970)
i1520-0469-54-16-2129-e1
where ũ is the angular momentum of the basic state, ζ is the eddy relative vorticity, Z is the total vorticity of the basic state, and α is the Rayleigh friction parameter. Other symbols have their conventional meanings. The internal forcing Fi has the form
Fim
where ψm is a measure of the wave amplitude in the stratosphere relative to the background wind there. It is given by the eddy geopotential at the level of the maximum wave amplitude below 60 km, multiplied by the ratio of the eddy to zonal mean wind,
i1520-0469-54-16-2129-e3
where zm is the altitude in the stratosphere where the wave amplitude is maximum. As shown in Fig. 9, β is nonzero only between 60 and 100 km. The form of Fi was chosen to represent the observed correlations between the stratospheric and mesospheric winds. In particular, the altitude dependence of β corresponds to the altitudes of largest correlation of mesospheric with stratospheric winds (see Figs. 1–2), the dependence on the maximum stratospheric geopotential (ϕ′) selects the stratospheric planetary wave phase that corresponds to the largest wave perturbation, and the (|u′/ū|) factor gives larger forcing when the stratospheric flow variations include weak or negative total zonal winds (see Fig. 6 and discussion). The results of the wave model with internal forcing are not very sensitive to the form of Fi as long as it is out of phase with the planetary wave in the middle to upper stratosphere.

Figure 10 shows the geopotential amplitude calculated for both model cases using the mean August zonal winds. Figure 11 shows the amplitudes and phases of theperturbation zonal and meridional winds at 42°S for these two cases. In the propagating wave case, the amplitude of the meridional wind has peaks at about 45 and 90 km while the zonal wind maxima are near 30 and 90 km, with a secondary maxima at 75 km. Phase tilts are small near the wave maxima and larger at other altitudes, and for the meridional wind they are westward with height everywhere. In the internal forcing case, the results are similar at all levels; the primary difference is that the mesopause maximum in wave amplitude is slightly larger. The zonal wind model results (Fig. 11) resemble the observations (Fig. 3) in a number of ways. There are maxima in the upper stratosphere (although the model maximum is at a lower altitude), in the middle mesosphere (75 km), and above the mesopause (95 km). The 75-km peak is underestimated in the model, while the 95-km peak is overestimated. Not all of the mesospheric variations of meridional wind are simulated. Note also that the model reproduces the dramatic decrease in vertical wavelength seen in the observations (Figs. 3 and 4)

The wave model results using NH February winds, shown in Figs. 1213, are quite different. In the propagating wave case, the wave amplitude is very small above 60 km. With the observed wind structure (Fig. 8), the planetary wave is unable to propagate vertically into the mesosphere, even though the damping rate is small. For this wind structure, there is a large difference in the mesospheric wave pattern between the propagating wave case and the internal forcing case. In the latter, another maximum appears in the upper mesosphere, and the model results more closely resemble the observations. For comparison, observed wave 1 structures from HRDI and NMC for February 1993 and 1994 are shown in Fig. 14. For wave 1 structure for 1992 and wave 2 for 1994, see Smith (1996, Figs. 12 and 13 respectively). In all the February cases, as well as in the results of the wave model, there is a minimum in the wave amplitude in the lower to middle mesosphere.

The differences in the August and February model simulations are due entirely to the background zonal wind structure. According to linear wave theory, the structure (magnitude and curvature) of the background wind determines whether wave propagation is directed vertically or meridionally. Wave activity that is focused equatorward will not penetrate to as high altitudes. In addition, damping is proportionately more effective when wind speeds are weaker.

The wave model with internal forcing is able to reproduce a number of aspects of the observations for both February and August. Without internal forcing, the model does an adequate job of reproducing the observations of August but not those of February. These wave model results support the conclusion of Smith (1996) that the mesospheric wave observed in the NH during February was caused primarily by in situ forcing due to dissipation of gravity waves. During August in the SH, however, the results are consistent with a planetary-scale wave in the mesosphere that has propagated upward from the stratosphere. The mechanism of filtering gravity waves in the stratosphere may be active but does not have a large enough impact on the observed wave structure to be identifiable.

The discussion has concentrated on the months of August (SH) and February (NH), primarily because there are more data for these than for other winter months, particularly at middle and high latitudes. It is also the case that the conclusions are consistent for all the years examined and there aredistinct differences between the mean winds during these two periods. The August stratospheric jet is much stronger and reaches to higher altitude than that during February.

During December in the NH, the upper stratospheric zonal wind is stronger and bears some resemblance to the flow pattern seen in August in the SH. The primary differences are that in December the jet maximum occurs at higher altitude and that the maximum winds in the stratosphere occur at higher latitudes. Unfortunately, the HRDI winds are not available at high latitudes, so the jet structure in the mesosphere is not well characterized. Figure 15 shows the wave structure at 40°N for three years. During 1992 and 1994, the wave structure has in common with that of August 1994 in the SH (Fig. 3) that the amplitudes are large through the middle atmosphere, with a minimum near 80 km. The meridional winds extend to the upper mesosphere without a minimum or discontinuity. The vertical phase tilts are much more pronounced during December than either August (SH) or February (NH).

On the other hand, in the SH during October, the stratospheric jet has weakened compared to that of August and the transition from westerlies to easterlies occurs at a lower altitude, so the winds have some similarities to those during February in the NH. Therefore, one might expect that the wave patterns during these periods might be intermediate between those seen in February and in August. During October, the SH wind reversal from westerly to easterly occurs between 50 and 60 km (between the NMC and HRDI observations), and therefore no planetary wave propagation into the mesosphere is expected. Figure 16 shows the wave 1 amplitude for October 1992, which resembles the pattern seen in February in that there is a minimum in the lower mesosphere and another maximum in the upper mesosphere that is out of phase with that in the stratosphere.

5. Conclusions

Large, persistent zonal asymmetries appear in the HRDI wind measurements of the upper mesosphere during winter. These are anticorrelated with longitudinal asymmetries in the stratospheric winds, which are associated with quasi-stationary planetary Rossby waves. Winds in late winter are analyzed to determine the nature of their relationship with the stratospheric winds.

The analysis suggests that during August (Southern Hemisphere late winter) planetary wavenumber 1 propagates to the mesopause. In contrast, during February (Northern Hemisphere late winter) the planetary waves propagate only as far as the lower mesosphere. However, the large variations in stratospheric winds filter gravity waves and hence introduce a longitudinal asymmetry in the spectrum of gravity waves that reaches, and dissipates in, the upper mesosphere. The gravity wave dissipation can generate planetary-scale variations in the mesospheric winds, and this is postulated to be the cause of the observed asymmetries during NH winter. The February case is described in more detail in Smith (1996).

The limited data available for NH midwinter (December) indicates that this period bears some resemblance to the SH August results in that the planetary wave appears to be continuous from the stratosphere through to the mesosphere. Spring in the SH (October) also has a period during which the planetary wave does not propagate into the mesosphere, and the gravity wave filtering mechanism therefore is assumed to be the cause of the mesospheric wave.

There are several features that go into the conclusion that the primary source of the upper mesospheric wave structure is not the same in all months and that there are systematic differences between the two hemispheres.

  1. In the SH winter, the wave characteristics such as amplitude and phase are continuous through the mesosphere, whereas in the NH late winter there is a minimumin wave amplitude in the lower or middle mesosphere.
  2. The horizontal component of the EP flux is equatorward in the SH winter in the upper mesosphere, characteristic of a Rossby wave, but in the NH upper mesosphere, the horizontal component of the EP flux is small and irregular in sign.
  3. A linear wave model using observed mean zonal winds indicates that during August waves propagate up to the mesopause but during February the wave decays strongly above the stratopause. The model was able to reproduce the NH observations when an additional forcing was included in the mesosphere that was out of phase with the stratospheric winds. This additional forcing had almost no impact on the August wave structure: first, because it was smaller since the stratospheric wind variations were smaller and, also, because it tended to be in phase with the propagating wave.
  4. Because of the larger background wind in the SH stratosphere, planetary waves there do not often give rise to local easterly winds, whereas they are common in the NH winter stratosphere. If the spectrum of gravity waves is centered near zero with respect to the earth’s surface, the longitudinal differences in filtering will affect a more energetic portion of the spectrum in the NH than in the SH and, as a consequence, will have a larger impact in the upper mesosphere.

Acknowledgments

I would like to thank B. Khattatov and R. Garcia for helpful discussions and comments. This research was supported by the NASA UARS Guest Investigator Program under Grant S-12896F. The National Center for Atmospheric Research is supported by the National Science Foundation.

REFERENCES

  • Abreu, V. J., P. B. Hays, D. A. Ortland, W. R. Skinner, and J.-H. Yee, 1989: Absorption and emission line shapes in the O2 atmospheric bands: Theoretical model and limb viewing simulations. Appl. Opt.,28, 2128–2137.

  • Bacmeister, J. T., 1993: Mountain-wave drag in the stratosphere and mesosphere inferred from observed winds and a simple mountain-wave parameterization scheme. J. Atmos. Sci.,50, 377–399.

  • Burrage, M. D., and Coauthors, 1993: Comparison of HRDI wind measurements with radar and rocket observations. Geophys. Res. Lett.,20, 1259–1262.

  • ——, D. L. Wu, W. R. Skinner, D. A. Ortland, and P. B. Hays, 1995: Latitude and seasonal dependence of the semidiurnal tide observed by the High Resolution Doppler Imager. J. Geophys. Res.,100, 11 313–11 321.

  • ——, and Coauthors, 1996: Validation of mesosphere and lower thermosphere winds from the High Resolution Doppler Imager on UARS. J. Geophys. Res.,101, 10 365–10 392.

  • Dunkerton, T. J., and N. Butchart, 1984: Propagation and selective transmission of internal gravity waves in a sudden warming. J. Atmos. Sci.,41, 1443–1460.

  • Fritts, D. C., 1984: Gravity wave saturation in the middle atmosphere: A review of theory and observations. Rev. Geophys. Space Phys.,22, 275–308.

  • Gelman, M. E., A. J. Miller, K. W. Johnson, and R. M. Nagatani, 1986: Detection of long term trends in global stratospheric temperature from NMC analyses derived from NOAA satellite data. Adv. Space Res.,6(10), 17–26.

  • Hays, P. B., V. J. Abreu, M. E. Dobbs, D. A. Gell, H. J. Grassl, and W. R. Skinner, 1993: The High Resolution Doppler Imager on the Upper Atmosphere Research Satellite. J. Geophys. Res.,98, 10 713–10 723.

  • ——, D. L. Wu, and the HRDI Science Team, 1994:Observations of the diurnal tide from space. J. Atmos. Sci.,51, 3077–3093.

  • Holton, J. R., 1982: The role of gravity wave induced drag and diffusion in the momentum budget of the mesosphere. J. Atmos. Sci.,39, 791–799.

  • ——, 1984: The generation of mesospheric planetary waves by zonally asymmetric gravity wave breaking. J. Atmos. Sci.,41, 3427–3430.

  • Lindzen, R. S., 1981: Turbulence and stress owing to gravity wave and tidal breakdown. J. Geophys. Res.,86, 9707–9714.

  • ——, and H.-L. Kuo, 1969: A reliable method for the numerical integration of a large class of ordinary and partial differential equations. Mon. Wea. Rev.,97, 732–734.

  • Matsuno, T., 1970: Vertical propagation of stationary planetary waves in the winter Northern Hemisphere. J. Atmos. Sci.,27, 871–883.

  • McLandress, C., and N. A. McFarlane, 1993: Interactions between orographic gravity wave drag and forced planetary waves in the winter Northern Hemisphere middle atmosphere. J. Atmos. Sci.,50, 1966–1990.

  • Miyahara, S., 1985: Suppression of stationary planetary waves by internal gravity waves in the mesosphere. J. Atmos. Sci.,42, 100–107.

  • Morton, Y. T., and Coauthors, 1993: Global mesospheric tidal winds observed by the High Resolution Doppler Imager on board the Upper Atmosphere Research Satellite. Geophys. Res. Lett.,20, 1263–1266.

  • Ortland, D. A., P. B. Hays, W. R. Skinner, M. D. Burrage, A. R. Marshall, and D. A. Gell, 1995: A sequential estimation technique for recovering atmospheric data from orbiting satellites. The Upper Mesosphere and Lower Thermosphere, Geophys. Monogr., Ser. No. 87, R. M. Johnson and T. L. Killeen, Eds., Amer. Geophys. Union.

  • Randel, W. J., 1987: The evaluation of winds from geopotential height data in the stratosphere. J. Atmos. Sci.,44, 3097–3120.

  • Schoeberl, M. R., and D. F. Strobel, 1984: Nonzonal gravity wave breaking in the winter mesosphere. Dynamics of the Middle Atmosphere, J. R. Holton and T. Matsuno, Eds., Terra Scientific, 45–64.

  • Smith, A. K., 1996: Longitudinal variations in mesospheric winds: Evidence for gravity wave filtering by planetary waves. J. Atmos. Sci.,53, 1156–1173.

Fig. 1.
Fig. 1.

Correlation of HRDI zonal (left) and meridional (right) winds with the vertical average of NMC stratospheric winds for the same latitude, longitude, and date for August 1992, 1993, and 1994. The shaded areas indicate where the correlation is significant at the 95% level.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 2.
Fig. 2.

As in Fig. 1 but for September 1992, 1993, and 1994.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 3.
Fig. 3.

Average zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) from NMC and HRDI at 40°S for August 1992, 1993, and 1994.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 4.
Fig. 4.

Longitude × altitude cross sections of perturbation zonal winds from NMC and HRDI averaged for the month of February 1994 (40°–50°N) and August 1994 (30°–40°S).

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 5.
Fig. 5.

Correlation of the longitudinally asymmetriczonal and meridional HRDI winds at the same locations and times for August of 1992–94.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 6.
Fig. 6.

Longitude × latitude cross sections of wind vectors computed from NMC at 1 mb on 24 February 1993 and 24 August 1994.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 7.
Fig. 7.

Average of August zonal wind speeds from NMC (below 48 km) and HRDI (above 60 km) for all available data from August 1992–94. On the right is the complete model zonal wind field obtained by interpolation/extrapolation.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 8.
Fig. 8.

As in Fig. 7 but for February.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 9.
Fig. 9.

Rayleigh friction coefficient (solid line), Newtonian cooling coefficient (short dashed line), and timescale for internal forcing (long dashed line) used in the linear wave model.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 10.
Fig. 10.

Latitude × altitude cross section of the geopotential wave amplitude (in meters) calculated by the linear wave model for the propagating wave case (left panel) and the internal forcing case (right panel), using the August mean winds.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 11.
Fig. 11.

Zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) from the linear wave model for August at 42°S for the propagating wave case (upper panels) and internal forcing case (lower panels).

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 12.
Fig. 12.

As in Fig. 10 but using the February mean winds.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 13.
Fig. 13.

As in Fig. 11 but for February at 54°N.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 14.
Fig. 14.

Average zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) at 52°N from NMC and HRDI for February 1993 and 1994.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 15.
Fig. 15.

Average zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) at 40°N from NMC and HRDI for December 1992, 1993, and 1994.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

Fig. 16.
Fig. 16.

Average zonal (left) and meridional (right) wind wave 1 amplitude (solid line; scale at bottom) and phase (open circles; scale at top) at 40°S from NMC and HRDI for October 1992.

Citation: Journal of the Atmospheric Sciences 54, 16; 10.1175/1520-0469(1997)054<2129:SPWIUM>2.0.CO;2

* The National Center for Atmospheric Research is sponsored by the National Science Foundation.

Save