Longitudinal structure of the equatorial ionosphere during the 24 h local time period is observed by the FORMOSAT-3/COSMIC (F3/C) satellite constellation. By binning the F3/C radio occultation observations during September and October 2006, global ionospheric total electron content (TEC) maps at a constant local time map (local time TEC map, referred as LT map) can be obtained to monitor the development and subsidence of the four-peaked longitudinal structure of the equatorial ionosphere. From LT maps, the four-peaked structure starts to develop at 0800–1000 LT and becomes most prominent at 1200–1600 LT. The longitudinal structure starts to subside after 2200–2400 LT and becomes indiscernible after 0400–0600 LT. In addition to TEC, ionospheric peak altitude also shows a four-peaked longitudinal structure with variation very similar to TEC during daytime. The four-peaked structure of the ionospheric peak altitude is indiscernible at night. With global local time maps of ionospheric TEC and peak altitude, we compare temporal variations of the longitudinal structure with variations of E × B drift from the empirical model. Our results indicate that the observations are consistent with the hypothesis that the four-peaked longitudinal structure is caused by the equatorial plasma fountain modulated by the E3 nonmigrating tide. Additionally, the four maximum regions show a tendency of moving eastward with propagation velocity of several 10 s m/s.
 Recent airglow observations show longitudinal structure of four enhanced equatorial plasma regions. This four-peaked longitudinal structure was seen by Sagawa et al.  using far ultraviolet (FUV) 135.6 nm emission observation on board the NASA Imager for Magnetopause-to-Aurora Global Exploration (IMAGE) satellite during equinox-to-early summer of 2002. They found that the feature could not be fully explained by magnetic declination, magnetic field strength, displacement of the magnetic equator from the geographic equator or empirical electric field and neutral wind models. They proposed a possible explanation that the eastward wave number three (E3) diurnal or semidiurnal nonmigrating tide, excited from the lower atmosphere, propagates upward to the lower ionosphere and subsequently affects the E region dynamo electric field may be the possible explanation. When viewing zonal wind and temperature amplitude from a global constant local time (or Sun-synchronous) perspective, the E3 diurnal nonmigrating tide exhibits four lower thermospheric or E region maxima [e.g., Oberheide et al., 2003; Forbes et al., 2006] at the same longitude locations of the four strong equatorial ionization anomaly (EIA) zones [e.g., Immel et al., 2006]. Similar feature is also seen in the equinoctial seasons of 2002 by Henderson et al. , who used the outputs of an empirical model based on the Global Ultraviolet Imager (GUVI) [Christensen et al., 2003] observations on board the NASA Thermosphere Ionosphere Mesosphere Energy and Dynamics (TIMED) satellite.
 Later, Immel et al.  superimposed the global distribution of the nighttime equatorial anomaly peaks in the northern hemisphere observed by the IMAGE FUV and the E3 nonmigrating tide (represented in neutral temperature) simulated by the Global Scale Wave Model [Hagan and Forbes, 2002, 2003]. The surprisingly good match of the four enhanced plasma regions and the four-maximum-temperature tidal signatures supports the hypothesis made by Sagawa et al. . England et al. [2006a] used magnetometer instruments on board three low Earth orbiting (LEO) satellites to derive the noontime equatorial electrojet (EEJ) current density. They found a similar four-peaked feature in the EEJ current from each of the three satellites, and the locations of the four enhanced regions coincide with those of airglow observations. The coincidence provides more evidence to support the hypothesis proposed by Sagawa et al. , since the stronger daytime EEJ current would lead to a stronger equatorial eastward electric field and results in a stronger equatorial plasma fountain and EIA.
 Although the airglow FUV observations of the IMAGE and the TIMED satellites showed the longitudinal structure of four enhanced EIA zones, the vertical electron density distribution of the longitudinal structure was not observed. Lin et al.  used the global three-dimensional electron density map constructed by the radio occultation observation on board the six-microsatellite constellation. They showed that the four-peaked feature mainly exists above 250–300 km altitude, around F region height, supporting the hypothesis made by Sagawa et al.  and Immel et al. . This result suggests that the atmospheric tide influences the F region plasma through altering the E region dynamo instead of propagating upward and modulating the F layer directly.
 Although the above observations have provided evidences to support the hypothesis that the E region dynamo modulated by atmospheric tides, however, these observations were all limited to the nighttime period when the E region has almost disappeared. Therefore to further prove the hypothesis, it is very important to demonstrate that the longitudinal four-peaked structure also exists during daytime when the E region is present with its strong interaction with the F region. It is the purpose of this paper to study the local time variations of the longitudinal structure based on radio occultation observations of the Formosa Satellite 3 (FORMOSAT-3).
2. FORMOSAT-3/COSMIC Observations
 The Formosa Satellite 3, also named as the Constellation Observing System for Meteorology, Ionosphere, and Climate (abbreviated as FORMOSAT-3/COSMIC or F3/C in short), is a constellation of six microsatellites, designed to monitor weather and space weather with its major payload, radio occultation experiment (GOX) instruments performing the radio occultation observations in both the troposphere and the ionosphere. Each microsatellite also has a triband beacon (TBB) transmitter to perform ionospheric tomography and a tiny ionosphere photometer (TIP) to observe the nighttime ionospheric airglow emission. In this paper, we mainly use the vertical electron density observations from the GOX payload. The constellation was launched into an initial circular low-Earth orbit at an altitude of 512 km and 72° inclination angle [Cheng et al., 2006] from the Vandenberg Air Force Base, California, at 0141 UTC on 15 April 2006. The six microsatellites were close to each other at the initial orbit. It will take about 16 months for the constellation to reach the mission orbit at around 800 km altitude, 72° inclination angle, and 30° separation in longitude between each microsatellite. Up to the last day of the data used in this manuscript, (as of 1 November 2006), the GOXs on four initial- and two mission-orbit microsatellites globally observe about 2500 vertical electron density profiles per day up to 500 (800) km altitudes.
 Averaged global ionospheric maps were constructed by binning measurements from 2 months (e.g., September to October 2006) of occultation data in every 2-h (or hourly), and taking median value of observations located in the same 2.5°-2.5°-1 km (longitude-latitude-altitude) grid. It is noted that observations during magnetically disturbed periods are excluded while binning the data.
3. Global Ionospheric TEC and Peak Altitude Maps
Figure 1 shows the global constant local time map of the electron density integrated between 400 and 450 km (here called local time total electron content (TEC) map and hereinafter referred as LT map) from 0000 to 2400 LT in 2-h segments. To better identify the wave-like four-peaked structure, Figure 2 shows the averaged TEC values of Figure 1 between ±15° magnetic latitudes. From Figures 1 and 2, we note that the four-peaked longitudinal structure starts to form at 0800–1000 LT in the four regions, west of South America, West Africa, India and Southeast Asia, and the central Pacific. The four regions move eastward and become most prominent at 1200–1600 LT. It is interesting to see that the four-peak feature becomes less prominent at 1600–2000 LT. At 2000–2200 LT, the four-peaked structure becomes more discernible again with weaker TEC magnitude. This trend continues until 0200–0400 LT. After that, the four-peaked structure disappears altogether. It is noted here that the altitudinal range of the daytime four-peaked structure varies at different local times. In general, the structure becomes more identifiable at higher altitude (figure not shown). Above 400 km altitude, the four-peaked structure exists at most local times, except during those times when the four-peaked structure disappeared. We therefore compare only the TEC integrated between 400 and 450 at different local times.
 To investigate whether the ionospheric peak altitude shows similar longitudinal variations, we construct ionospheric peak altitude maps at various local times in Figure 3, similar to Figure 1. Similar to Figure 2, the peak altitude values at the equatorial regions are plotted in Figure 4. From Figures 3 and 4, similar to TEC variations, the longitudinal structure in peak altitude starts to appear and manifest itself at the same local time periods as those in TEC. The major difference is that the four-peaked structure in peak altitude disappeared after 1800–2000 LT. Possible reasons accounting for this difference may be as follows. First, the electron content presented here is integrated between altitude ranges of 400–450 km, above the peak altitude. When a stronger plasma fountain effect brings more plasma to above 400 km altitude and accumulates there, it takes some time for the plasma to subside after the cessation of a stronger plasma fountain. Therefore the four-peaked feature of the electron content may last longer than that of the peak altitude. Second, the ionospheric layer height also varies with changes of the solar zenith angle (reduction of photoionization production) and the recombination loss effect, in addition to the equatorial plasma fountain effect that raises the ionospheric layer at the magnetic equator. Third, a transequatorial wind, which was thought not to change the ionospheric layer height near magnetic equator, may also actually increase the ionospheric layer height there during the evening hours [Maruyama, 1996]. The recombination and neutral wind effects make the longitudinal variation of the peak altitude more complex during the evening hours, especially when the prereversal enhancement effect is weaker during solar minimum condition. Therefore, it is more difficult to infer the four-peaked longitudinal variation of the E × B drift from the signature of the peak altitude variations during evening hours.
 The EIA is known to be produced by the equatorial plasma fountain, which lifts the plasma from the magnetic equator to higher altitudes, and then it diffuses down along magnetic field lines to higher latitudes, creating two ionization crests on both sides of the magnetic equator [Namba and Maeda, 1939; Appleton, 1946; Duncan, 1960; Hanson and Moffett, 1966; Anderson, 1973; Balan and Bailey, 1995; Rishbeth, 2000]. The two basic processes that affect the EIA formation significantly are the strength of the equatorial plasma fountain and thermospheric neutral winds [Balan and Bailey, 1995; Balan et al., 1997; Rishbeth, 2000; Abdu, 2001; Lin et al., 2005]. It is straightforward to conclude that a stronger plasma fountain lifts more plasma from lower to higher altitudes and results in stronger and poleward extended EIA crests. However, there are other factors that need to be considered. One is the faster downward diffusion at field lines with higher magnetic inclination. When a stronger plasma fountain transports more plasma upward and poleward to magnetic field lines with higher magnetic inclination, the plasma diffuses down to lower altitude faster than those at field lines with smaller magnetic inclination. This effect reduces the EIA plasma enhancement produced by a stronger plasma fountain effect and thus needs to be taken into consideration before concluding that a stronger plasma fountain effect can produce an enhanced and poleward extended EIA. On the other hand, stronger equatorward neutral winds help to sustain the plasma in higher altitude, resulting in an enhanced EIA strength. It is noted that the equatorward neutral winds may result in equatorward movement of EIA crests. These two effects are considered theoretically by Lin et al. , and they concluded that a stronger plasma fountain effect is indeed the major driver for producing a stronger and poleward extended EIA. If both the stronger plasma fountain effect and equatorward winds are present, the EIA strength would be enhanced more significantly, since the equatorward winds play a role in increasing the plasma accumulation at the poleward extended EIA crests. On the basis of Lin et al. , one can generally state that stronger and poleward extended EIA crests are mainly resulted from a stronger plasma fountain. The hypothesis put forward by previous authors [e.g., Sagawa et al., 2005; Immel et al., 2006; England et al., 2006a, 2006b; Lin et al., 2007] suggests that the stronger EIA zones are produced by a stronger plasma fountain effect in those four regions. To examine the hypothesis, Figure 5 shows two electron density cross-section plots, i.e., latitude versus altitude plots of the electron density, at meridians of stronger (−80°E) and weaker (−120°E) EIA regions at 1400–1600 LT when the four-peaked structure is significantly seen. Stronger and poleward extended EIA crests are seen in the South America region (−80°E) while equatorward contracted EIA crests are seen in the neighboring longitude region (−120°E) with weaker EIA. The comparison again supports the hypothesis of previous studies that the E3 nonmigrating tide modulated plasma fountain effect is the main cause of the four-peaked EIA structure.
 Since the strength of the EIA is positively correlated to the strength of the plasma fountain effect driven by equatorial upward E × B drift, we further compare diurnal variation of the intensity of the four-peaked structure with the empirical equatorial F region vertical drift model of Scherliess and Fejer , which is built based on climatological E × B drift observations in the equatorial region by both incoherent scattering radar (ISR) and satellite. From the Scherliess and Fejer  model, during equinoctial seasons under low solar flux condition, the daytime E × B drift turns from downward to upward at around 0600–0800 LT, reaches its maximum at around 1000–1100 LT, and decreases to its smallest at around 1700–1800 LT, prior to sudden increase of the upward E × B drift enhancement, known as the prereversal enhancement. Although there are approximately 2-h differences between the four-peaked TEC structure and the upward E × B drift, they are still in reasonable good agreements, since it took time for the plasma to be raised upward and then to diffuse down toward higher latitudes and accumulate there [e.g., Lin et al., 2005]. The four-peaked structure is formed when the E × B drift of the Scherliess and Fejer  model turns from downward to upward at around 0600–0800 LT, while the most prominent four-peaked structure occurred after the E × B drift reaches its maximum at around 1000–1100 LT. The weakening of the four-peaked structure between 1600 and 2000 may be due to the decrease of the upward E × B drift at around 1700–1800 LT. It is interesting to see that the four-peaked structure becomes discernible (but with smaller TEC, since the photoionization effect is much weaker) again at around 2000–2200 LT, after the sudden increase of upward E × B drift produced by the prereversal enhancement effect at around 1900 LT. Although this could apparently imply that the dynamo effect produced by the E3 nonmigrating tide may contribute to the prereversal enhancement effect, the conclusion is not straightforward as it will be discussed next.
 The peak altitude variations shown in Figures 3 and 4 show a trend similar to TEC variations during daytime, which again suggest that regions of TEC enhancement are resulted from stronger upward motion of the ionosphere or a stronger plasma fountain. Although the peak altitudes at 1800–2000 LT, around the prereversal enhancement period, are greater than at other local times the four-peaked structure becomes less prominent. This may be due to that the E3 nonmigrating tide has a weaker effect or no effect to the prereversal enhancement dynamo process. Additionally, other uncertain effects, such as recombination and neutral wind effects as discussed in the previous section, may also contribute to peak altitude variations at different longitudes. Therefore the longitudinal variation of the peak altitude becomes more complex and the four-peaked feature becomes less prominent after 1800–2000 LT. Unlike the recurrence of the four-peaked structure of the electron content, the four-peaked feature of the ionospheric peak altitude disappeared at 2000–2200 LT. Absence of the four-peaked feature in the peak altitude after 1800 LT is contradictory to the hypothesis made according to the recurrence of the four-peaked feature in electron content at 2000–2200, that there might be an E3 nonmigrating tide effect to the prereversal enhancement. Considering both the electron content and peak altitude observations and the fact that the peak altitudes vary more effectively with the E × B drift variations, it is more likely that the E3 nonmigrating tide contributes only with a small effect or no effect to the dynamo process of the prereversal enhancement. The four-peaked feature of electron content at 2000–2200 LT may be the residual effect of the enhanced daytime plasma fountain effect.
 It is interesting to see from Figures 2 and 4 that the four maximum regions of TEC and peak altitude show a tendency of moving eastward. Rearranging TEC values (TECs) from 0800 to 2400 LT in every 2-h interval, Figure 6 superposes the TECs from 0800 to 2400 LT. Note that the time labels in Figure 6 show only the central hour of every 2-h interval, that is, 0900 LT indicates data from 0800 to 1000 LT. The peaked TEC values at the four enhanced EIA regions show a general trend of eastward movement with some scattering westward movements in Figure 6. Further analysis of the propagation velocities of the four different regions is shown in Figure 7 by estimating the movement of each of the four peaks. From Figure 7 the propagation velocities in Central Pacific, South America, Africa, and Southeast Asia are 64 m/s, 88 m/s, 47 m/s, and 42 m/s, respectively. It is noted that the propagation velocity is much smaller than the phase velocity of the diurnal E3 nonmigrating tide. The phase velocity of the diurnal E3 nonmigrating tide can be derived as around 155 m/s by considering the Earth's circumference divided by the wave period and wave number. Parts of the difference may be explained by the satellite aliasing because the diurnal E3 nonmigrating tide is observed as a wave number four signature with an aliased phase speed of about 115 m/s. However, the physical processes that lead to the remaining differences are not yet clear and further modeling work is necessary to resolve this issue.
 Observations presented in this study show important results for better understanding of the local time diurnal variations of the longitudinal four-peaked structure in the equatorial ionosphere. These results will be useful to future modeling effort to study the physical mechanism responsible for the longitudinal structure. We summarize our main findings as follows.
 1. Global LT maps (Figure 1) indicate that the four-peaked longitudinal structure starts to form at 0800–1000 LT and become most prominent at around 1200–1600 LT. The structure is less distinguishable between 1600 and 2000 LT, and becomes discernible again at 2000–2200 LT. The longitudinal structure diminishes after 0000–0200 LT and disappears after 0400 LT. The time-sequence variations of the four-peaked structure show similar behavior as compared with the E × B drift of Scherliess and Fejer  model with 1–2 h delay.
 2. The peak altitude variations show similar trend as TEC variations before 2000 LT. This result again suggests that the four stronger EIA zones are caused by the E3 nonmigrating tide modulated equatorial plasma fountain. During evening hours, after 2000 LT, photochemical processes and/or neutral wind effects make it difficult to infer the strength of the plasma fountain from peak altitude variations.
 3. For the first time, the observational data allow us the opportunity to compare the shape of the EIA crests during the peak of the four-peaked structure at 1400–1600 LT in the enhanced EIA zone (around −80°E) and in the weaker EIA zone (around −120°E). The stronger and poleward-extended EIA crests in the enhanced EIA region again is consistent with the hypothesis.
 4. The four-enhanced EIA regions show a general tendency of eastward movement. The propagation velocities in Central Pacific, South America, Africa, and Southeast Asia are 64 m/s, 88 m/s, 47 m/s, and 42 m/s, respectively.
 CHL and CCH thank the FORMOSAT-3/COSMIC orbital operation team in National Space Organization (NSPO) for operating the satellite constellation to the mission orbit. NSPO is supported by the National Science Council of Taiwan. CHL thanks Jens Oberheide for valuable suggestions and discussions. The authors thank the two anonymous reviewers for their constructive suggestions and Clia Goodwin for the editorial suggestions. This work is partially supported by the Taiwan National Science Council under NSC 96-2111-M-492-001 and NSC 96-2628-M-008-005.
 Amitava Bhattacharjee thanks Inez Batista and another reviewer for their assistance in evaluating this paper.