The thermospheric mass density and zonal wind data used in this study are deduced from the observations of accelerometers on board the CHAMP and GRACE satellites. The CHAMP spacecraft was launched on 15 July 2000 into a near-circular polar orbit (inclination: 87.3∘) with an initial altitude of 456 km. By the end of the mission, 19 September 2010, the orbit had decayed to 250 km. The orbital period is about 93 min, thus circling the Earth about 15 times per day. The orbital plane covers all local times within 131 days, at a rate of 1 h in local time per 11 days. The Spatial Triaxial Accelerometer for Research (STAR) accelerometer developed by the Centre National d'Etudes Spatiales (CNES) on board CHAMP measures non-conservative forces exerted on the satellite with a resolution of < 10-9 m s-2 in the along-track and cross-track directions . This corresponds to a resolution of < 10-14 kg m-3 for mass density and < 10 m s-1 for the zonal wind measurements .

The GRACE mission consists of two identical spacecraft separated along-track by approximately 170–220 km. The twin GRACE spacecraft were launched on 17 March 2002 into a near-circular polar orbit (inclination: 89∘) with an initial altitude of about 490 km. By the end of 2012, the altitudes of the two spacecraft have gone down to about 460 km . The local time of the orbital plane precesses by 4.5 min every day, taking about 161 days to cover all local times. Both spacecraft are equipped with a three-axis SuperSTAR accelerometer with a resolution of < 10-10 m s-2 to observe the non-gravitational forces , which is about an order of magnitude better than the CHAMP mission.

The mass density data used in this study are derived from the calibrated accelerometer measurements in spacecraft x direction (along-track), together with the satellite attitude data, a structure model etc. Details about the processing of accelerometer measurements to obtain air mass density and cross-track wind data have been given by and references therein. In order to exclude variations caused by orbital height changes, the mass densities are normalized to common altitude. The method to obtain normalized densities at a fixed height is given as ρMea(h0)=ρMea(h)⋅ρMod(h0)ρMod(h), where h0 is the reference altitude, with values of 400 and 500 km for CHAMP and GRACE, respectively; h is the actual altitude of the satellites; ρMod(h) and ρMod(h0) are the densities calculated from NRLMSISE-00 at the satellite positions and at the reference altitude; ρMea(h) and ρMea(h0) are the observed density and normalized density, respectively. The wind data do not require altitude normalization since the thermospheric wind velocity depends only weakly on altitude.

The purpose of the study is to compare the tidal effects on the thermospheric mass density and zonal wind with that of the electron density; therefore, the same method for deriving the tidal components has been used as described in . The general mathematical formulation of the relationship between longitudinal patterns in satellite observations and the non-migrating tidal description in the Earth-fixed frame is given by and . In this study, the thermospheric mass density and zonal wind data has been first sorted into longitude (15∘) × local time (1 h) bins. Then, to suppress the prominent migrating tides, the longitudinal mean value has been subtracted hour by hour. After that, these mean-free data are processed by a Fourier transform in order to obtain the longitudinal patterns for each wavenumber up to 4 from which we identify visually the prominent tides. In the next step, a tidal model, containing all tidal and stationary components within the bandwidth of interest, is fitted to the longitudinal patterns. From the fitting we derive the amplitudes and phases of all tidal components. For validating our decomposition, we check the residuals, which are defined as observations minus synthetic signals, for systematic patterns. If no or small systematic patterns appear, the synthetic signal is considered reliable. The same approach has been used by for investigating the tidal signatures of the ionosphere at low and middle latitudes.

As we discussed above, the longitudinal wave structures observed by near-polar orbiting satellites can be caused by many different tidal components. By analyzing the propagating phase of the wave in local time, different tidal components can be separated. For each season, the mass density observations (ρ) is first sorted into MLAT (1∘) and geographic longitude (15∘) bins. Then, the mass density between ±40 to ±60∘ MLAT is averaged for each local time hour. To suppress the migrating tides the longitudinal mean value is subtracted hour by hour.

Figure 1 presents the local time versus longitudinal distribution of Δρ for four seasons in the Northern (top) and Southern (bottom) hemispheres both from CHAMP (left) and GRACE (right) mass density observations. Here Δρ means that the longitudinal mean value of ρ has been subtracted. At CHAMP altitude Δρ is found to be largest throughout the year around 150–30∘ W during 12:00–21:00 LT (except for the June solstice) in the Northern Hemisphere, while in the Southern Hemisphere a minimum value of Δρ appears in the same longitude region during the same local time. The peak value of Δρ in the Southern Hemisphere appears around 30–150∘ E during 12:00–21:00 LT, which is about 180∘ apart in longitude from the Δρ peak in the Northern Hemisphere. Interestingly, a mixture of longitudinal wave structures can be seen in the Northern Hemisphere during the June solstice. In the Southern Hemisphere during the December solstice, a second peak appears around 150–30∘ W during 21:00–09:00 LT. The two peaks can be considered as part of a wave-1 pattern, which is consistent with our earlier report about the wave-1 pattern of the southern MSNA feature .

The mass density at GRACE altitude exhibits mixtures of longitudinal wave structures in the Northern Hemisphere during the four seasons, the peak value of Δρ around 150–30∘ W during 12:00–21:00 LT is only prominent during the December solstice. In the Southern Hemisphere, the peak value of Δρ can also be found around 30–150∘ E during 12:00–21:00 LT. The larger value of Δρ around 150–30∘ W during 21:00–09:00 LT throughout the whole year makes the wave-1 pattern much more prominent in the Southern Hemisphere at GRACE altitude, compared to that at CHAMP altitude.

Figure 2 presents the amplitudes of the full non-migrating tidal and stationary spectrum for four seasons separately in the Northern (top) and Southern (bottom) hemispheres both from CHAMP (left) and GRACE (right) mass density observations. As can be seen, in the Northern Hemisphere SPW1 and D0 are found to be the most prominent stationary and tidal components throughout the year both at CHAMP and GRACE altitudes, but this is not so obvious around the June solstice. During the December solstice DW2 also shows comparable amplitudes. The semidiurnal tidal component SE2 is relatively large during the September equinox. In the Southern Hemisphere, spectral amplitudes are larger. SPW1, D0, and DW2 are found to be prominent throughout the whole year. The semidiurnal components SW1 and SW3 are also presented at all seasons with somewhat smaller amplitudes. Similar to the Northern Hemisphere, the tidal component SE2 shows comparable amplitudes during the September equinox in the Southern Hemisphere. To make it easier for the reader, the numerical values of amplitudes and phases for the two most important diurnal and semidiurnal tidal components are listed in Table 1.

As there are no zonal wind data from GRACE, Fig. 3 only presents the local time versus longitudinal distribution of zonal wind for four seasons in the Northern (left) and Southern (right) hemispheres from CHAMP measurements. As already mentioned above, the longitudinal mean value has been subtracted hour by hour to suppress the migrating tides. A mixture of longitudinal wave patterns are shown in the Northern Hemisphere, while a prominent wave-1 structure can be found in the Southern Hemisphere throughout the year. When compared to the longitudinal wave structures of the mass density in the Southern Hemisphere, the crest of wave-1 is found to be out of phase by about 12 h between these two neutral quantities. The amplitudes of the non-migrating tidal and stationary spectrum in the Northern (left) and Southern (right) hemispheres from CHAMP zonal wind measurements are presented in Fig. 4. Similar to Table 1, the amplitudes and phases of the two most important diurnal and semidiurnal tidal components from the zonal wind data are listed in Table 2. Prominent tidal components, D0, DW2, and SW1, can be found in both hemispheres throughout the whole year. Different from the tidal signatures in the electron and mass density, the amplitude of DW2 in zonal wind exceeds D0 during most of the seasons in both hemispheres (except the solstice seasons in the Southern Hemisphere). We will give a more detailed discussion about the tidal amplitudes and phases between the three quantities in the next section.

As already revealed by previous studies there are generally two major potential coupling mechanisms between the neutral atmosphere and the ionosphere. The first is upward propagation of non-migrating tides, since many of them are generated in the troposphere and stratosphere due to latent heat release and planetary wave/migrating tide interactions . Some of these tides can propagate directly to the upper thermosphere and generate the equally spaced patterns in longitude. The second is the E region wind dynamo. In this case, the non-migrating tides are not required to propagate all the way to the upper thermosphere, but electrodynamic coupling takes place between E and F regions, causing ion drifts at F region. As the E×B drift mainly affects the F region at equatorial latitudes, it will not be important for the longitudinal wave patterns of the plasma at midlatitudes. However, the wave patterns of the MSNA as reported by and imply that there should be some other coupling mechanism at midlatitudes. By model simulations, revealed that at low and middle latitudes the non-migrating tides of the zonal wind could be generated in situ through ion–neutral interactions due to the longitudinal ionospheric density variations imposed by the actual geomagnetic field configuration. Their results also show that the primary non-migrating diurnal tides excited in situ by ion–neutral interaction at middle latitudes are D0 and DW2; the secondary semidiurnal tides are SW1 and SW3 (see their Fig. 5).

Our observations confirm that the tidal components D0 and DW2 are clearly dominating the spectrum of thermospheric zonal wind at midlatitudes during all seasons (see Fig. 4). As opposed to the simulation, DW2 is on average slightly larger than D0. The phase of DW2 in the Northern and Southern hemispheres differs by about 12 h (see Table 2). The same is in principle true for D0. Unfortunately, reported no phase values for comparison. They suggested a possible generation mechanism for the patterns in the wind that is based on an in situ interaction between the diurnal migrating tide DW1 and the stationary planetary wave SPW1. Our result showed that the phase of the wind tidal components D0 and DW2 are rather similar during all the seasons (see Table 2). This means that the sum of tidal signatures has extrema near 0 and 180∘ longitude (cf. Fig. 3). Conversely, zero-crossing appears around ±90∘. These are the longitudes towards which the magnetic dipole axis is tilted. clearly showed that the prominence of the wave-1 tides appear when the actual geomagnetic field configuration is considered.

We use the same line of arguments to explain the prominent semidiurnal components SW1 and somewhat smaller SW3. In this case, it is the interaction of the migrating tide SW2 with SPW1. also predicted a prominent SW1 at middle latitudes and a smaller SW3. Also SW1 and SW3 vary practically in-phase and the two hemispheres vary again in anti-phase (6/12 h difference, see Table 2). Overall, our wind observations provide a good validation of the presented simulation results.

The ionospheric phenomena of MSNA have intensively been studied e.g.,. All these authors revealed that during local summer, the electron density in these regions is much larger at nighttime than at daytime. By using a similar approach as shown in Fig. 1 of , we subtracted the thermospheric mass density at daytime from the density at nighttime. But no prominent mass density enhancements can be found either in the Weddell Sea region or in the northern Atlantic (East Asian) regions during local summer. An explanation can be inferred from the large diurnal variations of the mass density that is masking the relatively small tidal effects. However, when the longitudinal mean value has been subtract to suppress the migrating tides, prominent wave-1 patterns of the mass density can be found in the Southern Hemisphere, which are similar to the wave-1 patterns of the southern ionospheric MSNA. For a quantitative assessment of relative tidal amplitude, we normalized D0 to the migrating tide DW1. During local summer we obtain at CHAMP altitude for the Northern (Southern) Hemisphere a ratio of mass density of 0.06 (0.12). The related ratio for electron density at the Northern (Southern) Hemisphere is 0.34 (1.0) (see Table 1 of ). This result shows that compared to the electron density, the effects of tide D0 in the thermospheric mass density is less by a factor of 6 to 8 at midlatitudes. As D0 is found to be the most prominent tidal component in both thermospheric mass density and electron density at midlatitudes; it may partly explain why the MSNA features can be seen directly in the F region electron density distribution, but not in the thermospheric mass density. Additionally, the interaction between the neutral wind and geomagnetic main field also contributes partly to the development of the ionospheric MSNA features. After subtracting the longitudinal mean, a coincidence of wave-1 structure in electron density, mass density as well as the zonal wind appears, implying that these non-migrating tides in the upper atmosphere are a result of a common driving mechanism. We think that the ion–neutral interaction in the upper ionosphere, as described by , is the primary cause.

As reported by , the northern MSNA shows a prominent eastward-propagating wave-2 pattern in electron density, and the dominating tidal and stationary components are found to be DE1, SPW1, D0, and DW2 (ordered by magnitude). However, this wave-2 pattern can neither be found in the zonal wind nor in thermospheric mass density observations in the Northern Hemisphere. Therefore, we believe that additional interactions might exist, for example, between the zonal wind and the geomagnetic field, additionally controlling the distribution of electron density in the F region. Further studies are needed for clarifying the issue.

The amplitudes and phases of diurnal tidal components DE1, D0, and DW2, as well as semidiurnal components SW1 and SW2 from electron density observations are listed in Table 3. A closer phase comparison of the most prominent tidal components in the ionospheric and thermospheric annually averaged quantities is shown in Table 4. The most interesting is tide D0. It appears at very similar phases in both electron and mass density, around 20:00 UT and 08:00 UT in the Northern and Southern hemispheres, respectively. This phase relationship for D0 in both neutral and ionization quantities suggests that an upward and downward motion of the thermospheric background controls at least partly the variation of electron density in the F region see. According to the Chapman layer formula, a larger atmospheric scale height enhances the ionization rate. For the other tidal components we cannot find such a phase relationship between neutral and ionized constituents.

Another mechanism exciting the tidal components in mass density is the converging and diverging zonal wind. As explained in the previous section, we find maxima of eastward wind shortly after midnight in the Northern Hemisphere (Fig. 3). This implies a convergence and air upwelling around 90∘ E in longitude as well as a divergence and downwelling around 90∘ W. The opposite wind direction are inferred from the tidal components D0 and DW2 shortly after noon. Consistent with this notion, we found at CHAMP altitude in the Northern Hemisphere mass density enhancements around 90∘ W and depletions around 90∘ E shortly after noon (Fig. 1). The opposite conditions prevail for density and wind in the Southern Hemisphere with mass density depletions around 90∘ W and enhancement around 90∘ E in longitude. This causal relationship suggest that the zonal wind is the main driver for the primary tides in mass density.

For completeness we want to comment on the appearance of the tide SE2 during the September equinox. This tidal component has earlier been observed (e.g., in thermospheric temperature at low-middle latitudes) and is related to wave-4 pattern caused by deep tropical convection in the troposphere e.g.,.