We examine for the first time the propagation of gravity waves (GWs) of lower atmospheric origin to the thermosphere above the turbopause during a sudden stratospheric warming (SSW). The study is performed with the Coupled Middle Atmosphere-Thermosphere general circulation model and the implemented spectral GW parameterization of Yiğit et al. (2008). Simulations reveal a strong modulation by SSWs of GW activity, momentum deposition rates, and the circulation feedbacks at heights up to the upper thermosphere (∼270 km). Wave-induced root mean square wind fluctuations increase by a factor of three during the warming above the turbopause. This occurs mainly due to a reduction of filtering eastward traveling harmonics by the weaker stratospheric jet. Compared to nominal conditions, these GW harmonics propagate to higher altitudes and have a larger impact on the mean flow in the thermosphere, when they are dissipated. The evolution of stratospheric and mesospheric winds during an SSW life-cycle creates a robust and distinctive response in GW activity and mean fields in the thermosphere above the turbopause up to 300 km.
 Sudden stratospheric warmings (SSWs) are spectacular events in the winter Northern Hemisphere (NH) first discovered observationally by Scherhag . They are accompanied by deceleration and even reversals of the zonal mean wind. Matsuno  was the first who demonstrated with a simple dynamical numerical model that planetary waves and their interactions with the zonal mean flow are responsible for SSWs. Further numerical studies confirmed the Matsuno  conclusion qualitatively [Holton, 1976; Palmer, 1981].
 An increasing number of observations indicate SSW signatures not only in the stratosphere but also in the thermosphere-ionosphere. Using the Millstone Hill Incoherent Scatter Radar ion temperature data,Goncharenko and Zhang  discovered alternating regions of warming in the lower thermosphere and cooling above ∼150 km, which were related to the minor SSWs. Goncharenko et al.  demonstrated that SSWs induced local time variations in the equatorial ionization anomaly. Pancheva and Mukhtarov  found in COSMIC (Constellation Observing System for Meteorology Ionosphere and Climate) data a systematic negative response of the ionospheric mean F2 layer peak plasma frequency (f0F2), the F2 layer peak ionization height (hmF2), and electron density to SSWs. Recently, a combined analysis of the observed total electron content (TEC) and the European Centre for Medium-Range Weather Forecasts reanalysis data showed that there is a coincidence between TEC enhancements and increases in stratospheric ozone due to planetary wave-mean flow interactions [Goncharenko et al., 2012].
 Since SSWs alter large-scale dynamical and thermal fields in the middle atmosphere, propagation and dissipation of small-scale GWs traveling upward from the lower atmosphere are expected to respond to these changes. Observations indicate considerable spatial and temporal variations of GW activity in the NH high-latitudes between 20 and 60 km during SSWs [Wang and Alexander, 2009; Thurairajah et al., 2010]. Global circulation models (GCMs) were used to demonstrate the link between SSWs and variations of GW activity at altitudes up to the lower thermosphere [Liu and Roble, 2002; Yamashita et al., 2010]. However, such studies were not performed in the thermosphere above the turbopause. Primarily, this was due to a lack of proper representation in GCMs of GWs at thermospheric heights.
 This paper reports on the first study of penetration of small-scale internal GWs of lower atmospheric origin into the thermosphere above the turbopause during an SSW in a three-dimensional GCM. It was conducted with the Coupled Middle Atmosphere Thermosphere Model-2 (CMAT2) interactively coupled with the subgrid-scale spectral nonlinear GW parameterization ofYiğit et al. .
 The main goal of this study is to quantify and analyze SSW-induced variations of GW dynamical effects (“GW drag”), and GW activity (“RMS wind fluctuations”) from the lower atmosphere to the upper thermosphere. The GCM, GW parameterization, and design of the model experiments are outlined insection 2. Results are presented in sections 3–5, and conclusions are given in section 6.
2. Numerical Tools and Model Experiments
 The CMAT2 GCM employed here is described in detail by Yiğit  and Yiğit et al. [2009, 2012]. In the vertical direction, the model uses a pressure level grid extending from the tropopause at 102 hPa (∼15 km) to the upper thermosphere at 1.432 × 10−6 hPa (250–600 km). The horizontal domain was represented by a 2° × 18° latitude–longitude grid.
 Because the model does not have a troposphere, National Centers for Environmental Prediction reanalysis data are used at the lower boundary to drive the GCM. Migrating solar tides with wavenumbers from one to three at the lower level are adapted from Hagan and Forbes [2002, 2003].
 The GW source spectrum at the lower boundary (∼15 km) is represented by 30 harmonics, whose momentum fluxes were normally distributed with respect to their horizontal phase speeds. Only harmonics with phase speeds up to 80 m s−1 (but traveling in both directions along the local wind at the source height) are considered. An illustration of the spectrum can be seen in Yiğit et al.  (Figure 1).
 The CMAT2 model with the extended GW scheme has been extensively validated against empirical models of the thermosphere, and agrees well with the observed temperature and wind distributions [Yiğit and Medvedev, 2009; Yiğit et al., 2009, 2012].
 The model was run for quiet solar conditions (F10.7 = 80 × 10−22 W m−2 Hz−1) for five months from September to January, outputting the data every three hours during a period of 6 weeks between 2 December and 16 January during a minor SSW.
3. Sudden Stratospheric Warming
 General characteristics of the simulated minor warming are presented next. Black lines in Figures 1a and 1b show the universal time (UT) variations of the zonal mean neutral temperature over the North Pole, and zonal mean zonal wind ū at 10 hPa pressure level (∼30 km) at 60°N, respectively, from 2 Dec to 16 Jan. The red line represents the zonal mean temperature difference between 60°N and the North Pole, . The results are indicative of a minor SSW, as defined in the seminal review of Schoeberl : “In a minor warming, …the meridional temperature gradients usually weaken but do not reverse.” In our simulations, increases rapidly by 18 K from 216 K on Dec 16 to 234 K on 4 Jan, and steadily decreases again after this date. decreases throughout the period when at the North Pole rapidly rises. This reduction of the equatorward temperature gradient occurs due to the poleward transport of heat by enhanced planetary wave activity. Figure 1b shows that before the minor warming commences, zonal mean zonal wind weakens, and decreases continuously from 52 to 24 m s−1 throughout the minor warming.
 To demonstrate the spatial evolution of the fields during the warming, the NH polar stereographic projections of geopotential height (color shaded) and zonal wind (contour) at 10 hPa are shown on three dates (15 Dec, 24 Dec, and 1 Jan) in row two of Figure 1. As the warming progresses, the polar night vortex splits into two and the geopotential height increases at high-latitudes in accordance with the rising temperature.
4. Variations of Gravity Wave Activity and Effects in the Thermosphere
 Next, we investigate how the minor SSW shown in Figure 1 impacts the GW propagation/dissipation, and what effects these waves produce above the turbopause (∼105 km). We evaluate the root mean square (RMS) wind fluctuations due to GWs, , M being the number of harmonics in the spectrum, σ2 is the horizontal wind variance, and is the wave amplitude in the zonal wind, as a quantitative measure of wave penetration into the upper atmosphere. GW dynamical effects are characterized by the zonal GW drag, , where ρ is the mean density.
Figure 2a presents temporal variations of the zonally averaged RMS at 60°N from 14 Dec to 7 Jan as a function of altitude. The white dashed line marks the onset of the warming when the westerly zonal wind started to weaken (cf. Figure 1b). GW-induced wind variations above 90 km steadily rise after that date, and reach their maxima around 25 Dec. The white dotted line on Jan 4 indicates the peak of the SSW. Before the warming, the maximum of RMS (∼3 m s−1) is in the mesosphere at around 80 km. During the warming, it increases to ∼6 m s−1, and is shifted higher into the lower thermosphere to around 120 km. Also, the overall GW penetration into the thermosphere increases during the event. Thus, the RMS rises to > 4 m s−1 at altitudes up to ∼250 km compared to 1–1.5 m s−1 before the onset.
 UT variations of the mean zonal GW drag ax at 60°N are plotted in Figure 2b. Remarkable changes occur with it during the SSW. Penetration higher into the thermosphere and larger amplitudes of waves result in the enhancement of ax, especially in the lower thermosphere at around 120 km, and in the upper thermosphere at ∼250 km. Before the SSW, there is only easterly drag below 120 km with the maximum of 100 m s−1 day−1 at around 80 km, and a weak (up to 10 m s−1 day−1) westerly drag in the upper thermosphere at 60°N. Mean zonal wind shown in Figure 2c demonstrates the role of selective filtering in GW effects in the thermosphere. After the warming onset, stratospheric westerlies weaken, thus permitting more eastward harmonics to propagate upward. As a result, the deposited momentum is markedly westerly between 120 and 180 km. Higher in the thermosphere at around 250 km, the westerly drag dramatically enhances after 21 Dec to more than 150 m s−1 day−1. As the warming progresses, the easterly wind at ∼120 km gradually weakens and reverses. In response, the lower thermospheric westerly drag also decreases and gradually changes the direction. Easterly propagating GW harmonics are Doppler shifted to faster intrinsic phase speeds by the westerlies in the lower thermosphere, thus propagating favorably upward. Upon their dissipation in the thermosphere below 180 km, they deposit the negative momentum (blue shades) to the mean flow. This descending with time pattern in the middle atmosphere and thermosphere is seen in Figure 2b. In the upper thermosphere, strong temporal alternations between easterly and westerly GW momentum deposition in the course of SSW are created by faster harmonics in both directions surviving the weaker filtering below, and modulated by solar tides.
5. A Closer Look at Gravity Wave Penetration into the Thermosphere
 In order to investigate further how the GW penetration into the thermosphere and their effects vary with time during a minor SSW, three representative times are chosen: a) few days before the onset (15 December), b) the ascending phase (24 December), and c) towards the end of warming phase (1 January). In Figure 3, the RMS wind fluctuations at (60°N, 18°W) on these days are plotted in red color shades. The same color scheme is used to enable an intercomparison of this measure of GW activity. The local zonal wind and the momentum deposition by GWs are overplotted in black and color contour lines, respectively.
 Before the onset of SSW (Figure 3a), there is a general lack of westerly propagating harmonics (c > 0) in the thermosphere, because these waves are filtered below by the strong westerly stratospheric jet. Easterly waves propagate upward into the middle atmosphere, and break/saturate in the mesosphere, depositing the easterly momentum shown with dashed color lines. The upward propagation of easterly GWs (c < 0) depends on local time. Between 0 and 12 UT, the easterly local wind filters out these harmonics, as is seen from the rapid vertical decay of the RMS. After 1200 UT on 15 Dec, the fast easterly waves populate thermospheric altitudes, owing to favorable propagation conditions. Their intrinsic phase speeds, c − u, increase above 120 km when the local wind is westerly (u > 0). Therefore, these GW harmonics have longer vertical wavelengths, less affected by molecular viscosity, and propagate higher. This is seen in Figure 3a, where the RMS wind fluctuations grow up to 4.5 m s−1 at 2100 UT. Eventually, owing to dissipation due to molecular viscosity, waves induce a strong easterly drag of more than 100 m s−1 day−1 for the mean flow.
 During the mid-phase of the warming on 24 Dec, the stratospheric westerlies are much weaker, and, therefore, faster westerly waves propagate upward to the thermosphere. Most of the surviving waves withc > 0 are attenuated primarily by nonlinear processes of breaking/saturation after their amplitudes sufficiently grow. The associated westerly acceleration exceeds 100 m s−1 day−1 throughout the day around 120–150 km, as seen in Figure 3b. Propagation of fast westerly waves is favored further by the easterly wind in the thermosphere above 150 km during the first half of the day. The increased vertical wavelength makes them less susceptible to molecular viscosity, enabling them to propagate higher up into the upper thermosphere. The enhanced RMS wind fluctuations and positive GW drag seen in Figure 3b are created by these harmonics. After 1200 UT, the RMS below 150 km is due to GW harmonics with phase velocities of both signs. Higher, the westerly wind filters harmonics with c > 0, and the wave variance and easterly momentum deposition are created by the dissipation of easterly GWs, much like on 15 Dec.
 In the final phase of the warming (Figure 3c), the westerly momentum deposition is confined to the MLT region only. As before, the weaker background stratospheric jet permits more GW harmonics with c > 0 to propagate to the MLT, where they impart westerly acceleration upon their dissipation. At this stage, however, the westerlies have replaced the easterly wind between 90 and 120 km due to the westerly acceleration during the mid-phase of the warming. These winds in the MLT are overall responsible for the enhanced dissipation of westerly waves. Only few fast westerly waves withc ≥ 60 m s−1 are capable of penetrating above 120 km into the thermosphere, where they are in competition with easterly waves that have much more favorable propagation conditions after 1200 UT. The simulated GW activity and the easterly drag during that time is a result of this competition.
6. Summary and Conclusions
 Propagation and dynamical effects of gravity waves (GWs) of lower atmospheric origin in the thermosphere above the turbopause (∼105 km) during a minor sudden stratospheric warming (SSW) have been studied for the first time using a three-dimensional general circulation model (GCM). Simulations with an implemented spectral GW parameterization [Yiğit et al., 2008] accounting for thermospheric dissipation of GWs demonstrated a robust and distinctive response in the upper thermosphere to the SSW event. It manifests itself in enhanced GW activity and drag in the thermosphere, and in changes in the mean circulation. The main mechanism that controls this response is the selective filtering of GWs traveling upward from the lower atmosphere by changes in the stratospheric jet and mesospheric winds over the course of the warming.
 The major inferences of this investigation are summarized below.
 1. Weakening of the stratospheric polar night jet during the warming's mid-phase permits more fast westerly GW harmonics to penetrate into the lower and upper thermosphere. The associated RMS wind fluctuations increase by several times above 120 km, while the eventual dissipation of westerly waves due to nonlinear breaking/saturation and molecular viscosity and thermal conduction enhances the westerly drag.
 2. Towards the end of the warming, the imposed westerly GW drag slows down and even reverses the easterlies in the high-latitude mesosphere and lower thermosphere (MLT). This process has a descending with time pattern, which begins at around 150 km.
 3. In turn, the weaker easterlies in the MLT enable more easterly GWs to penetrate the upper thermosphere. They also increase the RMS wind fluctuations by several times and create an enhanced easterly drag when they are dissipated in the upper thermosphere.
 The above processes are modulated by strong solar tides in the thermosphere, and, therefore, have distinct diurnal variations. Up to 270–300 km, these variations have larger magnitudes during SSWs. Thus, an ionospheric response to SSW events is quite plausible and yet to be investigated. Tidal effects will be considered in detail in a future study. The simulated warming is relatively weak. In case of a stronger warming event, such as a major SSW, in which the zonal mean zonal wind reverses its direction, more dramatic variations of GW propagation and effects are to be expected in the thermosphere. In particular, the westerly harmonics are likely to be affected to a minor degree and could easily propagate to thermospheric altitudes, potentially producing larger eastward drag.
 ASM was partially supported by the German Science Foundation (DFG), project HA3261/5 and EY was supported by NASA grant NNX11AQ73G.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.