The Far Ultraviolet Imager (FUV) on board the IMAGE satellite provides an instantaneous global view of the OI 135.6-nm nightglow with 2 min time resolution. Because the OI 135.6-nm emission from the nighttime ionosphere is determined by the line-of-sight integrated plasma density, the nightglow images are useful for studying the nighttime low-latitude ionosphere globally. With the IMAGE/FUV 135.6-nm observations from March to June 2002, we have examined the global characteristics of the nighttime equatorial anomaly (EA) by constructing a constant local time map (LT map), in which pixels within an assigned local time range are extracted from the IMAGE/FUV nightglow images obtained over an observation period of 3 days or more and are put together to compose a global distribution map of emission intensities at that local time. These LT maps show that the development of the EA has a significant longitudinal structure, in which peaks and dips of the crest emission intensity and the crest latitude have about 90° longitudinal separation in the longitude range from 0° to 250°. Although there is not enough data over the American sector, this result suggests that the EA longitudinal structure has a prominent zonal component of the wave number 4. The observed longitudinal structure of the nighttime EA could not be fully explained by factors such as the empirical electric field and neutral wind models, the geomagnetic declination angle, or the displacement of the geomagnetic equator from the geographic equator. To explain the observed longitudinal structure of the EA, in particular, the wave number 4 feature, we may need to consider other forcing, for example, nonmigrating tide originated from the lower atmosphere.
 In the equatorial F layer ionosphere an eastward zonal electric field generated through both the ionospheric dynamos and magnetospheric process causes the upwelling of the plasma over the geomagnetic equator. The subsequent redistribution of the plasma along the magnetic field line creates two crests in the plasma density at both sides of the geomagnetic equator. This phenomenon is known as the “equatorial ionization anomaly” or simply “equatorial anomaly (EA).” Analyzing the world-wide ionosonde data available at that time, Namba and Maeda  first reported that there are two peaks of foF2 at both sides of the geomagnetic equator, although this phenomenon only became well-known after its first introduction to the western world by Appleton ; thus it is often called the “Appleton Anomaly.” In the 1960s, computer models of the EA were developed [e.g., Bramley and Peart, 1964; Hanson and Moffet, 1966], taking into account various processes such as electron production by photoionization, loss by recombination, diffusion along magnetic field lines, and an upward electromagnetic drift. With these works, it was shown that the electric field over the equator is an essential factor in the development of the EA. Without it, only a shallow and narrow “equatorial trough” could be expected. Bramley and Young  examined the effect of the field-aligned plasma flow driven by the neutral wind in the F layer on the EA development and showed that the asymmetrical EA profile about the equator can be accounted for by the neutral wind.
 The global structure of the EA has been studied by using data from the ground-based ionosondes distributed worldwide. Walker  discussed the longitudinal structure of the EA. He concluded that the following three longitudinal dependences are the basic reasons of the EA longitudinal dependences: (1) displacement of the magnetic and geographic equators, (2) magnetic declination angle at the magnetic equator, and (3) equatorial electric field and magnetic field. However, he pointed out these reasons could not explain all aspects of the observed longitudinal structure, in particular, a large change in relatively small longitude distance (about 40°), and suggested that forcing from below in the form of tide and planetary wave might be important in determining the EA development.
 Besides the ionosonde network, ground-based optical observations by using 630-nm emission from atomic oxygen showed a strong latitudinal peaks known as the “intertropical arcs” which corresponds to the enhanced ionization of the EA. Thuillier et al.  examined the 630-nm nightglow data with respect to its longitudinal variation, and concluded that the field-aligned plasma flow driven by the neutral wind in the F layer is responsible to the observed longitudinal variations. By using the global 630-nm nightglow data obtained by the UARS satellite, Thuillier et al.  further affirmed his original conclusion. Note that volume emission rate of the 630-nm nightglow depends on both O+ and O2 densities.
 With these early works, the basic physical processes that cause the EA are now well understood; that is, the equatorial electric field and the neutral wind are important factors in determining the EA development. However, these parameters can show a very complex behavior [Rishbeth, 2000; Abdu, 2001] and also are difficult to measure in comparison with the EA itself. It can therefore be claimed that an EA study at this time is a valuable tool for researching into the nature of the low-latitude electric field and neutral wind because a full understanding of the global behavior of these quantities is still lacking.
 The EA is the most prominent optical feature in the satellite nightglow observations not only at the 630-nm emission but also at OI emission bands in the far ultraviolet (FUV) wavelengths. In particular, as well as the 130.4-nm emission, the 135.6-nm emission from atomic oxygen excited through the recombination of atomic oxygen ions with electrons is one of the strongest lines in the FUV nightglow at low latitudes. Owing to its relatively higher transparency in the upper atmosphere, the 135.6-nm emission line is more suitable for sensing the ionosphere remotely than the 130.4-nm emission [Meier, 1991]. The first FUV nightglow observation was carried out by the OGO 4 satellite, in which the EA was clearly seen as the latitudinal distribution of the 135.6-nm emission intensity having two crests at both sides of the geomagnetic equator [Barth and Schaffner, 1970; Hicks and Chubb, 1970; Hicks and Chubb, 1974]. The 135.6-nm emission intensity corresponds to the line of sight integration of the product of O+ (nO+) and electron (ne) number densities that is approximately equal to ne2 in the F layer. We could ignore the minor effects of the transmission loss and the contribution from the secondary process which produces the 135.6-nm photon through negative ion neutralization reactions [Hanson, 1970]. Therefore the global imaging of the 135.6-nm nightglow is a powerful tool for diagnosing the ionosphere remotely. However, there have been few global images of the 135.6-nm nightglow taken by satellites until the IMAGE satellite was launched in 2000 [Carruthers and Page, 1976; Frank and Craven, 1988; Burch, 2000]. In the this study, we used the 135.6-nm nightglow imaged by the spectrographic imaging component of the Far Ultraviolet Imager (FUV) on the IMAGE satellite [Mende et al., 2000a, 2000b; Sagawa et al., 2003; Immel et al., 2003, 2004] to investigate the EA development on a global scale. The data used in this paper were taken from March to June 2002, when the apogee of the satellite reached to the low latitude from where the equatorial region is visible. Because the coverage of local time was changing with the speed of one degree per day, the nightside region was only visible during this period.
2. IMAGE/FUV 135.6-nm Nightglow Observations
 The IMAGE satellite was placed in a highly elliptical orbit (perigee altitude: 1000 km, apogee altitude: 7.2 Earth radii (RE)) in March 2000. Apsidal motion of the orbit has caused the latitude of apogee to move from 40° to 90° and back down to cross the equator in early 2003. In March 2002, the apogee latitude was sufficiently low to consistently view the northern nightglow arc due to the enhanced plasma density of the EA, with slightly improved viewing of the southern anomaly in May and June. During these months when the local time of apogee was in the evening sector, a clear view of two nightglow arcs from the evening terminator to postmidnight local times was provided along the geomagnetic equator. The FUV imaging cadence matches the rotation rate of the satellite, namely, about 2 min, with 5 s of integration time per image. The FUV consists of three channels, that is, WIC for the Lyman Birge Hopfield bands of N2, SI-12 for the Doppler-shifted Lyman alpha, and SI-13 for the 135.6-nm emission [Mende et al., 2000a]. In this study the SI-13 channel was used, which, with its 5-nm passband centered at 135.6-nm, is specifically designed for observations of the emission of atomic oxygen [Mende et al., 2000b]. For auroral or dayside studies, care must be taken to interpret the images with the knowledge that particular emissions of the N2 LBH bands also lie in the imager's passband. Since the molecular emission is only produced by direct impact of >∼10 eV electrons on N2, either photoelectrons or auroral electrons, it does not contribute to the SI-13 signal during observations of low latitudes on the nightside.
 Because typical nightglow intensity of the OI 135.6-nm emission near the EA peak is the order of 100 Rayleigh (R), and the SI-13 has an approximate sensitivity of ∼1 count/100 R/pixel [Frey et al., 2003], in this data analysis, we averaged five or ten images to improve the signal-to-noise ratio. This leads to a temporal resolution of 10 or 20 min, which is still short enough for valid investigation of most of the ionosphere phenomena at low latitudes. For each pixel of the nightglow images, we obtained the intensity of the OI 135.6-nm emissions integrated along the line of sight. We then simply converted these images to the maps of the column emission rate by multiplying the observed emission intensity at each pixel with the cosine of the look angle measured from the local radial direction to the look direction of the SI-13. The error due to this simple conversion becomes larger as look angle increases because the assumption of a thin and uniform emission layer becomes invalid.
Figure 1 shows a sequence of the 135.6-nm nightglow maps observed by IMAGE/FUV on 14 April 2002 from 1430 to 1755 UT. Each panel shows the 135.6-nm emission intensity map averaged over 20 min (10 images) and is projected on to geographic coordinates from 30° to 150° in longitude and from −45° to 45° in latitude. In mapping the observed image on to geographic coordinates, we assume a constant emission layer height of 350 km, which corresponds to the F layer height of a typical nighttime low-latitude ionosphere. Owing to the satellite location above the Northern Hemisphere, there is no data beyond −30°. Similarly, the dark triangular part in the bottom left corner of each panel is caused by the limit of the field-of-view (FOV). The bright area in the upper left corner is due to the dayglow. The OI 135.6-nm emission under solar illumination mostly comes from a photoelectron excitation process and is not an indicator of the ionospheric plasma density. Observation times of four panels were separated by about one hour. For the first panel (Figure 1a), the local time at the center (90° longitude) was 2030 LT and increased by 1 hour for successive panels. Kp indices preceding the observation time were 3+, 3+, 1, 1, and 4+, so it was a period of moderate geomagnetic activity. Clearly shown in each panel are the two bands of the intertropical arcs, which are located at both sides of the geomagnetic equator (the middle of three dashed lines). It is a topic of this paper that the intertropical arcs corresponding to the EA are not strictly parallel with the geomagnetic equator. This feature is easily seen in Figures 1c and 1d, in which the distance between two arcs is not constant along longitudes, so that it reaches a minimum around 60° and a maximum around 90° longitude. Moreover, as most clearly seen in Figure 1c, the emission intensity at the EA crest and its latitude again decrease toward 140°. This is more easily seen at the northern crest than at the southern one because of the incomplete observational coverage in the Southern Hemisphere. It is also noted that the look angles of the FUV instrument were relatively large in the Southern Hemisphere because the satellite was located over the Northern Hemisphere.
 One possible explanation of the longitudinal dependence of the EA shown in Figure 1 is the local time difference because each panel covers almost 6 hours in local time. However, the comparison of the four panels shows that the EA location at 60° longitude is consistently closer to the equator than that at 90° during the 4 hours observation period shown here. The figure therefore suggests the existence of the longitude dependence of the EA that can not be attributed to the change of local time. To confirm this suggestion, we further analyzed the IMAGE/FUV data by constructing a constant local time map (LT map), in which pixels having a given range of local time (1 hour in our case) are extracted from each SI-13 image (20 min average) to be assembled in to a map. To complete an LT map, we need the data for at least 3 days because the IMAGE satellite returns to almost the same point above Earth in 3 days; that is, the orbital period of the satellite is 14.23 hours, and five revolutions around Earth takes 2.97 days. This is close enough to assume a 3-day recurrence period because the coverage of a single image is more than 120° in longitude. The top panel of Figure 2 shows an example of LT maps constructed by using data taken on 25–27 April 2002. The range of local time is from 2200 to 2300 LT. In the panel, geomagnetic coordinates are used, although the magnetic longitude is adjusted so that the 0° coincides with the geographic 0° for easy identification of geographic location. As noted before, because the satellite was positioned over the Northern Hemisphere during the period, the coverage of the Southern Hemisphere was limited. This causes dark triangular shapes near the equator. Because the magnetic equator is located in far south in the American sector, the latitudinal coverage near the 300° longitude zone is not good. While the northern crest of the EA was well sampled, the coverage of the southern crest was inadequate for the detailed analysis. Therefore in this paper, we only discuss the EA in the Northern Hemisphere.
 Because the LT map shown in the figure is in geomagnetic coordinates, the longitude variation of the EA crest latitude can be seen easily. The figure shows that the peak emission intensities at the EA crest show considerable variation with longitude. In the bottom three panels of the figure are plotted, from top to bottom, the EA crest latitude, the emission intensity at the crest, and the crest-to-trough ratio of emission intensities, respectively. These numbers are determined by fitting the observed latitudinal profiles of the emission intensity with the Gaussian function. Except for the small-scale variation, there is a trend of the gradual increase of the EA emission intensity and the gradual decrease of the EA crest latitude with the increase of longitude from 0° to ∼250°. The EA characteristics from ∼250° to ∼300° are not clear because of insufficient latitudinal coverage of the observation. Region from ∼300° to ∼345° have different characteristics from other longitudes, that is, there is no well-defined EA crest but the equatorial trough is wide and deep. If we ignore this large-scale variation (this will be discussed later in this section) and focus on smaller-scale features than this, clear minima of all three parameters are seen around the 60° longitude. Also, there are other longitude regions where the emission intensity, the crest latitude, and the crest-to-trough ratio are all low (around 160° and 250° longitudes), while higher emission intensities, higher crest latitude, and higher crest-to-trough ratio are seen at regions around 15°, 120°, and 230° longitudes as indicated by three gray arrows in the figure. That is, peak-to-peak or dip-to-dip separations are close to 90° in longitude. Thus the observed longitudinal structure of the small scale seems to suggest a strong zonal wave number 4 component, except for the region near 300° longitudes, partly because this region was not sampled well due to the large southward excursion of the geomagnetic equator.
 As shown in three lower panels of Figure 2, the longitudinal structure of the EA crest latitudes, emission intensities at the crest, and the crest-to-trough ratios are all well correlated except around 300°. This can also be seen in Figure 3, which plots cross-sectional latitudinal profiles of the 135.6-nm emission intensity at selected longitude regions where the EA development is relatively enhanced (Group A: 10°–35°, 120°–145°, 210°–235°) or relatively depressed (Group B: 40°–65°, 150°–175°) in terms of the emission intensity. The figure shows that the EA crest latitudes are higher in Group A than those of Group B. It should be noted that due to a large-scale trend described above, the crest emission intensities of the Group A are not necessary larger than that of Group B. Also shown in the figure is a peculiar latitude profile in the Atlantic sector (C: 320°–345°), where no clear crest exists, and emission intensity is lower near the equator than in other longitudes.
 The longitudinal structure shown in Figure 2 also exists in LT maps for other local times. Two panels of Figure 4 are keograms of the EA crest latitudes (top panel) and the emission intensities at the crest (bottom panel) in rectangular coordinates of the local time (vertical axis) and the longitude (horizontal axis). Color-coded latitudes (or emission intensities for the bottom panel) versus longitude plots are stacked vertically from 2000 to 0300 LT. To reduce statistical fluctuation, the 9 days of data obtained from 29 April to 5 May 2002 are used to construct LT maps for this plot. During that period, geomagnetic activities were relatively low except on 29 April when Kp reached to 3. Although region near 300° longitude seems to have a different characteristics from other longitude as already shown in Figures 2 and 3, three blocks of relatively bright region can be seen below 270° longitude in both panels, and each block has a similar longitudinal width of about 90°, i.e., the wave number of 4. The figure shows that the small-scale longitudinal structure with the wave number of 4 basically holds throughout the range of local times presented here, though we could not see the fourth peak (dip) of this component to be found around 300° (345°). The pattern of the longitude dependence is almost constant or moves slightly eastward with the local time change. Also note that there is a similar longitudinal pattern for both the EA crest latitude and emission intensities, indicating that both correlate well.
 To examine the long-term persistence of the above-mentioned longitudinal structure of the EA, we constructed a set of LT maps for the 3-month period from the beginning of March to early June 2002. Each LT map is based on the data obtained over 3 successive days. Because the local time of the orbit plane of the IMAGE satellite moves at the rate of one degree per day, nightglow observation was partial before and after this period. From a set of the 2300 LT maps, the top panel of Figure 5 shows a keogram of the emission intensities at the EA crest. The vertical axis is the day of year 2002, and the horizontal axis is the longitude. From each LT map, a strip of color-coded 135.6-nm emission intensities versus longitude was made by using the same technique as that used for the lower panels of Figure 2. Then all strips were stacked vertically according to their observation time. In the bottom panel of the figure, averaged longitudinal profile of emission intensities are plotted with vertical lines indicating standard deviations (±1σ). The profile is calculated by averaging all longitudinal profiles shown in the top panel. Even though this is a crude way to process the data, a similar longitudinal structure as suggested in Figure 2 can be seen. We can identify two bright bands running vertically near 30° and 120° and two dark bands near 60° and 150°, though the figure is fairly noisy. Note that these bands well correspond to peaks and dips shown in the bottom panel. We can see another peak in the bottom panel near 230°, that is slightly off to the west of the longitude where the third peak of the wave number of 4 structure is expected (210°). We discuss about this shift of the peak location later in this section. Therefore peaks and dips shown here are roughly separated by 90° in the 0°–270° longitude range, suggesting the existence of the zonal component of the wave number 4. The longitudinal structure of the EA shown in Figures 2 thus exists consistently throughout the observation period. Also plotted in Figure 6 is the crest latitude versus the day of year 2002 with the same format to Figure 5. Again the EA crest latitudes correlate well with the emission intensities in the 0°–270° longitude range. Bright area near the upper right corner, as discussed bellow, may be due to strong field-aligned plasma transport due to a large eastward declination angle in the 300°–330° region.
 If the zonal component of the wave number 4 is a significant feature of the EA longitudinal structure, another peak near 300° and a dip near 330° are expected. After about day 120, the observational coverage in this longitude region became much improved because of the drift of the apogee latitude toward the equator. However, even after day 120, no clear peak and dip are shown in the longitudinal distribution of the EA of Figure 5. Moreover, the peak near 230° longitude seems to jump to ∼260° around day 140. This fact suggests that because the observation period runs for 4 months, we must take into account the seasonal change and other changes in geophysical conditions. Figure 7 plots the daily emission intensities at the EA crest at 2300 LT for longitudes 115°–140°. Also plotted are Dst indices (shown with the offset of 400 nT) and daily F10.7 fluxes. Because clear correlation between geophysical condition (the solar flux and/or geomagnetic activity) and the emission intensity is not easily seen from the figure, we are able to discuss the seasonal change of the EA without normalization to the geophysical condition.
Figure 8 plots the emission intensities averaged over 30° in longitudes versus the day of year 2002 at five longitudinal regions. Three lines correspond to regions of intense EA development (A: 15°–45°, C: 115°–145°, and D: 215°–245°), while the 50°–80° (B) longitudes is the region of the minimum EA development, and the 325°–355° (E) region is where another dip of the emission intensity would be expected if we assume the prevailing zonal component of the wave number 4 in the EA longitudinal structure. Nine days running average was applied to each data set to reduce short-term fluctuations. It can be seen that the long-term variation in the 215°–245° (D) and 325°–355° (E) regions differ each other significantly; that is, the emission intensity in the 215°–245° (D) region was lower near the equinox and higher in the solstice that is opposite to the 325°–355° (E) region. This could be related to the fact that declination angles in two regions are large and have opposite sign. Declination angles are almost 0° in the 15°–45°, 50°–80°, and 115°–145° regions, then increase to about 10° in the 215°–245° (D) region, and then decrease to about −20° in the 325°–355° (E) region. Because the observation period was from the vernal equinox to the June solstice, the neutral wind pattern was changing during the period so that magnitude of the field-aligned plasma transport driven by the neutral wind were also shifting. The field-aligned plasma transport modifies the EA development so that the asymmetry of two crests increases with increasing transport velocity [Hanson and Moffet, 1966]. Thermospheric neutral wind, as a first approximation, blows approximately from the subsolar to the antisubsolar points. This means that in evening hours near equinox, the direction of the neutral wind is expected to be mostly zonal (eastward) at the equator. The neutral wind provides momentum to the plasma through collision. Because the plasma can only move freely along the field line, the zonal neutral wind can not cause the field-aligned flow of the plasma as long as the declination angle is zero. On the other hand, at regions with a large declination angle, the zonal neutral wind can drive the field aligned plasma flow so that the positive (eastward) declination angle causes the northward plasma transport and vice versa. Therefore at the 215°–245° region, the plasma transport is expected to be northward, while at the 325°–355° region, it is southward. As the subsolar latitude changes with time, the neutral wind direction in evening hours also changes so that the northward plasma transport becomes weaker at the 215°–245° region because the angle between the neutral wind direction and the magnetic field becomes closer to perpendicular than before, while the southward plasma transport at the 320°–350° region is enhanced as June solstice approached. The field-aligned plasma transport driven by the neutral wind causes an asymmetry of two crests so that the downwind crest has a smaller plasma density and a lower crest latitude than those of the upwind crest [Su et al., 1997]. This difference in the field-aligned plasma transport could explain the difference of seasonal dependences at these two regions (lines D and E of Figure 8) and also could explain the absence of the EA enhancement in the 325°–355° region as expected from the wave number 4 structure where a large declination angle causes very strong EA asymmetry. It could be also due to the strong field-aligned plasma transport that the latitudinal profile of the emission intensity in this region did not exhibit a typical EA crest form as shown by the line C of Figure 3. Thus the zonal component of the wave number 4 could be masked in the western longitude region where effects of declination angles are large. In particular, the large declination angle in the 325°–355° region dominates the EA development as the June solstice approaches. Also it is speculated that the EA development is affected by the field-aligned plasma flow in the region of the eastern declination angle (200°–270°), where the shift of the emission intensity peak from the wave number 4 variation is observed.
 The IMAGE/FUV observations of the 135.6-nm nightglow show that the EA has a longitudinal structure persisting during the observation period from March to June 2002. Compared to that used in early studies, the data provided by the IMAGE/FUV instrument are unique because that the hemispherical distribution of the plasma density (integrated along the line of sight) is taken every 2 min with the spatial resolution of about 100 × 100 km. Even though we needed to integrate five to ten images for the EA analysis (10–20 min data), it has enough time resolution for studying the equatorial ionosphere. We can therefore discuss a much finer scale of the plasma distribution over the globe both in terms of time and space, although we have only studied the northern part of the EA extensively because of the limitation imposed by the instrument's FOV. Furthermore, the 3 1/2 months of continuous observation presented here covers from the vernal equinox to the June solstice only.
 The observed longitudinal structure is characterized by the strong zonal component of the wave number 4. Among dips of the longitudinal structure, there is a marked depression of the EA development near 60° longitude. Longitudinal dependence of the EA development has been reported by many authors [Walker, 1981; Thuillier et al., 1976; Thuillier et al., 2002; Su et al., 1997] and is attributed to the following three factors: (1) relative locations of the geographic and magnetic equators, (2) declination variation, and (3) electric and magnetic field variations. However, Walker  pointed out that significant difference in the EA development occurs over a relatively small changes in longitude (about 40°). Also as reported by Thuillier et al. , results of the 630-nm nightglow observations by the WINDII instrument on board the UARS satellite suggests the existence of small-scale (several tens of degree in longitude) variation of the EA (their Figure 8). However, it is hard to reproduce such a small-scale longitudinal structure by three factors listed above. The first factor, that is, the relative location of the geographic and geomagnetic equators, causes variation in ionization in the ionosphere due to the change in incident angle of solar EUV illumination over the geomagnetic equator. It also may affect how dayside winds influence the EA development. Roughly speaking, the geomagnetic equator is located to the north of the geographic equator in the eastern-longitude region, while this reverses in the western-longitude region, though the distance between two equators is much larger in the western-longitude region than in the eastern-longitude region. This is expected to result in the zonal component of the wave number 1. The second factor, that is, declination angle variation, affects the plasma transport along the field line via interaction between the plasma and the zonal neutral wind. It is almost zero in the eastern-longitude region, while it varies largely in the western-longitude region. While, as shown in Figure 8, our results suggest that the effect of the declination angle exists in the western-longitude region, there exists a large variation of the EA within the eastern-longitude region, where the declination angle changes very little. This means that the variation in declination angle is not a major reason for the small-scale EA longitudinal structure. There is a third factor, that is, changes in the magnetic and electric fields with longitude that result in the variation of the uplift velocity of the plasma over the equator, the major cause of the EA. The total magnetic field intensity at the equator has a distinct minimum near 300° and a broad maximum near 100°. This fact corresponds to faster uplift velocity near 300° than near 100°, if the zonal electric field is equal. Thus this reason alone could not explain the zonal wave number of 4 observed by IMAGE/FUV.
 Because the emission intensities at the EA crest show similar longitudinal dependence to that of the crest latitude, we speculate that the electric field plays a major role in the observed EA longitudinal structure. The low-latitude ionospheric electric field, which determines the EA development, is mostly determined by the E and F layer dynamos under a quiet geomagnetic condition. Basically, the electric field during daytime is determined by the E layer dynamo, and the F layer dynamo dominate the electric field during night owing to the low conductivity of the E layer after the sunset. The evening plasma drift shows a complex behavior during the transition period from the E layer dynamo to the F layer dynamo [Richmond, 1995; Eccles, 1998]. There is a so-called prereversal (or evening) enhancement of the upward plasma drift followed by the downward drift. The nighttime EA is strongly affected by the prereversal enhancement. Basu et al.  showed that the peak-to-valley ratio of the equatorial anomaly TEC is linearly dependent on the strength of the vertical drift velocity of the prereversal enhancement, according to drift speed observations taken by the Jicamarca radar and an ionosphere model.
 The longitudinal dependence of the equatorial electric field is not simple. According to the empirical electric field model developed by Scherliess and Fejer , the prereversal peak velocity of the upward plasma drift over the equator, which corresponds to the zonal electric field strength, exhibits both longitudinal and seasonal dependences. During equinox, it peaks in the Brazilian-African sector, but during the June solstice, the drift velocity maximizes in the Pacific region. Assuming that the peak drift velocity at the prereversal enhancement period correlates to the EA development [Basu et al., 2004; Whalen, 2004], its longitudinal variation predicted by their model is not consistent with our results that show variation with a much smaller scale. This empirical model is based on the AE-E drift meter and the Jicamarca radar data and employs four longitudinal nodes in cubic-B splines. It therefore may not have enough longitudinal resolution to be compared with the IMAGE/FUV observational results.
 In order to examine the global EA distribution based on the current empirical models of the electric field, the neutral atmosphere, and the neutral wind, we present results of model calculation in Figure 9. The SAMI2 (SAMI2 is Another Model of the Ionosphere) available as an open source software [Huba et al., 2000] calculates the ionospheric plasma along the Earth's dipole magnetic field line from hemisphere to hemisphere and includes ion inertia in the ion momentum equation. An empirical electric field model [Scherliess and Fejer, 1999] is used to calculate the E × B drift of flux tube. Empirical models provide neutral air composition and density (NRLMSISE00) and the wind field (HWM93). Note that by adopting the offset dipole magnetic field model, the SAMI2 includes effects of varying declination angle longitudinally, although it is not perfect. The model calculation is carried out for field lines located from −40° to 40° in latitude and from 0° to 360° in longitude under conditions as of 24 April 2002 at 1700 UT. Input parameters to SAMI2 are F10.7 = 177, F10.7A (3 month average) = 198, and Ap = 7. From the model ionosphere, we calculate the 135.6-nm column emission by integrating the volume emission rates vertically. Figure 9 shows the SMAI2 results by plotting the longitudinal variations of the crest emission intensity and crest latitude. Solid and dashed lines indicate results for the northern and southern EA crests, respectively. A difference between the northern and southern crests indicates an asymmetric development of the EA. The figure shows that the large longitudinal variation mostly occurs in the western longitude region due to the large displacement of geographic and geomagnetic equators and due to the large declination angle. Thus it is clear that the model could not reproduce the small-scale longitudinal variation as observed by IMAGE/FUV, particularly in the eastern longitude region. Also it is shown that the large north-south asymmetry in the EA development occurs in the western longitude region.
 As suggested by Walker , forcing from below might be a better candidate for explaining the small-scale longitudinal variation. With recent progress in the MLT (mesosphere-lower-thermosphere) modeling research, it was demonstrated that the nonmigrating tide originated in the tropospheric latent heat release has a significant amplitude in the E layer [Forbes et al., 1997; Hagan and Forbes, 2002, 2003]. Because the nonmigrating tide includes a stationary perturbation, it could cause longitudinal difference in the neutral wind field. Using the GSWM model, Hagan and Forbes  showed that the global distribution of temperature amplitudes of the nonmigrating semidiurnal tide at 124-km altitude (their Figure 10) has a strong wave number 4 component, particularly during equinox. Their figure shows that the peaks of temperature amplitude are located very close to the longitudes where IMAGE/FUV observed the lesser development of the EA, that is, 50°, 140°, 230°, and 320°.
 It is quite possible that the neutral wind field due to the nonmigrating tide causes modification to the E layer dynamo electric field, although the wind field could not propagate up to the F layer. The neutral wind field at the F layer altitude does not have a large nonmigrating component because a tidal wave from below lose the most of its energy below 150 km, so the local heat source at the F layer altitude is the solar EUV which is migrating. However, because the prereversal enhancement of vertical plasma drift over the equator is controlled by both E layer and F layer dynamos, it is quite possible that the EA during the night is also affected by the E layer dynamo [Millward et al., 2001]. It should be noted here that the development of the EA is a fairly slow process because it involves the uplift of the plasma at the geomagnetic equator and the succeeding redistribution of the plasma along field lines. We thus suggest that the nighttime EA development, which is an F layer phenomenon, may be controlled by the E layer dynamo through the prereversal enhancement of the zonal electric field. The IMAGE/FUV global observation of the EA development shows that the F layer plasma has a small-scale longitudinal variation, which may be related to the nonmigrating tidal component in the E layer neutral wind field.
 The global EA development has been studied by using 135.6-nm nightglow observations by the IMAGE/FUV instrument from March to June 2002. This observation provides an instantaneous global distribution of the nightglow, in which the EA is the brightest feature. Because of the instrument's FOV limits, only the northern part of the EA was studied extensively. By composing a constant local time map (LT map) from the more than 3 days of observation, the following results of the EA characteristics are obtained.
 1. The EA development has a significant longitudinal structure fixed to Earth. In particular, it is suppressed mostly in the region around 60° longitudes.
 2. Although the data set used in this analysis is limited as regards studying the American sector, where the geomagnetic equator is located at the far south of the geographic equator, the observed EA longitudinal structure suggests a prominent zonal component of the wave number 4.
 3. The previous explanation for the EA longitudinal dependence, that is the difference between geomagnetic and geographic equators, the declination angle, and electric and magnetic field variations, could not fully explain our observational results. For example, although the geomagnetic and geographic equators run almost parallel from 30° to 150°, there is a significant difference in the observed EA development in this region.
 4. The emission intensities at the EA crest show similar longitudinal dependence to that of the EA crest latitude. This result strongly suggests that the electric field plays a major role in the observed EA longitudinal structure.
 The wave number 4 spatial scale of the EA longitudinal structure observed by IMAGE/FUV is hard to be explained by the factors such as the difference between geomagnetic and geographic equators or empirical models of the low-latitude electric field and the neutral atmosphere/wind. We have found that forcing from below may contribute to producing the structure of this scale in the F layer ionosphere.
 The IMAGE FUV instrument is supported by NASA at the University of California, Berkeley. IMAGE FUV analysis is supported by NASA through Southwest Research Institute subcontract number 83820 at the University of California, Berkeley, contract NAS5-96020. This work uses the SAMI2 ionosphere model written and developed by the Naval Research Laboratory.
 Arthur Richmond thanks Mangalathayil Abdu and another reviewer for their assistance in evaluating this paper.