The TIMED/GUVI thermospheric composition (O/N2 ratio) over 2009 reveals clear wave-3 (summer season) and wave-4 (Fall season) longitudinal dependence. The wave peak to valley varies from 7 to 11% of the background O/N2. The low Kp (yearly mean of 0.9) and low solar EUV flux indicates that the wave features were not caused by geomagnetic activity or solar EUV variation. A more likely explanation is nonmigrating diurnal eastward propagating DE2 and DE3 tides. Model calculations indicate that a 10% variation in O/N2 makes 1/3 contribution to the nightside equatorial 135.6 nm intensities compared to the variation due to the E-region dynamo electric field. This study confirms that nonmigrating tides can penetrate toward the upper thermosphere, both O/N2 and the E-region dynamo should be considered in analyzing ionospheric density variations due to nonmigrating tides. Comparison with MSIS results indicates that the nonmigrating tides have stronger effect on O/N2 than the migrating tides.
 The thermosphere and ionosphere conditions are continuously affected by energy inputs from solar radiation [Burns et al., 2004; Forbes et al., 2006], particle precipitation and Poynting fluxes from the magnetosphere [Fuller-Rowell et al., 1994], and waves from the lower atmosphere [Fesen, 1992; Sharon et al., 2004]. In addition to the solar radiation and the magnetospheric energy inputs that are usually the dominant drivers, the large scale waves (such as tides) also play an important role, especially during solar minimum years where the high latitude energy inputs are comparably very weak and the magnitude of the eastward propagating diurnal tide with zonal wave number 3 (DE3) in the thermosphere maximizes [Häusler et al., 2010].
 The nonmigrating DE3 tide has been shown to have a significant impact on both the thermosphere and ionosphere. The origin of DE3 is the tropical troposphere where it is excited by latent heat release in deep convective clouds [Hagan and Forbes, 2002]. DE3 has been attributed to the satellite observed wave-4 pattern in the equatorial ionosphere anomaly [Sagawa et al., 2005; Immel et al., 2006], electron density [Lin et al., 2007; Lühr et al., 2007; Kil et al., 2007], electric field [Hartman and Heelis, 2007], electrojet [England et al., 2006; Lühr et al., 2008], total electron content (TEC) of the ionosphere [Wan et al., 2008], and simulation of wave 4 ionosphere vertical drift [Ren et al., 2010, and references therein]. It has been proposed that the dynamo interaction of the tides with the lower ionosphere (E-layer) creates dynamo electric field in the E-region. The electric field modulates the ionosphere electron density in the equatorial and mid/low latitude regions via E × B vertical drift. In addition to DE3, DE2 [Forbes et al., 2008] is also one of the strongest and as such ought to have significant impacts on both the thermosphere and the ionosphere.
 While the dynamo electric field is important for modulating the ionosphere electron density, it is also well known that the thermosphere composition affects the ionosphere significantly. So, could these tides also modify the ionosphere through this pathway? In the recent work based on CHAMP (∼400 km) and GRACE (∼500 km) accelerometer measurements, Forbes et al.  found evidence that tidal perturbations extend all the way to the exosphere, and impose longitude variability there in much the same way that has been documented for the lower thermosphere [Zhang et al., 2006]. The longitudinal wave 4 pattern was also seen in the CHAMP thermospheric zonal wind [Häusler and Lühr, 2009]. Furthermore, TIMEGCM simulations and CHAMP data provided consistent results on CHAMP-altitude wave 4 pattern in zonal wind due to DE3 [Häusler et al., 2010]. Global tide simulations also indicate that nonmigrating tides have non-negligible amplitudes in the upper mesosphere and lower thermosphere [Miyahara and Miyoshi, 1997]. Oberheide and Forbes  revealed a large influence of nonmigrating tides on nitric oxide (NO) variability (wave 3 and wave 4 patterns) in the lower thermosphere (100–135 km). They suggested that the DE3 tidal component caused the observed wave-4 variation. The combination of DE2 and SE1 (semidiurnal eastward wave number 1) caused the observed wave-3 variation. All these results show that tides do affect thermosphere density, wind and temperature. As the thermospheric composition (O/N2) plays an important role in the F-region ionosphere density [England et al., 2010], we have to ask the question: Do nonmigrating tides change O/N2? What is the magnitude of the change?
 In this paper, we report that the nonmigrating tides do modulate the thermospheric column density O/N2 ratio [Strickland et al., 1995; Zhang et al., 2004] by a factor up to 10%. Such O/N2 changes make significant contribution to the nightside F-region variation.
 The Global Ultraviolet Imager (GUVI) [Christensen et al., 2003] on TIMED satellite has been simultaneously measuring thermospheric emissions in either imaging or spectrograph mode since December, 2001. The imaging mode data include the emissions in five “colors”: H Lyman α (121.6 nm), OI (130.4 nm), OI (135.6 nm) lines, N2 LBHS band (140.0–150.0 nm), and N2 LBHL band (165.0–180.0 nm). This eliminates the errors in data products (such as O/N2 [Zhang et al., 2004], and mean electron energy in the auroral region [Zhang and Paxton, 2008]) due to non-simultaneous measurement of these emissions. In addition to the GUVI imaging data, the GUVI spectrograph data can also be used to estimate the O/N2 ratio. To minimize the effect of geomagnetic activities, we selected the GUVI data in 2009, a solar minimum year. GUVI has been operating in spectrograph mode since 2008 at a fixed looking angle on disk. We used the same method of Zhang et al.  to calculate the O/N2 based on the radiances in “colors”: 135.6 nm and LBHS.
Figure 1 (top) shows yearly mean of GUVI O/N2 in 2009. Note the mean Kp in 2009 is only 0.9 indicating a mostly geomagnetic quiet year. This suggests that the effect of geomagnetic activities on O/N2 in 2009 can be neglected or minimized. The clear wave number 3 pattern in the O/N2 plot (Figure 1, top) is then likely not due to the energy input at high latitudes, but is more likely due to large scale waves in the thermosphere, such as nonmigrating tides. Note that low O/N2 (green color) marked with “SAA” is contaminated by penetrating particles in the Southern Atlantic Anomaly (SAA) region. To examine the amplitude of the wave 3 pattern, the O/N2 between latitudes 5 and 30 is averaged to avoid SAA and is plotted in Figure 1 (middle). The difference between the peak O/N2 (0.544) and bottom (valley) O/N2 (.508) is about 7% of the mean O/N2 (.526). The wave 3 pattern is confirmed by the Scargel spectra [Scargle, 1982] (Figure 1, bottom) of the averaged O/N2. The spectral peak is not at exact 3, but at 2.7. This could be due to the fact that the yearly mean O/N2 covers different local time as GUVI orbit precesses at a rate of 12 hours of local time every sixty days.
 To examine the O/N2 response to nonmigrating tides, the GUVI O/N2 data have been averaged over a much shorter period (20 days, during which the precession is just 4 hours). Figure 2 is obtained using the data from DOY (day of year) 181 to 201, 2009. The Scargel spectra (Figure 2, bottom) of the average O/N2 between latitudes 5 and 30 indicates the spectral peak at 3.2, slightly higher than 3. The difference between the wave peak (0.51) and valley (0.47) is about 8% of the background O/N2 (∼0.48, see Figure 2, middle).
 In additional to the wave 3 pattern, the GUVI O/N2 also reveals wave 4 pattern in the autumn season (Figure 3 for DOY 261 – 280, 2009). The wave 4 pattern is confirmed by the Scargel spectra peaked at 4. The difference between the wave peak (0.64) and valley (0.57) is about 11% of the background O/N2 (∼0.60).
 Tides lead to changes in the exosphere temperature [Forbes et al., 2009], thermosphere zonal wind [Häusler and Lühr, 2009], thermosphere density (e.g., NO) [Oberheide and Forbes, 2008]. Considering that the thermosphere remains thermo-equilibrium over the period of the tides, the thermosphere temperature should vary in the same way as the exosphere temperature. Changes in the thermosphere temperature and the difference in the scale heights of O and N2 lead to variations in the local O to N2 density ratio so as the column O/N2 ratio. The correlation between O/N2 ratio and exosphere temperature has been confirmed [Meier et al., 2005]. All of these results indicate that tides can modulate the thermospheric O/N2 ratio. Figure 4 shows the MSIS86 O/N2 with migrating tidal effect only. This provides additional evidence that tides can cause O/N2 changes.
 It is well known that global O/N2 distribution and its variation are mainly driven by geomagnetic activities and inter-hemispheric background wind. The geomagnetic activities usually lead to O/N2 depletion around the auroral oval and sub-auroral oval region depending on the strength of geomagnetic activities. Figures 1–3 do not show such a feature indicating the effect of geomagnetic activities was very weak or negligible at the latitudes shown in the plots. This is consistent with the very low mean Kp in 2009. However, Figures 2 and 3 do show the season (inter-hemispheric background wind) effect. The summer hemisphere (northern in Figure 2, southern in Figure 3) tends to have lower O/N2. But the inter-hemispheric background wind (meridian wind) does not play any role for the wave 3 or 4 pattern in O/N2. The observed wave patterns are very likely due to tides. Our results on O/N2 wave 4 in September – October and wave 3 in June–July 2009 are consistent with the wave patterns seen in the ROCSAT electron density (H. Kil, private communication, 2009). This supports the concept that the wave patterns seen in the GUVI O/N2 were due to nonmigrating tides. Considering the relative fixed local time of TIMED satellite, the observed wave 3 patterns may be theoretically due to tidal components DW4, DE2, SW5, SE1 etc. The observed wave 4 could be due to DW5, DE3, SW6, SE2, etc. [Häusler and Lühr, 2008]. The strong DE3 detected in the autumn season [Häusler and Lühr, 2008; Oberheide and Forbes, 2008] suggests that DE3 is the major driver for the observed wave 4 variation in O/N2 (Figure 3). On the other hand, DE2 and SE1 peaked around June [Häusler and Lühr, 2008; Oberheide and Forbes, 2008]. DE2 and/or combination of DE2 and SE1 (suggested by Oberheide and Forbes ) may contribute to the observed wave 3 pattern in the O/N2. These results are consistent with the findings of Forbes et al.  that DE3 and DE2 are the most important tidal components for generating the E-region dynamo electric field.
 The wave 3 pattern in MSIS O/N2 (Figure 4, top) is due to migrating tides. Regardless the higher MSIS O/N2 (compared to the GUVI O/N2), the peak (0.945) to valley (0.895) difference in the O/N2 (due to migrating waves) at equator is about 5% of the background O/N2. This indicates that the nonmigrating tidal effect on O/N2 (GUVI observation) has the same or larger magnitude than the migrating tidal effect. Therefore, it is very important to include nonmigrating tidal effect in thermosphere models (MSIS does not include the effect of nonmigrating tides (J. Forbes, private communication, 2009)).
 Simultaneous observations of both the lower thermosphere and F-region ionosphere have shown that both DE3 and DE2 are capable of significantly affecting the ionosphere [England et al., 2009], with the DE3 usually being dominant, producing the well-known wave 4 pattern in the ionosphere. The DE3 effect is also seen in the GUVI O/N2 (Figure 3) in the autumn season when the DE3 amplitude reaches its peak in a year. However, as the DE2 has a longer vertical wavelength, it is entirely possible that its effects would be dominant over a vertical column of the thermosphere over a long (one year) period such as that observed by GUVI (Figure 1). The DE2 tide would produce a zonal wave 3 structure in a fixed local time frame, which matches that observed by GUVI (Figures 1 and 2). Thus, the DE2 is likely the cause of the modified thermospheric O/N2 seen, although some contribution from other tides cannot be ruled out without further observations.
 Assuming DE3 or DE2 tide causes the GUVI O/N2 patterns, we can assess how significant the O/N2 variation affects ionosphere as the production rate of the most abundant ionosphere O+ strongly depends on O while N2 strongly controls the loss rate of O+. Following the methodology of England et al. , we first perform a control simulation using the normal inputs to the SAMI2 model for equinox, running the model at 190° longitude to minimize the effect of the offset between the geomagnetic and geographic equators. In the second simulation, we imposed a change in the climatological O and N2 values that matches the amplitude of the change observed by GUVI. That is, a diurnal variation in the climatological O of 8% and a simultaneous variation in N2 of 4% that is out of phase with the O variation. This equates to a 10% variation in the ratio of these two species. As the GUVI observations represent a change in the total column O/N2, we apply this change in the SAMI2 model at all heights throughout the thermosphere.
 The resultant impact on the ionosphere can be assessed in a variety of ways, but perhaps the most instructive is to compare the impact of the change in O/N2 with the well-known wave 4 longitudinal changes in O+ FUV brightness [e.g., Immel et al., 2006] that have been attributed to changes in the E-region dynamo. Figure 4 (bottom) shows the simulated brightness in the 135.6 nm emission from O+ in the early evening from the SAMI2 model for the control, perturbed O/N2 and perturbed E-region dynamo cases [after England et al., 2010]. Figure 4 shows that the change on O/N2 observed by GUVI could produce a change in the O+ brightness that is ∼1/3 of the amplitude of that produced by changing the E-region dynamo fields. Thus, this observed change in O/N2 could be expected to have a significant impact on the ionosphere.
 The GUVI O/N2 data in 2009, a solar minimum year with low mean Kp (0.9) and low solar EUV flux, reveal wave 3 and wave 4 longitudinal variations during summer and autumn seasons, respectively. These waves are very likely due to nonmigraitng DE2 and DE3 tides. The wave peak to valley varies from 7% to 11% of the background O/N2. These results are consistent with other measurements (neutral density, zonal wind etc) and ionosphere electron density where DE3 and DE2 are the major sources of variations. Comparison with MSIS results shows that the effect of nonmigrating tides is larger than that of the migrating tides. This suggests that nonmigrating tides should be included in thermospheric models for a better and more realistic specification the thermosphere. Our model calculations indicate that a variation of 10% in O/N2 makes 1/3 contribution to the nightside equatorial intensities of 135.6 nm compared to the variation due to the E-region dynamo electric field by the nonmigrating tides. This study confirms that both O/N2 and the E-region dynamo electric field play important roles in the ionospheric density variations due to nonmigrating tides.
 This work was supported by NASA grants NNX07AC57G (Y.Z. and L.J.P.), NNX07AT80G and NNX07AG44G (S.L.E.).