A study of a self-generated stratospheric sudden warming and its mesospheric–lower thermospheric impacts using the coupled TIME-GCM/CCM3



[1] A stratospheric sudden warming episode was self-consistently generated in the coupled National Center for Atmospheric Research Thermosphere, Ionosphere, Mesosphere, and Electrodynamics General Circulation Model/Climate Community Model version 3 (TIME-GCM/CCM3). Taking advantage of the unique vertical range of the coupled model (ground to 500 km), we were able to study the coupling of the lower and upper atmosphere in this warming episode. Planetary wave 1 is the dominant wave component in this warming event. Analysis of the wave phase structure and the wave amplitude indicates that the wave may experience resonant amplification prior to the peak warming. The mean wind in the high-latitude winter stratopause and mesosphere decelerates and reverses to westward due to planetary wave forcing and forms a critical layer near the zero wind lines. The wind deceleration and reversal also change the filtering of gravity waves by allowing more eastward gravity waves to propagate into the mesosphere and lower thermosphere (MLT), which causes eastward forcing and reverses the westward jet in the MLT. This also changes the meridional circulation in the upper mesosphere from poleward/downward to equatorward/upward, causing a depletion of the peak atomic oxygen layer at 97 km and significant reduction of green line airglow emission at high latitudes and midlatitudes. Planetary waves forced in situ by filtered gravity waves in the MLT grow in the warming episode. Their growth and interaction with tides create diurnal and semidiurnal variabilities in the zonal mean zonal wind.

1. Introduction

[2] A stratospheric sudden warming (SSW) is a dramatic event in the winter middle atmosphere which involves profound changes of temperature, wind, and circulation in a short period of time. It has been under extensive observational and theoretical investigation for many years and reviewed, for example, by Andrews et al. [1987], McIntyre [1982], Holton [1980], and, more recently, Labitzke and van Loon [1999]. The key mechanism, initially proposed by Matsuno [1971] and now widely accepted, is the growth of upward propagating planetary waves from the troposphere and the interaction between the transient wave and the mean flow. The interaction decelerates and/or reverses the eastward winter stratospheric jet and also induces a downward circulation in the stratosphere causing adiabatic heating and an upward circulation in the mesosphere causing adiabatic cooling. Further studies have then concentrated on the essential ingredients of this mechanism, namely, the amplification of planetary waves and the planetary wave and mean flow interaction.

[3] Possible mechanisms for the planetary wave amplification have also been studied. Tung and Lindzen [1979a, 1979b] used a linear model to describe resonant Rossby waves for atmospheres with both idealized and realistic wind structures, and found that resonant wave (wave number 1 and 2) amplification is possible under winter conditions. They also proposed a resonant condition relating the altitude of the wave “turning point” (roughly corresponding to the altitude below which the wave is trapped) to the vertical wavelength. Resonant wave amplification has also been studied by Smith and Avery [1987]. They were able to identify a resonant wave 2 mode for the 1979 major warming using a linear model and realistic atmosphere. To obtain the resonance solution, they found that wave disturbances at the upper troposphere were not maintained by waves propagating from the surface but likely due to wave 2 blocking. Wave resonance by a different mechanism has been proposed by Plumb [1981], who showed that a tropospheric stationary wave and a traveling planetary wave aloft may engage in a self-tuning near resonance that leads to wave amplification.

[4] Propagating planetary waves and their interaction with the mean flow leading to SSW have been studied in detail using Lagrangian and quasi-Lagrangian diagnostics on atmospheric measurements and model results to gain more insight into the dynamics involved in SSW. Palmer [1981] found that variations in the high-latitude jet core in the winter stratosphere can alter the refractive index in such a way that it can favor the propagation and growth of low wave number planetary waves in the high-latitude middle atmosphere. He suggests that this may be an important precondition for a SSW to develop. A study by Dunkerton et al. [1981] also showed the importance of the initial wind in producing SSW through a diagnostic analysis of a warming modeled by Hsu [1980]. Butchart et al. [1982] designed a series of numerical experiments using a full nonlinear model with a realistic atmosphere background, and imposed planetary waves at the lower boundary to study the 1979 major warming and found that the anomalously high-latitude jet core is essential for the warming to occur. The experiments also showed that it is important for the imposed wave 2 to slowly propagate eastward. On the other hand, the wave-wave interaction was found to play no significant role in the numerical experiments.

[5] There is a mesospheric cooling accompanying the SSW as shown by Matsuno's [1971] modeling study. Evidence of the mesospheric cooling were also found in observational studies [Gregory and Manson, 1975; Myrabø et al., 1984; Whiteway and Carswell, 1994]. More recently, Walterscheid et al. [2000] studied the 1993 warming by comparing lidar and airglow mesospheric temperature measurements over Eureka, Canada (80°N) and NCAR Thermosphere, Ionosphere, Mesosphere, and Electrodynamics General Circulation Model (TIME-GCM) simulations that used the 1993 NCEP analysis geopotential height for its lower boundary forcing. The mesospheric cooling was observed in the OH airglow temperature measurement, and the comparative model simulation indicated alternating regions of cooling and warming above the main warming in the lower stratosphere.

[6] In the current study, the flux coupled NCAR TIME-GCM and Community Climate Model version 3 (TIME-GCM/CCM3) model system is used to further study a SSW and its impact on the MLT. Because the planetary waves and their amplification are self-consistently resolved in the model rather than imposed as an external forcing at the lower boundary, the simulated warming occurs spontaneously. Therefore, it is possible to treat the warming event in a consistent and integrated fashion and perform detailed analysis of the planetary wave amplification, propagation, interaction with the mean flow, and SSW impacts in the MLT.

[7] The rest of the paper is arranged as follows: A detailed description of the coupled NCAR TIME-GCM/CCM3 is given in section 2. Then in section 3 analyses of different aspects of the simulation are presented, including general features of the simulated warming, planetary wave amplification, planetary wave and mean flow interactions, and mesospheric and lower themospheric impacts. The paper concludes in section 4.

2. TIME-GCM/CCM3 Flux-Coupled Model

2.1. Flux Coupling of the Thermospheric and Lower Atmosphere Models

[8] To obtain some insight of how the variability of the lower atmosphere affects the upper atmosphere, the TIME-GCM has been flux coupled to the NCAR Community Climate Model CCM3. The CCM3, described by Kiehl et al. [1998], is a spectral model with a horizontal T42 spectral resolution (approximately 2.8° × 2.8° transform grid) and has 18 hybrid levels extending from the ground to the 2.9 hPa level (∼40 km) which change from terrain following coordinate at the ground to pressure coordinate above 83 hPa. The model time step is 20 minutes and includes a diurnal cycle in which radiative fluxes are calculated every hour.

[9] TIME-GCM is a finite difference grid point model with 4th order horizontal differencing on a 5° × 5° latitude/longitude grid. It has 45 pressure surfaces extending from 10 hPa (about 30 km height) to above 500 km with a vertical resolution of 2 grid points per scale height and a model time step of 5 minutes. Leapfrog scheme is used for time integration of advection and implicit scheme is used for vertical diffusion. Details of the numerical framework of the model are given by Dickinson et al. [1981] and Washington and Williamson [1977]. For the flux-coupled mode the lower boundary has been raised to the 2.9 hPa level, the upper boundary of CCM3. It includes a diurnal cycle for all chemical species and physical processes.

[10] To couple the two models a message passing flux-coupler is used to synchronize the model time steps and provide the interpolation of quantities in both time and space that are passed between the two models as described by Roble [2000]. Thus information at the CCM3 upper boundary is transferred to the lower boundary of TIME-GCM and vice versa. The physical quantities used in the transfer are temperature, zonal and meridional winds, geopotential height and the mass mixing ratios of water vapor and methane. Since the time constants are much longer in CCM3, the combined models are started from a 10 year run of CCM3 in a stand-alone simulation.

[11] Except for certain long-lived chemical species in the TIME-GCM, the temperature and dynamics of the middle and upper atmosphere adjust to an imposed lower boundary forcing in about 20 days of simulation time. Therefore, the coupled system is allowed to adjust for several months before histories are recorded for analysis. The primary motivation for this initial investigation is to determine how variability generated in the lower atmosphere propagates into the upper atmosphere and ionosphere. The replacement of the “rigid lid” upper boundary of CCM3 with the TIME-GCM does affect the upper stratosphere layers within CCM3 somewhat, but generally these effects do not propagate deeply into the stratosphere.

[12] The flux-coupling of two dissimilar models at an interface in the free atmosphere is an exploratory exercise to obtain some idea on how a self-consistent model of the entire atmosphere would behave. It is basically a feasibility study to determine just how processes in the lower atmosphere affect the upper atmosphere. From previous studies of the upper atmosphere using the TIME-GCM only, it is clear that solar and auroral variability alone cannot represent the variability observed by ground-based and satellite instruments, especially the day-to-day variability in the upper mesosphere and lower thermosphere. Another motivation for the development of a General Circulation Model of the entire atmosphere is to examine how deep solar-terrestrial effects propagate into the Earth's atmosphere. Here we use data output for the first 2 months of “year 4” of the coupled model simulation, with constant solar activity and no geomagnetic disturbances superimposed.

2.2. Version of TIME-GCM

[13] The NCAR TIME-GCM is the latest in a series of three-dimensional time-dependent models that have been developed over the past two decades to simulate the circulation, temperature, and compositional structure of the upper atmosphere and ionosphere. It combines all the previous features of the TGCM [Dickinson et al., 1981, 1984], TIGCM [Roble et al., 1988], and TIE-GCM [Richmond et al., 1992]. The model has been extended downward to 30 km altitude, including aeronomic processes appropriate for the mesosphere and upper stratosphere, as described by Roble and Ridley [1994], Roble et al. [1987] and Roble [1995]. The differences between the model described in previous papers and that used for the present work include the following:

  1. Solar ionization rates are calculated using the EUVAC solar flux model and absorption cross-sections from Richards et al. [1994]. Solar photodissociation rates for the mesosphere and upper stratosphere are determined using the parameterizations given by Brasseur and Solomon [1986] and Zhao and Turco [1997].
  2. The chemical reaction rates for the aeronomic scheme described by Roble [1995] have been updated to be consistent with those of DeMore et al. [1997].
  3. The background diffusion used by Roble and Ridley [1994] has been reduced by two orders of magnitude, consistent with the findings of Akmaev et al. [1996] in their simulation of the diurnal tide. Because the parameterized gravity wave forcing does not appear to provide the necessary drag on the jet between 30–60 km, Rayleigh friction is used. The Rayleigh friction has a small value during equinox (∼10−8 s−1) and a larger value during solstice (∼10−7 s−1) with a sinusoidal variation in between.
  4. The CO2 infrared cooling parameterization has been updated to include the model of Fomichev et al. [1998] to account for a variable CO2 mixing ratio, important for nonlocal thermal equilibrium (non-LTE) processes in the upper mesosphere and lower thermosphere. All calculations assume an O-CO2 vibrational relaxation rate of 3 × 10−12 cm−3 s−1, which works reasonably well for terrestrial planetary thermospheres [Bougher et al., 1999].
  5. The gravity wave parameterization is based on Lindzen [1981]. The CCM version used in this coupling study does not have propagating gravity wave components, thus the gravity wave parameters are specified at the lower boundary of TIME-GCM according to our studies using the uncoupled model: A discrete spectrum of gravity waves, with phase speed from −60 to 60 ms−1 and 10 ms−1 intervals, are specified at the lower boundary of the TIME-GCM in both zonal and horizontal directions. The spectral shape is Gaussian, similar to those used in the Middle Atmosphere CCM (MACCM) [Boville, 1995]. Our study using the uncoupled TIME-GCM indicates that the zonal gravity wave spectrum at the model lower boundary needs to be anisotropic with the Gaussian peak at eastward 10 ms−1 (meridional spectrum still isotropic), so that the simulated wind agrees with the UARS wind measurements [McLandress et al., 1996]. The orographic component is adapted from the calculation by McFarlane [1987], primarily in the Northern Hemisphere during winter.
  6. For the uncoupled TIME-GCM, it is necessary to specify tidal forcing at the lower boundary near 30 km, due to diurnal and semidiurnal components excited in the troposphere. However, these tidal components are resolved by the model coupling and it is thus unnecessary to specify tidal forcing at the lower boundary of TIME-GCM.

[14] TIME-GCM calculates neutral gas heating, photoionization, and the compositional structure of the middle and upper atmosphere and ionosphere for a given solar irradiance spectrum, which in the present version is computed for a 10.7 cm solar radio flux (F10.7) and is held constant throughout the model year. The geomagnetic activity held constant throughout the year at a very quiet level (Ap = 4), and the model includes the appropriate level of auroral precipitation, cross-cap potential fields, and hemispheric power [Roble and Ridley, 1987].

3. Analysis of Simulation Results

3.1. General Description of the Simulated Warming

[15] The calculated zonal mean temperature at 72.5°N is shown in Figure 1a for January and early February. A stratospheric warming can be clearly seen during this period of time. The warming is mainly below 60 km and starts around day 13 (January 13). It peaks on day 25 and the temperature at 45 km increases by more than 25 K within 12 days. At the same time, the mesosphere (above 60 km) becomes cooler and at 80 km the temperature decreases by about 40 K. There is also an indication of warming in the lower thermosphere (105–120 km). At the early stage of the event (day 13–20), the cooling in the mesosphere is quite significant compared with the warming in the stratosphere. For example, the cooling rate at 80 km on day 15 is ∼−3.2 Kd−1 while the heating rate at 45 km is ∼2.2 Kd−1. There is a slight cooling at 72.5° starting as early as day 8 near 105 km as shown in Figure 1a, but the cooling rate is insignificant before day 13. By comparing Figures 1a and 1c, it is found that this cooling corresponds to the slight decrease of the jet around day 8. The temperature increase in the stratosphere at the same time is barely noticeable. We will examine if this early cooling in the mesosphere is related to the SSW event in later sections.

Figure 1.

(a) Calculated zonal mean temperature (K) at 72.5°N from day 1 to day 36 between 10 and 120 km. The contour interval is 5 K for temperatures below 260 K, and 10 K for higher temperatures. (b) Zonal mean temperature differences between 87.5°N and 62.5°N from day 1 to day 36 between 10 and 200 km. (c) Zonal mean zonal wind (ms−1) at 62.5°N from day 1 to day 36 between 10 and 200 km.

[16] Figure 1b shows the calculated zonal mean temperature difference between 87.5° and 62.5° (T(87.5°)−T(62.5°)) from day 1 to day 36. Starting from day 12, the poleward temperature drop in the stratosphere and lower thermosphere and the temperature increment in the mesosphere begin to decrease. In the stratosphere, the temperature drop reverses sign between day 18 and 21. At the peak of the warming (day 25), a temperature increment of 15 K is seen at 45 km. In the upper atmosphere, the temperature increment first begins to reverse sign on day 16 at 100 km and then the reversal height extends downward into the mesosphere at a rate of about 3 km d−1. On day 26, the temperature drop reaches 20 K at 85 km. An opposite reversal occurs in the lower thermosphere with 10 K drop. The mean zonal wind changes at 62.5°, as shown in Figure 1c, are closely correlated with the changes of the meridional temperature gradient shown in Figure 1b, with the wind reversal first appearing at higher altitudes and then extending downward.

[17] The global structures of the calculated zonal mean temperature and zonal mean zonal wind are examined on day 10 (prior to the warming) and 25 (peak warming) as shown in Figure 2. Both change significantly in the winter (northern) hemisphere high-latitudinal region. On day 10, our control case before the SSW, the zonal mean temperature structure displays a cold winter stratosphere and warm winter mesosphere and mesopause, with the winter mesopause near 100 km and summer mesopause near 90 km. The zonal mean eastward jet in the winter stratosphere/mesosphere peaks at 60°N and 60 km altitude with the maximum value larger than 70 ms−1. The jet core is at its climatological winter solstice position, rather than at the higher latitude (75°N) position indicated to be essential for the wave 2 major warming in 1979 by Butchart et al. [1982]. On day 25, the peak stratospheric temperature in the winter polar region is more than 260 K at 45 km, which is more than 40 K warmer than that at the same location on day 10. The mesospheric temperature, on the other hand, decreases by more than 50 K at 80 km. The meridional temperature gradient in the polar region at these altitudes also changes significantly. The eastward jet in the stratosphere and mesosphere is reduced and between 40 and 80 km it reverses to westward, while an opposite change occurs in the lower thermosphere.

Figure 2.

Zonal mean temperature (K) on (a) day 10 and (b) day 25, and zonal mean zonal wind (ms−1) on (c) day 10 and (d) day 25.

[18] The difference fields of mean temperature and wind between the two days are given in Figure 3. It is found from Figure 3 that there is indeed a lower thermospheric warming associated with the stratospheric warming and mesospheric cooling, with a temperature increase of 20–30 K between 120–130 km. The maximum westward acceleration of the jet is at about 55 km between the stratospheric warming and mesospheric cooling regions. Similarly, the maximum eastward acceleration is at ∼110 km between the mesospheric cooling and lower thermospheric warming regions.

Figure 3.

Difference fields of zonal mean temperature (K) (color contours) and zonal mean zonal wind (ms−1) (line contours) between day 25 and day 10 in the stratosphere, mesosphere, and thermosphere.

[19] From these model results, it can be seen that the model simulates a complete stratospheric sudden warming (SSW) event, though it would not be counted as a major warming because the wind reversal in the stratosphere does not reach the 10 hPa level (the reversal stops 10 km above). However, the warming is still a strong event with more than 40 K increase in the stratosphere and 90 ms−1 wind change within two weeks. Figure 3 shows that the stratosphere, mesosphere, and lower thermosphere are closely coupled in the event. The dynamical processes associated with this event, the coupling of the different regions, and the implications for middle and upper atmospheric observations will be discussed in the rest of this section.

3.2. Growth of Planetary Waves

[20] Figure 4 shows the amplitudes of zonal harmonic waves, with zonal wave number 1–3, of the geopotential height perturbation at 62.5° from day 1 to day 36. It is immediately evident that wave 1 is the dominant feature in the stratosphere and mesosphere in this simulation. Wave 1 amplitude increases with altitude below 60 km, and it begins to grow after about day 5: There is a rapid wave growth from day 5 to day 8 followed by a decrease of the amplitude between day 8 and 13, and then the growth resumes and continues till day 18–21 (4–7 days before peak of the warming) with growth at lower altitudes lasting slightly longer than at high altitudes, and both decreasing rapidly afterwards. The wave 2 and 3 amplitudes also increases rapidly from day 15 to day 22 between 30 km and 70 km, with the maximum values lagging the wave 1 maximum by about 4 days at 55 km.

Figure 4.

Amplitudes of geopotential height perturbation (in meters) at 62.5°N from day 1 to day 36 between 10 and 120 km for (a) wave 1, (b) wave 2, and (c) wave 3. Contour interval is 100 m for Figures 4a and 4b and 20 m for Figure 4c.

[21] Time and latitude plot of wave 1 amplitude and phase at 250 hPa are shown in Figure 5a. The phase speed of the wave is approximately 0 from day 10 to day 18, 19, or 20 (depending on the latitudes). By comparing the phase (contour lines) and the amplitude (contour shade), it is worth noting that the wave 1 amplitude growth and decay between 40° and 70° track the wave phase structure closely. This correlation between the wave amplitude and phase is also found at higher altitudes as shown in Figures 5b and 5c. From Figure 5b, it is seen that wave 1 is baroclinic and has a westward phase tilt before day 21. From day 10 to 18, the phase speed at 250 hPa is near 0 as seen in Figure 5a, and the phase speeds at 100 hPa, 30 hPa, and 10 hPa are eastward and approximately 3.5° d−1. Starting from day 18 (19 or 20 for other latitudes), wave 1 at 250 hPa moves rapidly westward (about 140° westward within 6 days). The phase speed at upper levels slows down and reverses accordingly, though at a slower rate and with certain lag time. As a result of these phase changes, the westward tilting phase line becomes more vertical, and by day 25 at the peak of the warming, the phases at different levels coalesce, consistent with satellite observations of SSW [Quiroz, 1975]. From Figure 5, it is seen that the wave amplitude evolution has a good correlation with the phase change: the wave growth corresponds roughly to the period of quasi-stationary phases and the decay occurs as the phase changes around peak warming. The evolution of the vertical phase structure of the model at 62.5° throughout the stratosphere and mesosphere is shown in Figure 6a. From this figure it is clear that in reaching the phase coalescence between 10 and 70 km, the phases above ∼30 km move rapidly eastward while those below move rapidly westward.

Figure 5.

Evolution of the wave 1 phase and amplitude. (a) Wave 1 geopotential height perturbations at 250 hPa: phase (line contour in longitude degrees) and amplitude (shaded contour in meters) change between day 1 and day 36. The magnitude range of the shaded contour is from 120 m (dark) to 200 m (light) with 10 m interval. (b) Phase and (c) amplitude of wave 1 geopotential height perturbations at 62.5°N and 250 hPa (solid line), 100 hPa (dotted line), 30 hPa (dashed line), and 10 hPa (dash-dotted line) between day 1 and day 36.

Figure 6.

(a) Wave 1 phase of the geopotential height perturbation at 62.5°N from day 1 to day 36 between 10 and 120 km. The phase values may differ from those in other plots by ±360° to eliminate phase discontinuity in the plotting area. (b) Phase difference between approximate turning point and 250 hPa. 83 km and 70 km are taken as the approximate turning points before and after day 21, respectively, in the calculation. The dashed line is the theoretical value for resonance amplification.

[22] We will attempt to relate the wave growth described above to resonance amplification by examining the mean atmosphere state and the wave amplitude and phase structures. The zonal mean wind structure from Figure 2c is similar to those prescribed by Tung and Lindzen [1979b], which may favor a wave 1 or 2 mode with zero phase velocity. From the zonal mean wind and temperature, we can derive the square of the quasi-geostrophic refractive index [Matsuno, 1971] for wave 1 as shown in Figure 7. Day 15 and 21 illustrate the time when wave 1 is growing and beginning to decay, respectively, according to Figure 4. It is seen that in the high-latitude region, wave 1 is trapped at about 85 km on day 15 and at lower altitudes (55–75 km) on day 21, consistent with the wave 1 amplitude plot in Figure 4a. This change in refractive index is mainly due to the decrease and/or sign reversal of the meridional gradient of the zonal mean quasi-geostrophic potential vorticity, which is determined by the mean wind structure.

Figure 7.

Square of the refractive index for wave 1 on (a) day 15 and (b) day 21.

[23] By comparing the wave 1 amplitude (Figure 4a), phase (Figure 6), and zonal mean zonal wind (Figure 1c), it is evident that the wave 1 decay, its phase coalescence, and zonal mean wind deceleration and reversal are closely related. According to Tung and Lindzen [1979b], the wave will be in resonance if the height range between the ground and the altitude where the wave is trapped (the “turning point” in Tung and Lindzen [1979b]) is approximately equal to (1/8 + n/2)λz where n is a nonnegative integer and λz is the vertical wavelength of the wave. In other words, the phase difference between the turning point and the lower boundary should be −45°, −225°, etc for a westward tilting wave. The approximate phase difference between the turning point and 250 hPa from the model is shown in Figure 6b, and it is seen that the phase difference is near −225° between day 10 and day 18. Therefore, the wave 1 in the numerical model is likely to be in resonance during this period of time. On day 21, however, the phase difference is reduced to about −140° due to the rapid westward phase shift at 250 hPa, starting from day 18, and the descent of the turning point due to wind deceleration. The wind deceleration may also cause planetary wave breaking, which will be discussed later. The rapid phase shift at 250 hPa is probably caused by the resonance condition change due to wind changes in the stratosphere or by the damping in the troposphere. Though the temporal scale of wave growth is roughly the same as the damping time (∼5 days) due to Ekman pumping at 250 hPa as predicted by Tung and Lindzen, it is much longer at upper levels (∼15 days). The amplitude grows with altitude approximately as ez/2H below 60 km, with H being the scale height, consistent with Tung and Lindzen [1979b].

[24] From Figure 6, it is worth noting that before day 10, when the wave phase becomes quasi-stationary and the resonance condition is approximately satisfied, the phase changes quite rapidly. Starting from the lower part of the atmosphere, the phase moves first eastward and then briefly westward before it settles into a quasi-stationary phase. The phases at higher altitudes follow this pattern but with time lag and larger phase speeds. As a result of this time lag and the difference in the phase speeds, the phase difference between the turning point and 250 hPa reduces from −270° on day 3, overshoots the resonance phase of −225° to −190° on day 7, and then adjusts back to ∼−225°. This process with the transient phase adjustment may explain the brief wave 1 growth between day 5 and 8 before its major growth starting on day 12.

[25] Given the nonlinear and complex nature of this numerical model, it is difficult to make detailed comparisons between the model results and the aforementioned theory to determine the exact cause of the wave amplification. However, the heuristic comparison above does show some characteristics of resonance amplification. Furthermore, if resonance amplification is the mechanism responsible for the wave growth, it is then essential for a GCM to extend from the ground to the mesosphere or higher to self-consistently resolve a SSW event, because the turning point of the wave is usually located in the mesosphere. It also seems that the mean zonal wind structure changes due to wave 1 resonance amplification ultimately breaks the wave resonance condition and lead to the decay of the wave.

3.3. Planetary Wave and Mean Flow Interactions

[26] Figure 8 shows the zonal mean zonal wind acceleration rate due to Eliassen-Palm flux divergence (unit: ms−1 d−1) at 62.5° and 77.5° from day 1 to 36 (only TIME-GCM domain is shown), and the latitude-height structure on days 10 and 15. Starting from day 4, the acceleration in most of the middle atmosphere equatorward of ∼70° is weakly westward. The eastward forcing in the stratosphere and lower mesosphere before day 12 and after day 25 can be associated with the decay of the planetary wave. From around day 12, the westward acceleration in most of the middle atmosphere increases rapidly, reaching its peak around day 21 between 60 and 70 km. This correlates well with the zonal mean wind reversal and mean temperature gradient change shown in Figures 1c and 1b, respectively, and the start of the event at higher altitudes and its downward extension. Corresponding to the wave 1 growth between day 5 and 8, there is a quite strong westward forcing which causes the slight decrease of the jet around day 8 (Figure 1c). Therefore, the slight temperature decrease on day 8 is related to the rapid but brief wave 1 growth in the phase adjustment process.

Figure 8.

EP flux divergence (ms−1 d−1) between day 1 and day 36 at (a) 62.5°N and (b) 77.5°N in the computational domain of TIME-GCM (lower boundary at around 40 km). Latitude/altitude plot of EP flux divergence on (c) day 10 and (d) day 15, including computational domains from both TIME-GCM and CCM. Solid contour lines are for eastward forcing.

[27] The westward forcing also induces changes of the mean meridional residual circulation in the middle atmosphere. From Figure 9, it is clear that on day 21 the poleward/downward circulation below 60 km is much stronger than on day 10, and the previously strong poleward/downward circulation in the polar mesosphere between 60 and 80 km becomes poleward/upward. As shown by previous studies [cf. Matsuno 1971], the circulation change leads to adiabatic heating in the stratosphere and cooling in the mesosphere in the polar region. From day 12 to day 21, the enhancement of the mean meridional circulation and adiabatic heating below 60 km is accompanied by strong wave growth. However, this heating process is not uniform. As shown in Figure 10, the heating is asymmetric and it occurs mostly in the warm phase of the wave 1 planetary wave while the heating rate is about zero in the cold phase. As a result, large temperature gradients in both of the vertical and the longitudinal direction are produced between the cold and the warm phases during the warming.

Figure 9.

Mean residual circulation in the stratosphere and mesosphere on (a) day 10 and (b) day 21. A viewing factor has been multiplied by the vertical velocity to compensate for the biased aspect ratio. Vector with unit length designates 5 ms−1.

Figure 10.

(a) Temperature perturbation (K) and (b) total heating rate (Kd−1) due to horizontal and vertical advection at 62.5°N on day 17.

[28] The westward reversal of the mean zonal wind produces zero wind lines at high latitudes (at around 55°N between 40 and 80 km on day 25 from Figure 2). Therefore, the planetary waves encounter critical layers near the zero wind lines and break down [cf. Andrews et al., 1987]. Following McIntyre and Palmer [1983], the Ertel potential vorticity (PV) is examined. Figure 11 shows the three-dimensional normalized isentropic PV structures before the warming (day 9) and at the peak of the warming (day 25). The vertical range is from 500 K isentrope level to 3200 K isentrope level, corresponding approximately to 20–70 km, and the PV values are normalized by the maximum PV values on the corresponding levels. On day 9, the PV “core” column looks quite uniform and upright, indicating that PV peaks at the north pole at all levels. On day 25, almost the whole core column is displaced off the pole due to wave growth. As highlighted by the vertical lines at 55°N and above ∼40 km, large PV is stretched over the longitudes and around the critical layer, and PV with smaller values from lower latitudes is brought into high-latitude regions. It is also seen from Figures 11c and 11d that the PV stretching is more pronounced at higher altitude, due to earlier formation of the critical layer at higher altitudes and wind shear increase with altitudes. This is demonstrated more quantitatively from the time series of PV polar stereographic plots on isentropes 850 K (near 10 hPa) and 1500 K (near 1 hPa) in Figure 12. According to the wind plot in Figure 2, the wave may not encounter a critical layer yet at 850 K level but it does at 1500 K level. On day 5, the PV on isentrope 850 K is quite asymmetric in longitude: the latitude distance between the 3 and 4 × 10−4 Km−1 s−1 contours is about 5° at 0° longitude and is 30–35° at 180° longitude. There is a “surf zone” between 105–225°. As the planetary wave grows, the PV core is displaced equatorward and the 4 × 10−4 Km−1 s−1 contour is expanded and stretched, preferentially into the surf zone. In the process of wave decay, the PV core swings back to the pole and its value decreases. A new broader surf zone is left between the expanded 4 × 10−4 Km−1 s−1 contour line and the core (40–60°) after the warming. On isentrope 1500 K, the stretching of the PV is more pronounced than on 850 K (day 25) due to stronger wind shear. By comparing day 5 and 30, the 4 × 10−3 Km−1 s−1 contour line only expanded slightly, but the core PV value has dropped almost by half and the area enclosed by 6 × 10−3 Km−1 s−1 has shrunken significantly, mostly due to wave decay. As a result, the PV gradients between 4 and 6×10−3 Km−1 s−1 is greatly reduced. Therefore, the PV structure in the stratosphere and stratopause bears a signature of a breaking wave consistent with the planetary wave breaking process described by McIntyre and Palmer [1983].

Figure 11.

Three-dimensional (3-D) Ertel potential vorticity structures on isentropes from 500 K to 3200 K on day 9, (a) top view and (b) bottom view, and on day 25, (c) top view and (d) bottom view. The potential vorticity is normalized by the maximum values at each level and the plotting range is 0.45–1. The latitudinal range of the plot is 30–90°N, with the north pole designated by the vertical line in the center (visible at the bottom panels when the PV core is displaced off the pole). Vertical lines are also placed every 30° along 55°N latitude extending from about 40 km upward, highlighting the approximate location of the critical layer.

Figure 12.

Polar stereographic plots of Ertel potential vorticity. (a, b, e, and f) Days 5, 15, 25, and 30 on the 850 K isentrope (unit: 10−4 Km−1 s−1), respectively. (c, d, g, and h) Same days on the 1500 K isentrope (unit: 10−3 Km−1 s−1). Latitudinal range of the plots is 20–90°N.

Figure 12.


[29] Planetary waves with higher wave numbers (2 and 3) increase from day 13 to 22, coincident with the wave 1 growth, as shown in Figure 4. Analysis of the geopotential height longitude profile shows that in this simulation this is mainly due to the displacement and distortion of the vortex off the pole. Wave 2 refractive index is imaginary poleward of 50° at most altitudes below 10 km or above 30 km before day 15. Therefore, wave 2 is unable to propagate into the middle atmosphere. However, the boundary of the region with imaginary refractive index is shifted poleward (by almost 10° in this case) as the eastward wind decelerates. So it can be expected that the wind deceleration during wave growth and before the formation of critical layers around peak warming may favor wave 2 growth. This may explain the fact that in some major warming events, there is a wave 2 maximum between the time when wave 1 maximizes and the time when the warming reaches its peak [Labitzke, 1977].

3.4. Gravity Wave Filtering and Changes of Circulation and Transport in the Mesosphere and Lower Thermosphere (MLT)

[30] The deceleration and reversal of the eastward jet in the high-latitude stratosphere also changes the filtering of internal gravity waves and allows eastward propagating gravity waves from lower atmosphere to penetrate into the MLT while blocking gravity wave components with zero and westward phase speed, as shown in Figures 13a and 13b. Therefore, the previously dominant westward forcing in the high-latitude MLT region is replaced by increasing eastward forcing prior to the maximum warming, and the corresponding westward zonal wind becomes eastward (Figure 2d). This process is reversed after the maximum warming.

Figure 13.

Zonal mean zonal gravity wave forcing (ms−1 d−1) on (a) day 5 and (b) day 21. Solid lines are for eastward forcing. Residual circulation with the vector lengths normalized to unit values on (c) day 5 and (d) day 21. A viewing factor has been multiplied by the vertical velocity to compensate for the biased aspect ratio.

[31] The analysis in the previous section shows that the residual circulation changes in the stratosphere and lower mesosphere due to changes in planetary wave forcing. In the MLT region, on the other hand, the residual circulation changes as a result of the gravity wave forcing change, as shown in Figures 13c and 13d where all vector lengths are normalized to unit value so that the circulation pattern within the whole altitude range from 30 to 120 km can be visualized. It is seen that the eastward gravity wave forcing in the winter hemisphere induces an equatorward and upward flow between 90–105 km and downward above. This circulation change enhances the adiabatic cooling in the polar mesosphere and adiabatic heating in the polar thermosphere, as shown in Figures 2b and 3. It can also be seen from Figure 1a that the temperature change in the mesosphere starts from day 13 (12 days before the peak warming), almost as soon as the stratospheric jet begins to decrease (Figure 1c).

[32] The upward-equatorward circulation in the MLT region also changes the transport and distribution of chemical species. Atomic oxygen has a relative long chemical life time during polar night and its mass mixing ratio increases with altitudes. An upward-equatorward circulation in the MLT region causes a net divergence of atomic oxygen around its number density peak at ∼95 km, as shown in Figure 14a The peak atomic oxygen number density at 62.5°N decreases from its maximum value of 9.6 × 1011 to about 5.5 × 1011 on day 25, and the corresponding green line emission, calculated according to McDade et al. [1986], also decreases sharply during the warming (Figure 14b). The decrease of the atomic oxygen and green line emission starts around day 9, but the rate of decrease only becomes large starting from day 15. After day 25, the atomic oxygen and green line emission recovers rapidly. This rapid recovery coincides with the large westward gravity wave forcing around 95 km after the warming, which restores the poleward and downward circulation. The relative changes of the midnight atomic oxygen column density and height integrated green line emission rate between day 25 and day 10 are plotted in Figure 14c as a function of latitude. The decrease of these quantities extends well into midlatitudes. For example, the height integrated green line emission rate decreased by about 10% at 30°N and 35% at 40°N. This is consistent with the residual circulation change as shown in Figures 13c and 13d.

Figure 14.

(a) Zonal mean atomic oxygen number density (cm−3) and (b) green line (557.7 nm) emission rate (photons cm−3s−1) at 62.5°N between day 1 and 36. (c) Latitudinal distribution of the relative changes of atomic oxygen column density (dotted line) and height integrated green line emission rate (solid line) at local midnight between day 10 and day 25.

[33] It should be pointed out the altitude range of the gravity wave breaking is dependent on the gravity wave specification at the lower boundary which is not well quantified and is tuned so that the model results agree with various observations [e.g., McLandress et al., 1996]. Therefore, there is uncertainty, and very likely variability in the real atmosphere, in the wave breaking altitudes. The changes in the MLT residual circulation and atomic oxygen distribution during SSW will thus vary with the conditions at the lower atmosphere.

[34] The changes of the green line emissions during SSW at high latitudes have been reported previously by Ismail and Cogger [1982]. A detailed examination of the green line emission intensity measured at Thule and the temperature at 30 hPa during a major warming event in 1973 (Figure 5 in their paper) reveals that the temperature increase in the stratosphere started from January 11 and reached the peak (an increase of about 40 K) around February 1. It is different from this numerical simulation in that the simulated stratospheric temperature stayed high for about 10 days before it began to cool down again. The green line intensity, on the other hand, started to decrease rapidly on January 22 and reached minimum on February 1, with its value about 20% of its peak value. It then had a fast recovery within a week. These are in qualitative agreement with our simulation results. The evidence of green line emission change during a SSW is also found from cross examining the green line measurements at Meteorological Institute of Stockholm University (MISU) (60°N, 20°E) [Shepherd et al., 1999] and the north pole temperature at 10 hPa [Shepherd et al., 2002]. In 1992, there is a warming peaking near day 84 according to the north pole temperature at 10 hPa, and the green line emission rate at MISU is at local minimum on days 84 and 85 and quickly recovered afterwards. In 1993, the peak warmings near day 43, day 64, and day 78 also correspond well with green line emission minima. Therefore, the “springtime transition” in the mesospheric airglow and temperature, as demonstrated by Shepherd et al. [1999] and Shepherd et al. [2002], might be closely related to late or final stratospheric warmings.

[35] A detailed examination of Figure 6a reveals that there is a transition region between 80 and 100 km where wave 1 undergoes a relatively rapid phase change (≥100° within 10 km), and this region is also where the wave amplitudes are locally minimal (Figure 4a). For example, the wave amplitude on day 21 minimizes near 90 km and then increases to more than 1000 m around 110 km, and the phase difference between 85 and 95 km is about 100°. Furthermore, the wave is about 190–200° out of phase between the two local maxima at 60 and 110 km. These features are typical of the in situ forced planetary waves due to filtered gravity wave forcing, as shown by Smith [1996, 1997] and Meyer [1999a, 1999b]. This is also evident for wave 2 and 3 in Figures 4b and 4c, especially around maximum warming.

[36] The growth of the in situ forced planetary waves in the MLT has important implications for the wave interaction and wind variation in that region. From spectral analysis of the model results, it is found that semidiurnal wave 1 westward propagation increases at high latitudes during the SSW event. This is probably due to the nonlinear interaction between migrating semidiurnal tide and the quasi-stationary planetary wave 1. The interaction of these components causes zonal mean wind variations in the region, as shown in Figure 15a. Above 80 km, the zonal mean wind variation has an obvious period of a half day; the maximum variation at 110 km is 10 to 15 ms−1. These changes are the combined results from the forcing of planetary waves, tidal waves, and gravity waves as shown in Figures 15b and 15c. Between 80 and 100 km, the EP flux divergence has both diurnal and semidiurnal periods, while above 100 km, the period is dominantly semidiurnal. The gravity wave forcing is a response to the local mean wind change.

Figure 15.

(a) Zonal mean zonal wind at 62.5°N between UT00 of day 23 and UT00 of day 25. (b) Similar to Figure 15a but for EP flux divergence (ms−1 d−1). (c) Similar to Figure 15a but for gravity wave forcing (ms−1 d−1).

[37] To understand the EP flux divergence and its variation, we partition the forcing by different components. Because the EP flux of components with the same wave number are not linearly additive under zonally averaging, different combinations of these components are examined. The results associated with wave 1 and 2 components are plotted in Figure 16, where the EP flux divergence profiles are taken at 62.5°N. The three columns, from left to right, are profiles at UT00, UT06, and UT12, respectively. The first row shows the profiles due to wave 1 components, total (solid lines) and the 3 single components (migrating diurnal tide, quasi-stationary planetary wave 1 and semidiurnal wave 1). The second row is partition of wave 2 into migrating semidiurnal tide, quasi-stationary planetary wave 2, and their combination. From Figure 16, it is seen that EP flux divergence due to wave 1 components is dominant between 80 and 100 km, while EP flux divergence due to wave 2 components is dominant above. For wave 1, single components EP flux divergence from migrating diurnal tide and semidiurnal wave 1 are insignificant (Figures 16a–16c). Quasi-stationary wave 1 is significant in the stratosphere, but in the mesosphere its forcing is relatively small and does not produce the variations seen in the total forcing. For the superposition of two components (not shown), migrating tide/semidiurnal wave 1 EP flux divergence are small at all three UT. Both migrating tide/quasi-stationary wave 1 and quasi-stationary wave 1/semidiurnal wave 1 have relatively large forcing between 80–100 km. The former has a variation period of one day while the latter of a period of half day. This is due to the relative phase change of these components. Similarly, the single components EP flux divergence due to migrating semidiurnal tide and quasi-stationary wave 2 are small but EP flux divergence of the superposed components are much larger above 100 km and with a half day period (Figures 16d–16f) and variation amplitude of about 35–40 ms−1 d−1. Therefore, the EP flux divergence of superposed wave components with the same wave number but different period drives the zonal mean wind variations in MLT as seen in Figure 15. From the simulation, such variations become stronger during SSW due to growth of the in situ forced planetary wave in MLT and their nonlinear interactions with the tidal waves.

Figure 16.

Partition of EP flux divergence (ms−1 d−1) at 62.5°N on day 23. (a)–(c) EP flux divergence profiles at UT00, UT06, and UT12, respectively, including migrating diurnal tide (DT) (dotted lines), stationary planetary wave 1 (P10) (dashed lines), and semidiurnal wave 1 (P12) (dash-dotted lines). The solid lines are the profiles of the EP flux divergence due to the superposed perturbation from the 3 components (DT + P10 + P12). (d)–(f) Similar to Figures 16a–16c but for partition of wave 2 into single components, including migrating semidiurnal tide (SDT) (dotted lines) and stationary planetary wave 2 (P20) (dashed lines). Solid lines: SDT + P20.

4. Conclusions

[38] A coupled NCAR TIME-GCM/CCM3 simulation was run for 3 consecutive years and in January of year 4 a self consistently generated stratospheric sudden warming event was detected. By analyzing the simulation results from the coupled model system, we can gain a comprehensive understanding of the process of planetary wave growth and decay, planetary wave interaction with the mean flow, and changes of the mean circulation and transport in the stratosphere, mesosphere, and thermosphere during the stratospheric warming event.

[39] In this simulation, wave 1 is the dominant wave component at most altitudes and its growth precedes the simulated SSW and persists for about 2 weeks. The period of the wave 1 amplitude growth coincides with the stationary phase of the wave in the lower part of the atmosphere (∼250 hPa) and a weakly eastward phase propagation aloft. Furthermore, the vertical phase difference of the westward tilting wave between 10km and the turning point of the wave (at about 85 km from the analysis of the refractive index) is close to 225°, a wave resonant condition proposed by Tung and Lindzen [1979b]. Therefore, the wave 1 growth in the model is probably due to wave resonance. The model also shows that this vertical phase difference goes through a quite rapid adjustment process before settling into the resonance phase. The stratospheric jet produced by the model prior to the warming agrees with climatology, so a high-latitude jet core is not essential for the wave 1 warming in this simulation. The perturbation growth is asymmetric in longitude and the model demonstrates that the total heating due to both horizontal and vertical advection coincides with the warm phase of wave 1 and there is no heating or cooling in the cold phase. This results in large temperature gradients in both vertical and longitudinal directions between the warm and cold phases of the wave. The decay of the wave, on the other hand, is accompanied by a rapid decrease of the vertical phase difference and a final phase coalescence near the time of the peak stratospheric warming, consistent with previous observations. The decay is probably due to the a decrease in the forcing from the lower boundary and/or the zonal wind change in the mesosphere and upper stratosphere.

[40] The amplification of the quasi-stationary wave 1 produces strong westward forcing in the winter stratosphere, especially at high latitudes, causing the reversal of the stratospheric jet. The reversed jet forms a critical layer for upward propagation of the quasi-stationary planetary wave 1 and leads to the breakdown of the wave. The potential vorticity, with its core displaced off the pole due to wave growth, is stretched over longitudes; air with large PV is entrained to lower latitudes. This is more pronounced at stratopause and lower mesosphere in the model than in the stratosphere, probably due to larger meridional shear of the zonal wind. As the wave breaks down and decays, the PV core also decays and recedes to the polar region, leaving a surf zone at the midlatitudes/high latitudes. At higher altitudes (stratopause and above), the surf zone is not as clearly defined as that at lower altitudes, because the PV core is less tight after the wave breaks down and the PV meridional gradient is more uniform.

[41] The westward forcing due to planetary waves in the winter stratosphere also induces a poleward/downward residual circulation and adiabatic warming in the stratosphere, as shown by previous studies. The weakening and reversal of the eastward stratospheric jet allow increasing amounts of eastward propagating gravity waves to penetrate into the mesosphere and lower thermosphere and break there, while decreasing the amount of westward propagating gravity waves reaching these altitudes. Therefore, the net eastward forcing due to gravity wave breaking is increased in the MLT and the jet reverses from westward to eastward. The mean residual circulation also changes from downward to upward, leading to strong adiabatic cooling in the mesosphere (resulting in a temperature decrease of about 50 K at high latitudes). It is also noted that the cooling in the mesosphere occurs about two weeks prior to the peak stratospheric warming in response to the weakening of the stratospheric jet. A lower thermospheric warming is also detected in the model due to a secondary downward circulation induced by the equatorward mesospheric circulation.

[42] The equatorward/upward circulation in the MLT also causes a decrease in the atomic oxygen peak number density and thus in the green line airglow emission rate. In the model, the decrease of the green line emission reaches as much as 75% in the polar region and 30–50% in the midlatitudes around the peak of the stratospheric warming. A fast recovery of atomic oxygen and green line emission follow the event, along with the recovery of the normal circulation pattern.

[43] Planetary waves forced in situ by filtered gravity waves are also discovered in the MLT region. Their growth and interaction with tides have significant contributions to the variability in the MLT region in this stratospheric warming event. The nonlinear interaction between the quasi-stationary wave 1 and the migrating semidiurnal tide generates a semidiurnal wave 1 (westward) component. The forcing by the combined wave 1 components (quasi-stationary, migrating diurnal tide, and semidiurnal wave 1) has both diurnal and semidiurnal periods and its magnitude is much larger than simply summing the forcing of each individual component, because EP flux is not linearly additive for components with the same wave number. Similarly, the forcing by the combined wave 2 components (quasi-stationary and migrating semidiurnal tide) has a variation period of a half day and again a magnitude much larger than each individual component or the summation of the two. As a result, the zonal mean flow in the MLT shows large diurnal and semidiurnal variations.


[44] The authors wish to thank A. K. Smith for helpful discussion and comments. Comments by V. Yudin and an anonymous reviewer are acknowledged. We would also like to acknowledge E. C. Ridley and B. Foster for their contribution to the development of the coupled TIME-GCM/CCM3. The authors' efforts are in part supported by the NASA Sun Earth Connection Theory Program (S-13753-G). The National Center for Atmospheric Research is sponsored by National Science Foundation.