3.1. Seasonal Variations
 The low-latitude ionosphere TEC morphology exhibits large seasonal variations [e.g., Codrescu et al., 1999, 2001; Jee et al., 2004] due to seasonal changes in the equatorial vertical drift, in the strength and direction of the meridional neutral wind, in the neutral composition, and in the solar illumination of the northern and southern hemispheres. At equinox, the TEC pattern is largely symmetric about the geomagnetic equator, but large asymmetries are observed during solstice conditions [e.g., Jee et al., 2004]. This basic morphology can be understood by examining the solar illumination in the southern and northern hemisphere and the direction of the neutral meridional wind. During the solstices, the neutral wind blows from the summer to the winter hemisphere, raising the F region ionization in the summer hemisphere and lowering it in the winter hemisphere [Schunk and Nagy, 2000]. This wind effect, coupled with the asymmetry in the solar ionization and neutral composition in the two hemispheres, accounts for the observed asymmetries during the solstices [e.g., Liu et al., 2007]. Furthermore, seasonal variations in the equatorial vertical drift [Fejer et al., 1995; Scherliess and Fejer, 1999] lead to differences in the strength and location of the Appleton anomaly peaks from season to season.
 Consequently, in our analysis, we separated the normalized TEC data into their respective seasons. In detail, the data were separated into three 4-month long seasonal bins; Equinox (March–April, September–October), June solstice (May–August), and December solstice (November–February). In our initial analysis of the longitudinal TEC variability, data from all solar flux conditions were used. In order to minimize effects associated with geomagnetic activity, only data obtained during extended geomagnetically quiet conditions were used. For this analysis, geomagnetically quiet periods were defined when the average Kp index was below 3.0 for the 9 h prior to the time of the observation. The resulting data were binned into eight 3-h long magnetic local time bins (0000–0300 MLT, 0300–0600 MLT, up to 2100–2400 MLT) and into 15° × 1° geomagnetic longitude/latitude bins every 10° in longitude. Finally, in each of the 51,840 bins (3 seasons × 8 MLTs × 36 longitudes × 60 latitudes) the median value was calculated. In addition, the corresponding average values were also calculated for each bin, but no significant differences between the results using medians or averages were found. Also note that our criteria for geomagnetically quiet periods might not exclude perturbations in the low-latitude electric fields and neutral composition and winds associated with periods of extended geomagnetic activity [Fuller-Rowell et al., 1994; Scherliess and Fejer, 1997]. An extension of our Kp criteria from 9 to 36 h, however, did not result in significant differences in our median values.
 Figures 5a–5c show the median normalized TEC values as a function of geomagnetic latitude and longitude for equinox (Figure 5a), June solstice (Figure 5b), and December solstice (Figure 5c). Each figure is separated into eight panels corresponding to the eight magnetic local time bins starting in the upper-left panel near sunrise (0600–0900 MLT) and ending in the lower-right panel in the late night (0300–0600 MLT). In order to better geolocate the observed features on these maps, we shifted the geomagnetic longitude of each bin by 69° to the east. This applied shift approximately colocates the 0° longitude with the geographic 0°-meridian. The white areas in Figures 5a–5c indicate that TOPEX did not obtain observations in these longitude/latitude bins associated with its traversing over the continents. The color scale used for the normalized TEC maps ranges from 0.4 to 1.2 for all panels.
Figure 5a. Normalize TEC maps for eight 3-h long magnetic local time intervals during equinox conditions, showing the median normalized TEC values versus geomagnetic latitude and longitude (shifted by 69° to the east). The values correspond to the median for all solar flux conditions and geomagnetically moderate activity (9-h Kp average < 3.0). The color scale for the normalized TEC ranges from 0.4 to 1.2.
Download figure to PowerPoint
 The most prominent features seen in Figures 5a–5c are strong enhancements in the relative TEC variations that exhibit significant longitudinal variations. During equinoxes (Figure 5a) two separate enhancements in relative TEC develop after sunrise (0600–0900 MLT) over east Asia between about 80°–120°E longitude and over the Pacific Ocean at 170°–270°E longitude. Initially, these enhancements form near the geomagnetic equator and extend up to about ±10° in geomagnetic latitude on both sides of the equator. Over the Atlantic ocean, between about 300°E to 0°E longitude, these enhancements are absent after sunrise, indicating that the TEC values over the Atlantic are significantly smaller (≈20%) than the TEC values in the Pacific region at this local time.
 In the next panel, which shows the normalized TEC values 3 h later (0900–1200 MLT), the two equatorial enhancements over the east Asian and the Pacific region have moved poleward in both hemispheres and have separated into northern and southern enhancements corresponding to the development of the equatorial anomaly at these local times. Furthermore, the strong enhancement over the Pacific Ocean has began to split into two separate longitudinal enhancements on both sides of the equator located at about 170°–200°E and 230°–270°E longitude, separated by a local minimum located near 220°E longitude (more clearly seen in the southern hemisphere).
 This longitudinal separation becomes even more evident in the next panel from 1200 to 1500 MLT. At this time, the equatorial anomaly peaks are clearly established and nearly symmetrically located on both sides of the geomagnetic equator at about ±15° geomagnetic latitude. Three separate enhancements on both sides of the equator can clearly be identified over the east Asian and the Pacific sectors with their maxima located near 100°E, 190°E, and 270°E longitude. A fourth enhancement is also now evident over the Atlantic. The strongest enhancement at this time is over the east Asian region and the weakest enhancement is over the Atlantic.
 At 1500–1800 MLT the four enhancements continue with a strengthening over the Atlantic Ocean. The variations between the four maxima and their neighboring minima is of the order of 20–30%, which is similar to the relative variations between the anomaly enhancements and the equatorial valley regions. After sunset (1800–2100 MLT), the four peaks in both hemispheres are still clearly observable but are starting to diminish in peak strength. This weakening of the peak values is largely due to a smearing of the peaks due to a dependence of the geomagnetic latitude extent of the anomaly on the solar flux conditions at these local times. This can be understood by examining the dependence of the equatorial vertical drift on the solar flux conditions. During daytime, the equatorial vertical drift, which is the main driver of the anomaly, is nearly independent of the solar flux conditions [e.g., Fejer et al., 1995; Scherliess and Fejer, 1999], and as a result the locations of the anomaly peaks appear at nearly the same geomagnetic latitudes during low and high solar flux conditions. During the evening hours, however, the prereversal enhancement in the equatorial vertical drift as well as the nighttime drifts exhibit a strong solar flux dependence, resulting in a significant solar flux variation in the locations of the anomaly peak values. This leads, when data from different solar flux conditions are combined (as done in Figure 5a) to a washed out map of the enhancements. In section 3.2, we separate the equinoctial data into low and high solar flux conditions, which provides a better representation of the four low-latitude enhancements during the night.
 Immel et al.  used nighttime UV observations from the IMAGE satellite to study the longitudinal variation of the low-latitude ionosphere during 30 d in March/April of 2002. Their study revealed a global wave number four longitudinal pattern of the nighttime airglow intensity associated with the equatorial anomaly in agreement with our results. They speculated that the observed nighttime variations are the result of longitudinal variations in the amplitude of the diurnal atmospheric tides in the daytime E region ionosphere and a corresponding modulation of the daytime zonal electric field. Their nighttime observations, however, could not directly confirm that the observed longitudinal variation was not the result of a modulation in the F region wind near the dusk terminator and/or due to a longitudinal modulation in the prereversal enhancement of the F region vertical drift near dusk. England et al. [2006b] demonstrated that the wave number four pattern also exists in the daytime electrojet sheet current density, which provided further evidence for the importance of the daytime ionospheric E region in generating these patterns. Their study, however, could not directly relate their daytime E region observations to the nighttime F region UV observations. Our results clearly show that the observed wave number four longitudinal structure in the low-latitude plasma density is established during the daytime hours and is not the result of a longitudinal modulation in the prereversal enhancement near dusk. This strongly supports the interpretation that the wave number four pattern is the result of a longitudinal variation in the E region daytime tidal amplitudes and a resulting longitudinal variation in the daytime equatorial zonal electric field (vertical drift).
 The Immel et al.  study also reported a phase shift in the peak brightness of their nighttime UV observations of about 20° to the east when compared to the diurnal peak amplitudes of the temperature variation at 115 km as given by the Global Scale Wave Model (GSWM) [Hagan et al., 2001; Hagan and Forbes, 2002, 2003; Forbes et al., 2003]. A detailed inspection of Figure 5a reveals that the locations of our nighttime enhancements are shifted by about 20° to the east from the 1500–1800 MLT bin to the 2100–2400 MLT bin. For example, the peaks located near 100°E longitude and 270°E longitude at 1500–1800 MLT are shifted toward 120°E longitude and 290°E longitude at 2100–2400 MLT, respectively. From our analysis it is not possible to determine if this shift is the result of a virtual drift of the peaks (e.g., a decay on the western crest and a buildup on the eastern side) or due to an actual drift of the entire enhancement to the east. We speculate that this drift is the result of an upward-poleward electric field, causing the plasma to drift to the east, which is similar to the movement of low-latitude plasma bubbles during the night [Immel et al., 2003; Makela, 2006]. The eastward plasma velocity required to move the plasma over a 20°-longitude region in a 6-h period is approximately 100m/s, which corresponds to typical zonal drift velocities at equatorial and low-latitudes in the evening hours [Fejer et al., 1991].
 Figure 5b shows the normalized TEC values for June solstice conditions. The data correspond again to the mean values during moderate geomagnetic activity and all solar flux conditions. Similar to the development during the equinoxes, two separate enhancements develop after sunrise (at 0600–0900 MLT), one over east Asia (70°–120°E) and one over the Pacific Ocean (150°–260°E). During the next several hours, the Pacific enhancement separates in both hemispheres into two distinct peaks, which are located near 190° E and 260°E longitude at 1200–1500 MLT. However, in contrast to the equinoctial case, these two enhancements exhibit a strong hemispheric asymmetry, with initially larger values in the southern hemisphere at 0900–1200 MLT followed by larger values in the northern hemisphere at 1200–1500 MLT. A forth enhancement can also be seen to develop over the Atlantic Ocean with a peak location near the 0° meridian. The dominant enhancement during these times is, however, located over the east Asian region (60°–120°E), which is about 30% larger than the Pacific enhancements. This strong enhancement of the east Asian anomaly was pointed out by Vladimer et al.  and attributed to an enhancement in the daytime equatorial vertical drift velocity over this region. With regard to the wave number four pattern, Immel et al.  suggested that it is the result of electrodynamic effects on the plasma distribution of nonmigrating tides, which are associated with latent heat release during raindrop condensation over the tropics [Hagan and Forbes, 2002, 2003]. Currently, it is not clear if other nonmigrating tidal modes and/or stationary planetary waves play a role in the generation of the strong enhancement seen in the east Asian sector during this season. Clearly, a more detailed analysis, including modeling and data analysis studies, is necessary in order to better understand the observed variations.
 During the course of the afternoon and early evening hours, the strength of the east Asian enhancement diminishes and appears to drift eastward from a peak location near 90°E at 1200–1500 MLT to a location near 110°E longitude at 1800–2100 MLT. Furthermore, the two distinct enhancements over the Pacific region (near 190°E and 260°E) that could be seen at 1200–1500 MLT are combined at later times into one extended enhancement from about 190°E to 280°E longitude. This extended enhancement is the most dominant feature in the premidnight local time sector and can be seen all through the night.
 England et al. [2006a] noted that the wave number four structure is not seen during solstice periods in the UV nighttime observations. However, Figure 5b shows that during June solstice the wave number four structure is present during the afternoon but vanishes during the night when the UV observations were obtained.
 Figure 5c shows the normalized TEC values during December solstice (November–February) conditions. Initially, after sunrise a broad enhancement forms over the Pacific region extending from about 190°E to 300°E longitude and a second smaller enhancement appears over east Asia from about 80°E to 110°E longitude. These two enhancements further develop until about 1500 MLT, after which the enhancement over the Pacific weakens and the enhancement over the Atlantic becomes strong and broad (300°–20°E). This Atlantic enhancement is the major feature during the late evening and premidnight hours. Note that although some evidence of a weak wave number four pattern can be seen in the afternoon hours during December solstice this structure is largely absent or washed out during this season.
 Figure 6 shows an overview of the afternoon relative TEC variations for equinox (top), June solstice (middle), and December solstice (bottom) conditions. Clearly, the four separate enhancements can be seen during equinox and June solstice in both hemispheres. During December solstice conditions, however, the separate enhancements seen during the other two seasons over the east Pacific near 270°E longitude are absent, and instead, a broad enhancement is observed over the Pacific extending from 180° to 280°E longitude.
Figure 6. Overview of the normalized TEC variations versus geomagnetic latitude and longitude (shifted by 69° to the east) for the three seasons (top) equinox, (middle) June solstice, and (bottom) December solstice and for magnetic local times from 1200 to 1600 MLT. The results correspond to moderate geomagnetic activity (9-h Kp average < 3.0) and all solar flux conditions. The maps correspond to the median normalized TEC value in each bin and the color scale is the same as in Figure 5a.
Download figure to PowerPoint
 Owing to observational constraints, most of the previous studies of the wave number four pattern have focused on the vernal equinox (March–April) and only very limited evidence exists for the development of this pattern during the autumnal equinox (September/October). Henderson et al.  used 6 d of nighttime (2130 to 2230 LT) GUVI 135.6 nm UV observations from 2 to 7 October 2002 and found that the wave number four structure was also present during these times. England et al. [2006a] used one day of OGO-4 640.0 nm observations obtained on 4 October 1967 and they also found four peaks in the nighttime emission rates. However, it is not clear if the observations during these few days are representative of the typical autumnal equinox pattern. Figure 7 shows a separation of the afternoon equinoctial data shown in the top panel of Figure 6 into vernal (March/April) and autumnal (September/October) conditions. However, in order to ensure that there is a sufficient number of data points in each bin, the MLT bins were increased to a 4-h bin from 1200 to 1600 MLT. As mentioned earlier, these data correspond to the entire 13 years of the TOPEX/Poseidon mission (August 1992 to October 2005). During both equinoctial periods, the longitudinal wave number four pattern of the TEC variations can clearly be seen in Figure 7, with the weakest enhancement over the Atlantic Ocean and the strongest enhancement over the east Asian sector. Furthermore, the locations and relative strength of the enhancements and valley regions are nearly identical during both equinoxes, indicating that the generation processes are the same during both periods. Also note that the strong similarities between the vernal and autumnal equinoxes seen in Figure 7 justified the combination of the equinoctial data into one season in Figure 5a.
Figure 7. Maps of normalized TEC variations versus geomagnetic latitude and longitude for (top) the vernal equinox and (bottom) the autumnal equinox. The results correspond to magnetic local times from 1200 to 1600 MLT and moderate geomagnetic activity (9-h Kp average < 3.0) and all solar flux conditions. The maps correspond to the median normalized TEC value in each bin and the color scale is the same as in Figure 5a.
Download figure to PowerPoint
3.2. Solar Cycle Variations
 There are important solar cycle variations in the overall equatorial and low-latitude TEC morphology. For example, the magnitudes of the equatorial anomalies are strongly influenced by the solar flux conditions [e.g., Jee et al., 2004] with the larger values observed during high solar flux conditions (also see Figure 4). However, a detailed analysis of the low-latitude longitudinal TEC variations has not been done. In particular, is it not clear if the wave number four structure of the afternoon and evening low-latitude ionosphere is dependent on the solar cycle conditions. Previous studies [Immel et al., 2006; Sagawa et al., 2005, England et al., 2006a] were based solely on nighttime UV observations obtained during high solar flux conditions, and Immel et al.  speculated that the troposphere-ionosphere coupling will be greatest at the peak of the solar cycle when the E region densities are highest, which would result in a less pronounced, or even absent, wave pattern during low solar flux conditions. England et al. [2006a] noted that away from solar maximum the 135.6 nm emission rates decrease rapidly, making extended observations with FUV impossible and observations with GUVI more difficult and less accurate. On the other hand, the 13-year database of TOPEX TEC observations provides the same high-quality TEC observations during low and high solar flux conditions, and consequently, is ideally suited to study solar cycle variations in the longitudinal variability of the equatorial and low-latitude ionosphere.
 In order to investigate a possible solar cycle dependence of the low-latitude longitudinal TEC pattern, we separated our normalized TEC data according to the solar flux conditions. Specifically, we created seasonal maps of the afternoon (1200–1600 MLT) and evening (1800–2100 MLT) median relative TEC variations for low (F10.7cm < 130) and high (F10.7cm > 130) solar flux conditions. The resulting maps correspond to an average F10.7cm value of 89 and 171 for the low and high solar flux periods, respectively, and, as before, represent geomagnetically moderately quiet conditions (9-h Kp average < 3.0).
 Although the absolute TEC values strongly vary with the solar flux conditions [Jee et al., 2004], Figure 8 shows surprisingly that the general longitudinal TEC morphology is largely unaffected by the solar cycle conditions. In particular, during both solar cycle conditions the wave number four pattern can clearly be observed during the afternoon and evening periods at equinox and during the afternoon period at June solstice. These results indicate that the coupling between the lower atmosphere and the ionosphere is largely independent of the solar flux conditions.
Figure 8. Overview maps of (top) the afternoon (1200 to 1600 MLT) and (bottom) evening (1800 to 2100 MLT) normalized TEC variations for the three seasons (left) equinox, (middle) June solstice, and (right) December solstice and for low solar flux (upper row) and high solar flux (lower row) conditions. The maps correspond to moderate geomagnetic activity conditions (9-h Kp average < 3.0) and show the median normalized TEC value in each bin. The color scale is the same as in Figure 5a.
Download figure to PowerPoint
 A comparison of Figure 8 with Figure 5a also shows that the evening wave number four characteristics of the equinoctial TEC variability is better represented, as mentioned in section 3.1, when the data are separated according to their solar flux conditions. This can be understood by considering that the evening equatorial vertical drift velocities are solar flux dependent, which results in a solar cycle dependent extent of the anomaly peaks from the geomagnetic equator.
 Although the equatorial and low-latitude TEC values exhibit a clear longitudinal pattern that is fairly independent of the solar flux conditions and almost identical between the vernal and autumnal equinoxes, a significant amount of scatter remains in the TEC data that is not accounted for by the longitudinal variability. Figures 9a–9c show scatterplots of the afternoon (1400–1600 MLT) normalized TEC values versus geomagnetic latitude for equinox (Figure 9a), June solstice (Figure 9b), and December solstice (Figure 9c) conditions. In contrast to Figure 8, the MLT bin used in Figures 9a–9c was reduced from a 4-h-wide bin to a 2-h-wide bin in order to reduce the amount of scatter associated with changes in local time. The data correspond again to moderate geomagnetic activity conditions (9-h Kp average < 3.0) and low (F10.7cm < 130) and high (F10.7cm > 130) solar flux conditions. Each column shows eight scatter plots vertically stacked with each panel corresponding to a 10°-wide longitudinal bin located near the maxima and minima regions of the TEC enhancements seen in Figure 8. Superimposed are the median values and the first and ninth decile. Clearly visible in Figures 9a–9b9c are the longitudinal variations of the equatorial anomaly peaks about an average location of about ±15° geomagnetic latitude (indicated as the vertical dashed lines in Figures 9a–9b9c). Although our normalization and binning of the data has mostly eliminated variability in the data associated with variations in the solar EUV fluxes as well as variations associated with longitudinal variability, Figures 9a–9b9c show that a significant amount of scatter remains in the normalized TEC values. This scatter is largest near the poleward edges of the anomaly peaks and is of the order of about 40%. Furthermore, the scatter is largely independent of the solar flux conditions, the seasons, and the longitude of the observations. We speculate that this remaining scatter is most likely due to variability in the thermospheric neutral winds and composition, and to variability in the effects of high-latitude forcing on the low-latitude regions even during geomagnetically moderately quiet conditions. Furthermore, day-to-day variability in the tidal, planetary, and gravity wave forcing from the lower atmosphere is expected to contribute to the observed scatter. More detailed observational and modeling studies are needed to further understand the origin of this remaining variability in the equatorial and low-latitude ionosphere.
Figure 9a. Scatterplot of the afternoon (1400–1600 MLT) normalized TEC values versus geomagnetic latitude for equinox conditions. The data correspond to moderate geomagnetic activity conditions (9-h Kp average < 3.0) and (left) low solar flux and (right) high solar flux conditions. Each column shows eight scatterplots vertically stacked with each corresponding to a 10°-wide longitudinal bin. Superimposed are the median values (black circles) and the first and ninth decile (depicted as error bars).
Download figure to PowerPoint