Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Corresponding author: H. Liu, Earth and Planetary Science Division, Kyushu University, Fukuoka, Japan. (email@example.com)
 The whole atmosphere model GAIA is employed to shed light on atmospheric response to the 2009 major stratosphere sudden warming (SSW) from the ground to exobase. Distinct features are revealed about SSW impacts on thermospheric temperature and density above 100 km altitude. (1) The effect is primarily quasi-semidiurnal in tropical regions, with warming in the noon and pre-midnight sectors and cooling in the dawn and dusk sectors. (2) This pattern exists at all altitudes above 100 km, with its phase being almost constant above 200 km, but propagates downward in the lower thermosphere between 100 and 200 km. (3) The northern polar region experiences warming in a narrow layer between 100 and 130 km, while the southern polar region experiences cooling throughout 100–400 km altitudes. (4) The global net thermal effect on the atmosphere above 100 km is a cooling of approximately −12 K. These characteristics provide us with an urgently needed global context to better connect and understand the increasing upper atmosphere observations during SSW events.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The stratosphere sudden warming (SSW) is a dramatic meteorological event in the winter polar stratosphere. Its formation mechanism and cooling impact on the mesosphere has been well demonstrated by Matsuno . Extension of SSW impacts to the lower thermosphere was predicted by Liu and Roble  using the Thermosphere-Ionosphere-Mesosphere Electrodynamic General Circulation Model (TIME-GCM), showing a warming effect near 120 km. SSW impacts on the upper atmosphere have been recently revealed in various observations. Using Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) satellite observations, Funke et al.  found warming effect in the polar lower thermosphere during the record-breaking major SSW event in January 2009, hence supporting the prediction of Liu and Roble  in the northern polar region. At middle latitude, Goncharenko and Zhang  reported ion cooling above 150 km, but warming around 120 km during the SSW2008 event. Since the ion temperature closely follows neutral temperature [Schunk and Nagy, 2000], their results indicate similar SSW impact on the thermosphere. However, this pattern is in opposite sense to the prediction of Liu and Roble [2002, see Figure 2 therein] at midlatitudes. At even higher altitude, Liu et al.  reported decrease of thermospheric density simultaneously observed by CHAMP and GRACE satellites at 325 and 475 km altitude in the dawn and dusk sectors. Therefore, one asks: What is the global context to consistently connect these observations at different altitude and latitude?
 Furthermore, the SSW impact on the ionosphere has been demonstrated to strongly depend on local time (LT), being roughly semidiurnal as observed in various ionospheric parameters like the total electron content, the Sq current, etc. [e.g., Goncharenko et al., 2010; Yamazaki et al., 2012]. Given the close coupling between the thermosphere and ionosphere via either ion drag or atmospheric waves, it is easy to postulate that SSW impacts on the thermosphere may well be local time dependent. So far, no thermosphere observation covering all local times has been reported during SSW events. The prediction of SSW effects on the lower thermosphere by Liu and Roble  was given as all local time averaged. This lack of local time information greatly limits our understanding of upper atmosphere observations during SSWs.
 Therefore, the present study seeks to obtain a global picture of the SSW impact on the upper atmosphere, with a focus on the local time and height dependence of thermosphere response in terms of temperature. For this purpose, we employ the fully coupled atmosphere-ionosphere model, GAIA.
2 GAIA Model
 The GAIA model is a self-consistent fully coupled model of the Earth's lower atmosphere, thermosphere, and ionosphere. It covers the altitude range from the ground to exobase (~500 km at solar minimum). Details about the model are described in Jin et al. . By interactively coupling the physical processes in the lower and upper atmosphere, GAIA has been proved to be highly capable in modeling prominent features in the thermosphere and ionosphere, e.g., the equatorial mass density anomaly, the equatorial wind jet, and the wave-4 structures [Stolle and Liu, 2013, and references therein]. The model's capability in characterizing gravity and tidal waves has also been demonstrated in various studies [e.g., Miyoshi and Fujiwara, 2003].
 The GAIA model is employed to simulate the upper atmosphere response to the major SSW in January 2009. A nudging technique is used to converge the model results below 30 km altitude to observations represented by the Japanese 25 year Reanalysis Project data [Jin et al., 2012]. During the simulation period of 1 November 2008 to 31 March 2009, the F10.7 index varied only about 1.5 % around 69.3. For investigating SSW effects, a fixed cross-polar cap potential of 30 kV and a quiet particle precipitation condition were held throughout the simulation period to exclude any geomagnetic activity effect. Simulation results on the stratosphere and ionosphere are presented in Jin et al. . Those results show good agreement to satellite observations in both the stratosphere and ionosphere, hence demonstrating the model's capability in capturing key processes during the SSW2009. This current paper reports the corresponding results on the thermosphere from the same simulation. Global features of SSW impacts are presented in terms of temperature and density.
3.1 Comparison with Thermospheric Density Observations
 As a validation of GAIA results on the thermosphere, we first compare with available observations. Figure 1 depicts thermospheric densities observed by CHAMP satellite near 17 and 05 LT at ~325 km altitude and that simulated by GAIA model. Longitudinally averaged data are used throughout the present study. The CHAMP density is normalized to Kp = 1 using NRLMSIS-00 model to minimize geomagnetic activity effects [Liu et al., 2011]. Good agreements between the observations and simulations exist in large-scale features, with prominent density drop during the SSW (pink line) at low and middle latitudes in both LT sectors. The obvious difference in absolute density values is due to difficulties in determining a true baseline for both the instrument and model. However, it is the temporal variation that is more pertinent to the physical processes in dramatically changing events like SSW.
 To examine the temporal variation more closely, the average density between 30°S and 30°N is extracted and line plotted in the third and fourth rows of Figure 1. Despite differences in absolute values, the same general trend emerges in both CHAMP and GAIA density. That is, the density starts dropping from day of year (DOY) 18/19, reaches a minimum near DOY 25/26, and gradually recovers afterward. Overlaid on this slow-varying trend, CHAMP density shows some rapid fluctuations during DOY 25–37. These might be remnant effects of imperfect removal of geomagnetic activity during the Kp normalization using NRLMSIS-00. As Kp briefly increased from 1 to 3 on DOY 26 and to 4 on DOY 35, density peaks are discernible around these days. Possible geomagnetic contribution to CHAMP observations was also pointed out by Fuller-Rowell et al. . Thus, the CHAMP density actually consists of a short time-scale (~1–2 days) perturbation driven by geomagnetic activity and a long time-scale (~30 days) perturbation driven by SSW. The GAIA density instead represents exclusively the SSW-driven component.
3.2 Local Time Dependence
 The polar-orbiting CHAMP satellite could sample only two local times 12 h apart during the SSW 2009, hence giving an incomplete picture. To complement this, we utilize the GAIA simulation. The good model-observation agreement on thermospheric density presented above, along with those in the ionosphere and stratosphere [Jin et al., 2012] warrants GAIA as a suitable alternative for investigating global features of SSW impacts.
 Figures 2a and 2b display the longitudinally averaged thermospheric mass density and temperature at 325 km altitude in tropical regions averaged between 30°S and 30°N. White line indicates SSW peak on DOY 23. A significant phase shift occurs around DOY 18–19, with temperature and density maxima shifting to earlier LT. To examine more closely perturbations during the SSW, we take the average density and temperature during DOY 1–10 as references and calculate the deviation from them. Figures 2c and 2d show perturbations averaged during DOY 25–30. Both density and temperature response are seen to strongly depend on LT, exhibiting a quasi-semidiurnal pattern with increase (warming) in the noon and pre-midnight sectors, but decrease (cooling) in the dawn and dusk sectors. The CHAMP observations near 05 and 17 LT fall both into cooling sectors, thus could observe only density decrease but not increase. Another feature to note here is that the warming magnitude (15 K near 21 LT) is significantly smaller than the cooling magnitude (−40 K near 17 LT). This indicates the drop of zonal mean during SSW, which is presented in section 'Global Net Effect'.
3.3 Height Dependence
 To see how persistent this quasi-semidiurnal feature is in altitude, we examine its height variation. Since perturbation patterns in density and temperature are nearly identical, only the temperature is presented in the following. Figures 3a and 3b display temperature perturbations in tropical and polar regions during DOY 25–30. In tropics (Figure 3a), the quasi-semidiurnal feature exists at all altitudes above 100 km, with interchanging warming and cooling sectors. Its phase remains constant above 200 km altitude, but propagates downward between 100 and 200 km. In northern polar regions (Figure 3b), warming of 20 K occurs at all LTs in a narrow layer between 100 and 130 km. Above 130 km, weak temperature perturbation occurs whose sign varies with LT. Below 100 km, little LT dependence is seen as generally known.
 Next, we examine temperature perturbations in the meridional plane. Figures 3c and 3d present perturbations in the warming and cooling sectors near 11 and 17 LT, respectively (warming and cooling refer to regions above 200 km in the tropics). At 11 LT in northern polar regions (Figure 3c), the well-known SSW feature appears below 100 km, with stratosphere warming of ~35 K and mesosphere cooling of ~30 K. In the lower thermosphere (100–150 km), warming of 10 K occurs north of 40°N while strong cooling of −30 K occurs near the equator. Above 150 km, warming occurs at most latitudes except for regions south of 30°S. The warming in northern polar thermosphere with peaks of ~15 K around 120 km is consistent with MIPAS satellite observations [Funke et al., 2010] and TIME-GCM predictions [Liu and Roble, 2002]. The GAIA model further reveals that this warming continues to extend to upper thermosphere around 60°N.
 At 17 LT (Figure 3d), the temperature perturbation below 100 km resembles that at 11 LT, reflecting the non-LT dependent characteristic below 100 km. Above 100 km, however, the perturbation structure significantly differs from that at 11 LT. In the lower thermosphere (100–150 km), warming occurs at most latitudes with peaks near the equator and northern middle latitudes. Above 150 km, cooling occurs south of 45°N, while slight warming of a few kelvin occurs around 60°N. The altitude dependence at northern middle latitudes shown in Figures 3c and 3d is consistent with ion temperature observations near 43°N [Goncharenko and Zhang, 2008], which show cooling above 150 km and warming around 120 km.
3.4 Global Net Effect
 The above results reveal that SSW impacts on thermospheric temperature significantly depend on LT, altitude, and latitude. To estimate the global net effect, we examine the zonal mean temperature perturbation (average over all LTs). Figure 4a shows that the zonal mean above 100 km is a slight warming in northern polar regions but cooling at other latitudes. Strongest cooling of −40 K occurs in southern polar regions. This zonal mean temperature drop during SSW has caused the asymmetric feature in Figures 2c and 2d, where the warming amplitude (15 K near 21 LT) is much smaller than the cooling amplitude (−40 K near 17 LT). When further averaged over all latitudes, the global net effect above 100 km is estimated to be cooling of approximately −12 K (see Figure 4b), with a cooling rate of about − 0.9 K/d during DOY 19–34. The high cooling rate during the 2 weeks period apparently cannot be explained by long-term seasonal variations. Furthermore, this upper atmosphere cooling is accompanied by the stratosphere warming, mesosphere cooling, and warming between about 80 and 100 km (see Figure 4c).
4 Discussions and Conclusions
 The GAIA model is used to simulate the atmosphere-ionosphere system response to the SSW 2009 event. The model results agree well with reported observations carried out at various time and locations. On top of these point-to-point agreements, the model reveals several distinct global features as discussed below.
 First, SSW impacts on the tropical upper atmosphere are primarily quasi-semidiurnal, with warming in the noon and pre-midnight sectors and cooling in the dawn and dusk sectors. This contrasts greatly to the lower atmosphere (below 100 km) where little LT dependence is seen (Figure 3a). This difference can be understood as the following. Although enhancement of semidiurnal and terdiurnal tides occurs at all altitudes during SSW, their magnitude is less than 2 K below 100 km but over 30 K above it [Jin et al., 2012]. Thus, the tidal signature (hence LT dependence) below 100 km is easily dominated by the zonal mean temperature changes and hard to be discerned in the total perturbation. The constant phase of the perturbation above 200 km altitude along with the downward propagation between 100 and 200 km seen in Figure 3a demonstrates the typical feature of an upward-propagating semidiurnal tides [Forbes, 1982].
 Second, due to the downward phase propagation, a rapid switch of SSW effects occurs near 150 km altitude above the equator (see Figures 3c and 3d). The switch direction depends on local time. It is from cooling below 150 km to warming above 150 km in sectors around 11 and 23 LT, but in the opposite direction near 05 and 17 LT. It would be interesting to compare these predictions with observations when available.
 Third, GAIA reveals a global net cooling of approximately −12 K above 100 km, with a cooling rate of about − 0.9 K/d during the SSW. At the moment, we do not have clear explanation for the rapid cooling during this period. However, one thing is clear that the seasonal variation cannot explain this. It is known that zonal mean temperature perturbations are largely caused by changes in the global circulation which are affected by various atmospheric waves [Matsuno, 1971; Liu and Roble, 2002; Pancheva et al., 2007]. Detailed analysis of the neutral wind, molecular diffusion, and ion drag in the GAIA simulation will be carried out to explore the underlying processes. In contrast to our result, a slight warming at 325 km altitude is reported by Fuller-Rowell et al.  using the Whole Atmosphere Model (WAM). This difference might be partly due to the fact that GAIA is an atmosphere-ionosphere coupled model, while WAM has no ionosphere included. However, closer examination is needed to clarify this.
 In summary, the GAIA model has revealed distinct local time and altitude dependence of SSW impacts on the upper atmosphere throughout 100–400 km altitude. These features provide us an urgently needed global context to better connect and understand upper atmosphere observations during SSW events.
 We thank S. Miyahara for helpful discussions and suggestions.