3.4. Enhanced Airglow Region
 The quasiperiodic southward moving waves are observed in and south of the enhanced airglow region, which probably corresponds to the equatorial ionospheric anomaly. As shown in the examples in section 2, this anomaly always shifts from higher geomagnetic latitudes to lower latitudes from evening to midnight. To see this feature more clearly, we made seasonal averages of the 630-nm intensity keograms in Figure 11. Because the sky conditions were not good except for May–July, as shown in Figure 1, the plots of May–July contain ∼30 nights of data and the other three seasons contain less than 10 nights of data. Nevertheless, the shift of the enhanced airglow region from south (higher geomagnetic latitudes) to north (lower geomagnetic latitudes) is clearly seen for all the seasons. The shift begins at earlier times in May–July (at 1900–2100 LT) than in the other three seasons (at 2200–0100 LT). The airglow intensity of the anomaly is weaker in November–January than in the other seasons. This is probably because the geomagnetic equator is ∼10° north of the geographic equator at this longitude, so the plasma density on the magnetic flux tube near Kototabang must be smaller in the winter of the Northern Hemisphere, in November–January.
Figure 11. Average nighttime variations of 630-nm airglow intensity in the north-south meridian (keogram) at Kototabang, Indonesia, in four seasons. This keogram was made from all-sky images, so the vertical axis is proportional to the zenith angle from Kototabang (−10.4° MLAT).
Download figure to PowerPoint
 The equatorward shift of the enhanced airglow region probably corresponds to the reversal of electric field from eastward in the sunset terminator to westward in the night. The eastward electric field near the sunset terminator generates the equatorial ionospheric anomaly through the fountain effect (upward E × B drift). When the electric field turns westward after the sunset at the equator, the anomaly shifts to lower geomagnetic latitudes, because the westward electric field at the equator causes equatorward E × B drift at the magnetic footprint of both hemispheres.
 Scherliess and Fejer  have shown a seasonal and longitudinal dependence of these electric field variations based on measurements of vertical plasma drift by the Jicamarca radar and the AE-E satellite. According to their Figure 5, the upward plasma drift at the sunset terminator (evening enhancement) is most intense in the equinox seasons for all longitudes. In the longitudinal sector of −20°–+180°E (including Indonesia), the time when the vertical drift turns from upward to downward is around 1930–2000 LT and does not change much with the seasons during high solar activity. The intense upward drift in the equinox seasons may cause the time delay of the anomaly shift to lower latitudes in Figure 11 compared with the solstice seasons. It is not clear why the anomaly shift is late in November–January compared with May–July.
 One may argue that the enhanced airglow region in Figure 11 corresponds to the neutral temperature enhancement in association with the MTM. However, the intensity of the 630-nm airglow often reaches 300–500 R in this enhanced airglow region, which is more than a few factors larger than that at midlatitudes (typically less than 100 R). On the other hand, the enhancement in neutral temperature in the MTM is less than 100 K [Herrero and Spencer, 1982], which is only ∼10% of the typical thermospheric temperature (∼1000 K). Thus the MTM is not likely to be the cause of this enhanced airglow region.
3.5. Possible Generation Mechanisms of Quasiperiodic Southward Moving Waves
 On the basis of the above considerations, we now discuss the possible generation mechanisms of the observed quasiperiodic southward moving waves. One of the common features of these waves is that the phase front is mostly aligned in the east-west direction. This fact suggests the generation of the waves by gravity waves in the thermosphere rather than by the oscillating electric field in the ionosphere. The oscillation of the neutral atmosphere by gravity waves causes a similar oscillation of plasma in the F layer through ion-neutral collision. Because the plasma can move only along the geomagnetic field line, the north-south oscillation in the neutral atmosphere can push plasma up/down along the field line [Hooke, 1970]. As discussed in subsection 3.2, the upward/downward motion of plasma causes reduction/enhancement of the 630-nm airglow. As a result, only the meridionally propagating gravity waves, which have a zonal (east-west) phase front, would be observed as wave structures in the 630-nm airglow images. On the other hand, the polarization electric field in the east-west direction, which pushes ionospheric plasma up/down to produce airglow variations, cannot develop in the observed east-west phase front. Thus the oscillating electric field is not likely to be the cause of the quasiperiodic southward moving waves reported in this paper.
 The other supporting evidence for the neutral wind perturbation is that, for some cases, the southward moving waves also show an ESE-WNW tilt of the phase front, as shown in Figures 6 and 8. This fact suggests that the waves have a westward moving component, which is opposite to the eastward plasma drift typically observed in the nighttime. Actually, when we observe the southward moving waves with plasma bubbles in Figure 8, the bubble continuously moves eastward, but the southward moving wave seems to have a component of westward motion.
 There have been several previous observations of gravity wave signatures in the variations of ionospheric plasma near the equator, on the basis of radio sounding techniques, such as IS radars, ionosondes, and HF Doppler signals [e.g., Sterling et al., 1971; Röttger, 1981, and references therein]. Röttger  reported medium-scale traveling ionospheric disturbances (TIDs) observed through the HF backscatter and Doppler experiments in the equatorial zones of Africa and South America. These TIDs propagate out from the equatorial zone with a mean phase velocity of 210 m/s and have periods between 5 and 30 min. He suggests the source of the TIDs as the cumulonimbus activity in the equatorial troposphere. Recently, Djuth et al.  reported gravity wave signatures with a timescale of 20–60 min, which is similar to the period of the waves in this paper, in the incoherent scatter power profiles in the thermosphere using the Arecibo radar in Puerto Rico (18.3°N, 293.25°E). These waves detected by the radio-sounding techniques may correspond to the waves reported in this paper. Simultaneous measurements by optical and radio techniques would be needed to fully characterize the gravity waves in the thermosphere.
 If we assume that the observed southward moving structures are acoustic gravity waves, they would follow the linear dispersion relation of the gravity wave [Hines, 1960], i.e.,
where m (=2π/λz), N, u, c, k (=2π/λh), and H are the vertical wave number, Brunt-Väisälä frequency, background wind velocity, apparent wave phase velocity, horizontal wave number, and scale height, respectively. In the thermosphere at 200–300 km, the scale height is 30–50 km. The Brunt-Väisälä frequency, which is given as N2 = 2g/5H (g: acceleration of gravity), is 8–11 × 10−3 rad/s. For typical values of the observed waves of λh = 700 km and c = 300 m/s and assuming u = 0 m/s, the vertical wavelength λz is estimated to be 200–250 km for these scale heights. λz becomes 160–200 km in case of a background southward wind u = 50 m/s. Since the thickness of the 630-nm emission layer is ∼100 km, these values of vertical wavelength are large enough to produce the observed variation in 630-nm airglow intensity by causing vertical motion of the F layer plasma. The molecular viscosity, which is not included in derivation of equation (4), may increase the actual vertical wavelength in the thermosphere.
 By neglecting the small term k2 in equation (4), the condition m2 > 0 gives ∣u − c∣ < 2NH. The value of 2NH is less than 300 m/s throughout most of the atmosphere below 100 km. Thus the observed waves with c = 300 m/s may be evanescent or ducted in the lower atmosphere. This consideration implies two possibilities. One is that the wave may be ducted below the mesopause with some energy leaking to higher altitudes [e.g., Francis, 1973], in which case they can travel large distances horizontally between the source and the F region observation location. The other possibility is that the wave may be generated above the mesopause. One such example is the secondary waves generated in the mesopause region by dissipation of small-scale gravity waves [e.g., Vadas et al., 2003].
 The gravity wave scenario has, however, several unclear points that require explanation. One is that the waves almost always move southward. Northward moving waves are observed on only two nights during the 2 years of observation. From the airglow imaging measurement of the gravity waves in the mesopause region (altitude of 80–100 km), the propagation directions of small-scale gravity waves vary depending on the season due to the filtering effect of the background wind [e.g., Nakamura et al., 2001; Ejiri et al., 2003; Suzuki et al., 2004]. However, the phase velocity of the waves in this paper is ∼310 m/s, which is much faster than the background wind velocity from the troposphere to the thermosphere. Thus the wind-filtering effect does not work for the present case. The propagation direction of the gravity waves would depend only on the relative location of the gravity wave sources and the observation point.
 Recently, Nakamura et al.  reported that small-scale (<100 km) gravity waves in the mesopause region observed by an airglow imager at the Tanjungsari observatory (6.9°S, 107.9°E), Indonesia, always move southward throughout the year. By comparing satellite cloud images, they concluded that this is because the cloud activity in the troposphere is always north of Tanjungsari due to orographic turbulences in the Sumatra Island. The airglow imager at Kototabang (0.2°S, 100.3°E) is located ∼700 km north of Tanjungsari, where the clouds are almost always in the field of view. Thus these clouds above the Sumatra Island would not be the source of the observed southward moving waves. It is known that strong tropospheric convection caused by the Asian monsoon exists north of Indonesia during Northern Hemispheric summer season. This is a possible source of the observed gravity waves, though some other source would be needed during the Northern Hemispheric winter season.
 The other challenge to the gravity wave scenario is that the southward moving waves are observed only in and south (geomagnetically poleward) of the equatorial ionospheric anomaly, and are not observed north (equatorward) of the anomaly. This observation may suggest a relation between the southward moving waves and the equatorial anomaly, which is the plasma structure in the ionosphere rather than the structure of the neutral atmosphere. However, this observation is probably due to the height variation of the F layer in the vicinity of the anomaly. The F layer height equatorward of the anomaly is very high (often above 500 km), particularly at premidnight local time, due to the upward drift of the plasma at the sunset terminator [e.g., Bilitza, 1990; Anderson and Roble, 1981]. Thus in the equatorward side of the anomaly the 630-nm airglow comes from higher altitudes. If the gravity waves that cause the observed wave structures in the 630-nm images are confined in the lower thermosphere, the waves would not be observed in the equatorward side of the anomaly. The observed high occurrence of the southward moving waves in May–July (Figure 10a) may partly be related to the observed early shift of the anomaly to lower latitudes in May–July (Figure 11).
 Recently, Vadas and Fritts  derived a gravity wave anelastic dispersion relation that included molecular viscosity and thermal diffusivity. Because molecular viscosity and thermal diffusivity increase rapidly in the thermosphere, and are the primary dissipative mechanism for high-frequency gravity waves there, this dispersion relation enabled the determination of gravity wave dissipation altitudes within the thermosphere via ray tracing. As shown by S. L. Vadas and D. C. Fritts (The influence of solar variability on gravity wave structure and dissipation in the thermosphere, submitted to Journal Geophysical Research, 2005), the dissipation altitude of a gravity wave depends sensitively on its intrinsic frequency and horizontal and vertical scales, as well as on the temperature of the thermosphere. Hot thermospheres enable deeper penetration than cool thermospheres, and gravity waves with larger vertical wavelengths penetrate to higher altitudes than those with smaller vertical wavelengths. For the typical parameters of the observed waves with a period of 40 min and a horizontal wavelength of 700 km, the waves are dissipated at altitudes of 200–340 km for thermospheric temperatures of 1000–2000 K (S. Vadas, private communication, 2005). This estimation is consistent to the fact that the waves are not observed at equatorward of the anomaly.
 The scenario of MBW generation from the MTM is that the pressure maximum, which may correspond to the maximum airglow intensity, launches waves away from itself. This scenario may connect the enhanced airglow region and wave generation. For the present case, however, this scenario would not be applicable, because the southward moving waves are also observed just north (equatorward) of the airglow intensity peak, as shown in the example of Figure 2.
 Finally, we note that waves with similar periods (0.5–1.5 hours) and smaller wavelengths (100–300 km) are observed at midlatitudes (20–40° MLAT) as medium-scale traveling ionospheric disturbances (MSTIDs) [e.g., Garcia et al., 2000; Kubota et al., 2000; Shiokawa et al., 2003b]. They always propagate equatorward and westward and have a peak occurrence of more than 50% in the Northern Hemispheric summer season. If the southward moving waves are a common feature in the equatorial thermosphere at all longitudes, there must be some boundary latitude of the southward (poleward) moving waves in the equatorial latitudes and the equatorward moving MSTIDs at midlatitudes. Shiokawa et al.  suggested the equatorward limit of the midlatitude MSTID propagation to be ∼18°MLAT on the basis of airglow imaging observations at a southern island of Japan. Otsuka et al.  and Shiokawa et al.  have shown, based on simultaneous airglow imaging observations in Japan and Australia, that the midlatitude MSTIDs have geomagnetically conjugate structures, indicating that they are generated by an oscillating electric field in the ionosphere. Similar conjugate measurements of airglow imaging may help to identify whether the electric field plays some role in the generation of the southward moving waves reported here.