Previous studies have established the existence of a four-cell longitude pattern in equatorial F region ionospheric parameters such as total electron content and electron densities and in daytime, equatorial E × B drift velocities. This paper, for the first time, quantifies the longitude gradients in E × B drift associated with the four-cell tidal structures and confirms that these sharp gradients exist on a day-to-day basis. For this purpose, we use the Ion Velocity Meter (IVM) sensor on the Communications/Navigation Outage Forecasting System (C/NOFS) satellite to obtain the daytime, vertical E × B drift velocities at the magnetic equator as a function of longitude, local time, and season. The IVM sensor measures the E × B drift velocity in three dimensions; however, we only use the E × B drift observations perpendicular to B in the meridional plane. These observations can be used to obtain the vertical E × B drifts at the magnetic equator by mapping along the geomagnetic field line. The period initially selected for this work covers several days in October, March, and December 2009. We find, on a day-to-day basis, that (1) sharp E × B drift gradients of −1.3 m s−1 deg−1 exist in the western Pacific sector during equinox, (2) sharp E × B drift gradients of +3 m s−1 deg−1 are observed in the eastern Pacific sector during equinox, and (3) sharp E × B drift gradients of −1.7 m s−1 deg−1 exist in the eastern Pacific sector during December solstice.
 In the Earth's ionospheric F region, between 200 and 800 km altitude, the daytime distribution of electrons and ions as a function of altitude, latitude, longitude and local time are determined by ionospheric production, loss and transport mechanisms. Production is primarily through photoionization of atomic oxygen by solar EUV (λ < 91.1nm) radiation, and loss is through charge exchange of O+ ions with N2 and O2, to give NO+ and O2+, followed by recombination with electrons. Transport of ionization perpendicular to B is due to E × B drifts and transport parallel to B is due to ambipolar diffusion and the component of the neutral wind parallel to B. At low latitudes, the primary transport mechanism is via E × B drifts in the vertical and meridional plane. At the magnetic equator, these E × B drifts are upward in the daytime and primarily downward at night. The daytime upward drifts are responsible for producing crests in the F region peak electron density, Nmax, at +/− 15° to 18° magnetic latitude, known as the equatorial anomaly [Hanson and Moffett, 1966; Anderson, 1973].
 A recent technique has been developed to infer the daytime, vertical E × B drift velocity from ground-based magnetometer observations [Anderson et al., 2002]. Utilizing a magnetometer located on the magnetic equator (Jicamarca, Peru) and one off the magnetic equator at 6°N mag. lat. (Piura, Peru), Anderson et al.  developed various relationships between the observed difference in the H component, ΔH (HJic − HPiura), and the vertical E × B drift velocity observed by the Jicamarca Unattended Long-Term Ionosphere Atmosphere (JULIA) coherent scatter radar measuring the Doppler shift of 150 km echo returns. These 150 km E × B drifts have been shown to be essentially equivalent to F region E × B drift velocities by comparing them with the Jicamarca Incoherent Scatter Radar (ISR) E × B drifts. Anderson et al.  developed a neural network technique that gave realistic, daytime E × B drift velocities. The neural network was trained with over 450 quiet and disturbed days between 2001 and 2004, using 5 min observations of ΔH and JULIA E × B drift velocities between 09:00 and 16:00 LT. Figure 1 compares the average, ΔH-inferred E × B drift velocity for equinoctial, quiet days with the Scherliess and Fejer  climatological E × B drifts in the Peruvian, Philippine and Indian longitude sectors. The excellent comparisons give us confidence that realistic E × B drifts can be obtained from the ΔH technique. A subsequent paper by Anderson et al. , demonstrated that realistic E × B drift velocities could be obtained with the Peruvian sector-trained neural network, when applied to other longitude sectors where appropriately placed magnetometers existed, such as in the Philippine, Indonesian and Indian sectors.
 Recently, several observational studies have identified the existence of a low-latitude, four-cell longitude pattern in various ionospheric parameters. The first evidence emerged from IMAGE satellite FUV (135.6 nm) radiance observations after sunset (20:00 LT) that clearly showed enhancements in airglow-inferred Nmax values at the crests of the equatorial anomaly in four specific longitude zones during March–April 2001 [Immel et al., 2006]. They attributed the four-cell pattern to the effects of a four-cell pattern in daytime, vertical E × B drift velocities associated with the diurnal, eastward propagating, nonmigrating, wave number 3 (DE3) tidal mode [Hagan and Forbes, 2002]. Since the IMAGE observations were at night, the authors could not rule out a four-cell pattern in the prereversal enhancement in E × B drift that occurs after sunset. A subsequent paper by England et al. , however, established that the four-cell pattern was observed in CHAMP satellite in situ electron densities at 12:00 LT. Further, the four-cell pattern has also been observed in ROCSAT-1, daytime electron densities at 600 km [Kil et al., 2008] and in COSMIC occultation observations [Lin et al., 2007; Liu et al., 2010] and ground-based, global ionospheric total electron content (TEC) maps – GIMs [Wan et al., 2008]. In addition, four-cell patterns in daytime, vertical E × B drift velocities have been reported from DMSP observations [Hartman and Heelis, 2007; Kil et al., 2008] and ROCSAT-1 observations [Kil et al., 2007, 2008].
 More recently, Scherliess et al.  analyzed TOPEX/TEC observations from 1992 to 2005 and binned the data for quiet days into equinox, June solstice and December solstice periods and by local time. The local time period from 12:00 to 16:00 LT displayed the four-cell pattern in the same specific longitude sectors seen in the previous studies during the equinoctial season (see Scherliess et al.  for details). The TOPEX/Poseidon satellite incorporates a dual-frequency altimeter operating at 13.6 GHz and 5.3 GHz to observe ocean surface heights. The dual-frequency allows the total electron content (TEC) to be measured from the satellite altitude of 1336 km to the ocean surface. The data set of TEC observations studied by Scherliess et al.  covers the period from August 1992 until October 2005. For the current study, a subset of their database has been used covering the years 2001–2002. As discussed by Scherliess et al. , difficulties arise for a statistical analysis of the TOPEX TEC values owing to the slow precession of the satellite orbit (2°/d). In their paper, Scherliess et al.  normalized the TEC data to a common baseline in order to circumvent this problem and the same normalization has been applied in the current paper. In a nutshell, the normalization is accomplished by first finding the maximum TEC value for each ascending and descending pass between +/− 30° geomagnetic latitude. Next, the peak values are longitudinally averaged to give daily values, again for ascending and descending passes, separately. These daily values were used as normalization factors, where each 18 s TEC data point was divided by its corresponding normalization factor. More detail concerning this applied normalization is given by Scherliess et al. , whose figures refer to “relative” TEC values that have been “normalized” using these factors.
 In order to quantify the daytime, vertical E × B drift velocities at different longitude sectors that might explain the TOPEX/TEC observations, Anderson et al.  analyzed the magnetometer-inferred, daytime E × B drift velocities in the Peruvian, Philippine, Indonesian and Indian sectors for both the years 2001 and 2002. They binned all of the quiet-day, E × B drifts into three seasons and found the average E × B drift pattern for each of the three seasons and in each of the four longitude sectors. Figure 2a presents the average, daytime vertical E × B drift velocities as a function of local time in four longitude sectors for equinox periods in 2001 and 2002. Note the large difference in maximum E × B drift velocity between the Philippine sector and the Indonesian sector, 22 m s−1 versus 15 m s−1. These two locations are only 15 degrees apart in longitude. In Figure 2b, the TOPEX/TEC values are plotted for equinox, 2001 and 2002, between 12:00 and 16:00 LT. The bottom portion of Figure 2b is simply the average of the Northern and Southern Hemisphere TEC values plotted as a function of geographic longitude and absolute value of the geomagnetic latitude, to emphasize the location of the longitude gradients. The longitude locations of the Peruvian (blue curve), Philippine (red curve), Indonesian (purple curve), and Indian (green curve) sectors are indicated in Figure 2a. Note the very sharp gradient in TEC between the Philippine (∼125°E geographic longitude) and the Indonesian (∼140° geographic longitude) sectors. This sharp gradient at the edge of the cell is presumably caused by the sharp gradient in the daytime E × B drift velocity between these two sectors. From Figure 2b, there also appear to be sharp longitude gradients at 220°E and 320°E, which would imply sharp gradients in the daytime, vertical E × B drift velocities at these longitudes.
 The results presented in Figure 2 are unique and the comparisons between ground-based, inferred E × B drift velocities and the satellite TEC observations that relate to the four-cell pattern and its seasonal and longitudinal dependence have not previously been compared. This paper addresses two fundamentally important and specific scientific questions: (1) How sharp are the longitude gradients in daytime, vertical E × B drift velocities that define the boundaries of each of the four cells? (2) Quantitatively, are these sharp longitude gradients in E × B drift velocities observed on a day-to-day basis? Previous studies have shown that the four-cell pattern exists day to day [Sagawa et al., 2005; Immel et al., 2006; England et al., 2006; Wan et al., 2008; Immel et al., 2009], but this is the first study to determine, quantitatively, that the sharp gradients in E × B drift occur on a day-to-day basis. Answering these questions from an observational standpoint will set the “benchmarks” that are needed by the theoretical modelers to understand the physical mechanisms and to compare model results with observations.
 In order to answer the two scientific questions listed above, we have incorporated E × B drift observations from the Ion Velocity Meter (IVM), which is one of the Coupled Ion-Neutral Dynamics Investigation (CINDI) sensors on board the Communications/Navigation Outage Forecasting System (C/NOFS) satellite [de la Beaujardiere and C/NOFS Science Definition Team, 2004]. The Coupled Ion-Neutral Dynamics Investigation (CINDI) on board the C/NOFS satellite is composed of two sensors, the Ion Velocity Meter (IVM) and the Neutral Wind Meter (NWM). The IVM sensor measures (1) cross-track E × B drift velocities with an accuracy of 2 m s−1 and a sensitivity of 1 m s−1 and (2) along-track, with respective E × B drift velocities of 10 m s−1 and 5 m s−1. The NWM sensor measures neutral wind velocities (1) cross-track, with an accuracy of 5 m s−1 and sensitivity of 2 m s−1, and (2) along-track, with respective velocities of 10 m s−1 and 5 m s−1. The C/NOFS IVM sensor has been used to obtain the daytime, vertical E × B drift velocities at the magnetic equator as a function of longitude, local time and season. The IVM sensor measures the E × B drift velocity perpendicular to B and can be used to obtain the vertical E × B drifts at the magnetic equator by mapping along the geomagnetic field line. This is possible because the magnetic field lines are equipotentials and the daytime E × B drifts at the magnetic equator vary linearly with altitude between 200 and 800 km [Pingree and Fejer, 1987].
 There are a number of constraints we have adopted in organizing the IVM E × B drift observations: (1) We incorporate IVM observations only between 10:00 and 12:00 LT because this is the approximate LT window for the maximum E × B drift velocities at all longitudes. (2) We only utilize IVM observations below 500 km, which is low enough to ensure that O+ is the major ion. (3) The IVM observations are averaged over each degree of longitude. (4) For this study, the periods for IVM observations are primarily in October, March, and December 2009.
 In Figures 3a, 4a, 5a, and 6a, we display the IVM E × B drift velocities as a function of geographic longitude for a number of consecutive days in October, March, and December 2009. In Figures 3b, 4b, 5b, and 6b, the longitude coverage is overlaid on the TOPEX contours of relative TEC values. Bear in mind that the beginning of each curve on the left corresponds to 10:00 LT, while the end of each curve corresponds to 12:00 LT. Figure 3a displays two curves for 6 and 7 October, respectively. The geographic coverage for the 6 October, 10:00 to 12:00 LT coverage is from 65° to 90°E, while the 7 October coverage is from 85° to 115°E. The altitude of the C/NOFS satellite varies from 414 km at 10:00 LT to 405 km at 12:00 LT for both days while the respective geomagnetic latitudes vary from 1° to −5°. The IVM observed E × B drift velocities are roughly 35 to 40 m s−1. Between 65° and 115°E longitude, there are no sharp longitude gradients in E × B drift implying that this longitude window is in the middle of one of the four-cell patterns (Indian sector). This is confirmed by Figure 3b, which overlays the longitude window on the TOPEX/TEC contour plot.
 In Figure 4, three consecutive days in October (11, 12, and 13 October) present IVM E × B drift velocities in the western Pacific sector between 112° and 147°E geographic longitude. For each of the curves, the sharp longitude gradient in E × B drift occurs between ∼133° and 145°E longitude. Note that the slope in E × B drift velocity versus geographic longitude is roughly equivalent for all 3 days and is about −1.3 m s−1 deg−1. The altitude of the C/NOFS satellite between 10:00 LT and 12:00 LT decreases from 450 km to 410 km while the geomagnetic latitude increases from −9° to −15°. Pingree and Fejer  have shown that over this height range, the vertical E × B drift velocity is essentially constant. The change in geomagnetic latitude when E × B drift sharply decreases between 135 and 145°E longitude is only 2° from −10° to −12° geomagnetic latitude meaning that the sharp change in E × B drift with longitude is not due to the change in geomagnetic latitude. The boundary of this western Pacific sector cell is located roughly at the longitudes discussed in section 1, where the ground-based magnetometer-inferred E × B drift velocity gradients were observed. Again, we have overlaid the longitude window with the TOPEX/TEC contours in Figure 4b.
 In contrast, Figure 5 depicts the IVM-observed E × B drift velocity gradient in the eastern Pacific sector between ∼ 240° and 265°E longitude for three consecutive days on 19, 20, and 21 March 2009. For these 3 days the slopes in the gradients are roughly + 3 m s−1 deg−1. The altitude and magnetic latitude changes are from 405 km to 450 km and from −1° to 11°, respectively. Figure 5b clearly demonstrates that the boundary of the eastern Pacific sector cell is being observed in going from ∼240° to 250°E longitude. Figures 3–5 display the day-to-day consistency in the IVM observed E × B drift velocity (1) within one of the four-cell structures and (2) at the boundaries of two of the four-cell structures for equinox conditions.
Figure 6 displays boundary conditions for the eastern Pacific sector for 2 days during the December solstice period, 8 and 9 December 2009. For this case the longitude gradient in E × B drift is exactly opposite to the gradient pictured in Figure 5 and is approximately −1.7 m s−1 deg−1. Scherliess et al.  observed that during the December solstice period, there were apparently only three cells observed rather than the four-cell patterns during equinox and June solstice periods. This DE2, nonmigrating pattern has also been reported by Forbes et al. . The altitude and magnetic latitude variations are from 406 km to 415 km and from −7° to 0.5°, respectively.
4. Summary and Future Work
 This paper has addressed the two important, scientific questions posed in the Introduction. Using the C/NOFS/IVM observations of the meridional E × B drift velocities for specific days in October, March, and December 2009, we have demonstrated that extremely sharp longitude gradients in E × B drift velocity exist in the eastern and western Pacific longitude sectors and that these longitude gradients appear on a day-to-day basis in each of these sectors. Table 1 lists the longitude gradients in E × B drift velocities for the days in October, March, and December 2009 and their respective longitudes. It is also shown that these sharp longitude gradients exist on a day-to-day basis as depicted in Figures 4–6. The constraints listed in section 2 limited the number of useful days for C/NOFS/IVM observations.
Table 1. Longitude Gradients in E × B Drift Velocities for the Days in October, March, and December 2009 and Their Respective Longitudes
E × B Drift Velocity Gradient (m s−1 deg−1)
8 and 9 December
 We have tentatively established that the longitude gradients in E × B drift velocities at the boundaries of the four-cell structures are very sharp and that they occur on a day-to-day basis. The next logical step is to determine whether or not existing theoretical, self-consistent, atmospheric-ionospheric models can account for these sharp gradients in observed E × B drift velocities at the boundaries of the four-cell structures. If they can, then the model results can be analyzed to determine the causes of the sharp boundaries and their longitude dependence.
 We intend to theoretically investigate these sharp gradients by incorporating the IDEA model. This theoretical model has been described in a paper by Fuller-Rowell et al.  to demonstrate the impact of terrestrial weather on the upper atmosphere. The Integrated Dynamics through Earth's Atmosphere (IDEA) model consists of the Whole Atmosphere Model (WAM) – a general circulation model (GCM) – built on the operational U.S. National Weather Service (NWS) Global Forecast System (GFS) model and extends to an altitude of 600 km by increasing to 150 layers and taking into account the different physical processes that occur in the upper atmosphere (see Fuller-Rowell et al.  for details). The resolution of the model is 1.8° × 1.8° in latitude and longitude. The WAM model is coupled to the Global Ionosphere Plasmasphere (GIP) model through a self-consistent, global electrodynamics solver that provides GIP with global electric fields as an input. Successful comparisons between observed sharp E × B drift gradients and the theoretically calculated E × B drift gradients would represent a “landmark” capability in explaining the interactions between tropospheric and ionospheric mechanisms.
 The C/NOFS mission is supported by the Air Force Research Laboratory, the Department of Defense Space Test Program, the National Aeronautics and Space Administration (NASA), the Naval Research Laboratory, and the Aerospace Corporation. At the University of Colorado, this work is supported by AFOSR grant FA9550-09-0408. This work is supported at the University of Texas at Dallas by NASA grant NAS5–01068.