We analyze in detail the zonal velocities of large-scale ionospheric plasma depletions over two conjugate stations inferred from OI 630 nm airglow all-sky images obtained during the Conjugate Point Equatorial Experiment (COPEX) campaign carried out in Brazil between October and November 2002. The conjugate stations were Boa Vista (BV) (geogr. 2.8N, 60.7W, dip angle 22.0°N) and Campo Grande (CG) (geogr. 20.5S, 54.7W, dip angle 22.32°S). Over Campo Grande, the zonal velocities were measured also by a system of spaced GPS scintillation receivers. The airglow zonal velocities at the conjugate sites were seen to agree very closely, except for a slightly increased velocity over CG which we attribute to the presence of the geomagnetic anomaly. The results show a high degree of alignment of the bubbles along the geomagnetic field lines during the bubble development phase and as the bubbles travel eastward, thereby suggesting that the neutral zonal wind effect in the zonal plasma motion is an integrated effect along the flux tube. The zonal velocities obtained from the GPS technique were always larger than those calculated by the airglow technique, which permitted observation of zonal plasma velocity shear between the altitudes of the airglow emitting layer and of the GPS scintillation. Theoretical ambient plasma zonal velocities calculated using the formulations by Haerendel et al. (1992) and Eccles (1998) are compared with the experimental results. Our results also reveal some degree of dependence of the zonal velocities on the solar flux (F10.7) and magnetic (Kp) indices during the COPEX period.
 Plasma bubbles develop in the postsunset equatorial ionosphere, and large-scale plasma bubbles, extend over a remarkably large area of the South American continent and almost all of the Brazilian territory as seen in the all-sky images in the 630 nm emission (Figure 1). Figure 1 also illustrates drastic variation of the alignment of the magnetic equator over South America. The plasma bubbles along with plasma irregularity scales spanning about 6 orders of magnitudes 0.1–105 m [Kelley, 1989, p. 154], constitute the phenomenon of equatorial spread F (ESF). The wide spectral distribution of plasma irregularities are well known to cause interference with satellite-based telecommunication channels, while they are also known to sustain frequent interhemispheric radio amateur communication in the VHF bands 50 MHz - 432 MHz (personal communication from a group of local radio amateurs, February 2007).
 In this work we study plasma bubble zonal velocities simultaneously measured at two geomagnetically conjugate sites in Brazil. A previous study of plasma depletions using OI 630 nm airglow imagers at the conjugate sites of Sata, Japan (geomagnetic latitude 24°N), and Darwin, Australia (geomagnetic latitude 22°S), carried out by Otsuka et al.  has shown symmetry conditions of plasma depletions with respect to the geomagnetic equator. That study was conducted at conjugate sites farther away from the equator and hence permitted the investigation of vertically extended and predominantly fossil plasma depletions. The present experiment involving conjugate stations closer to the equator, at ±12° geomagnetic latitude, permits study of the depletions in their growth as well as fully developed stages. The observations reported here were performed from October to December 2002 as part of the Conjugate Point Equatorial Experiment (COPEX) campaign. The main scientific objectives of the COPEX campaign and some complementary results are presented in a companion paper by Abdu et al. . As explained by Abdu et al. , an important logistical motivation of the COPEX campaign was the consideration that the north-south geographical extension of the Brazilian territory permitted ionospheric conjugate point observations satisfying the condition that the conjugate E region magnetic field lines map to the dip equatorial F layer peak and bottom side heights.
 The COPEX campaign ionospheric instruments were deployed at three sites. Two of them at conjugate locations are Boa Vista (geogr. 2.8° N; 60.7° W, dip angle 22.0°N) and Campo Grande (geogr. 20.5° S; 54.7° W, dip angle 22.3°S) where the all-sky imagers were operated, and a third site was Cachimbo located near the magnetic equator. We use data from only the conjugate sites, henceforth referred to as BV and CG, respectively. The observational results of the plasma bubble zonal velocities are compared with theoretical calculations of ambient plasma zonal velocities. The results show that the zonal velocity of the ambient plasma is approximately equal to the bubble zonal velocity later in the evening. Theory predicts that the expected zonal velocities of the bubbles which is equivalent to the ambient plasma zonal drift velocity should be approximately equal to that of the zonal wind velocity conductivity weighted along the flux tube. The results from the present analysis are discussed below.
2. COPEX All-Sky Imagers: Setup and Data Analysis
 This section describes the experimental setup of the all sky imagers and the techniques used in the determination of the plasma bubble zonal velocities. At the sites BV and CG chosen for this experiment special arrangements and temporary infrastructure were set up with the collaboration of the Brazilian Air Force and UNIDERP University to carry out the experiments of this campaign.
2.1. OI 630 nm All-Sky Imager Data
 Digital all-sky images of the OI 630 nm airglow were obtained at BV and CG during the nights of 2, 4, 8, 10 and 11 October 2002 and 1 November 2002. Figure 1 shows an example of a pair of conjugate point all-sky images on the night of 8 October. The maps correspond to a radius of ∼1800 km at a reference height of 250 km centered over the imager zenith. The northern image is surrounded by a thin dark rim that contained no airglow intensity information and reduced the apparent radius of the image. The two objects that appear in the images, at top left in the CG image and the right hand side of the BV images are parts of nearby objects in the imagers field of view. They do not affect in any way our determination of the depletion zonal velocities. The series of images presented in Figure 2 in the night of 8 October 2002 clearly shows an eastward propagation of the depletions in the successive images.
2.2. Bubble Zonal Velocity Determination
 The signatures of the bubbles in airglow images are wavelike structures with near magnetic north-south aligned fronts. A zonal cut through the images presents intensity minima that are used as reference in successive images to calculate the zonal velocities by the method described by Sobral and Abdu . Wavelike signatures can result also from a meridional cut through the images [Sobral et al. 1980] which will not be discussed here. For the zonal velocity calculation it was first necessary to linearize all these images. This was done using a digital technique developed by Garcia et al. . Zonal profiles of the airglow intensity were then prepared from the linearized images. Each zonal profile had an extension of 1024 km, at an assumed reference altitude of 250 km, which was presented in 512 pixels, rendering a spatial resolution of about 2 km per pixel. The resolution becomes somewhat larger than 2 km at locations further away from vertical because of minor distortions introduced by the linearization process. The zonal velocity presented is therefore limited to a range of 1024 km centered over the zenith of the station. From the linearized zonal profiles the position and time of airglow intensity minima were obtained. The zonal velocities were obtained by dividing the distance between the minima in two consecutive linearized profiles by the corresponding time offset. The deduced velocities refer to 250 km, the assumed height for the airglow emitting layer. The local times used in present calculations refer to the longitudes of the respective stations. The local time difference between BV and CG is ∼20 min. The sampling period of the images is approximately 7 min. The present airglow data did not suffer any interference from moonlight.
2.3. Zonal Drift Velocity From GPS Scintillation by the Spaced-Receiver Technique
 The data from two closely spaced single-frequency (at the L1 = 1.575 GHz) GPS scintillation monitors (SCINTMON) operated at CG were used to calculate the zonal drift velocities of the ionospheric plasma irregularity patches assuming the GPS signal pierce points to be near the F region peak [Kil et al., 2000; de Paula et al., 2002; Ledvina et al., 2004]. The distance between the two GPS receivers was 100 m in the magnetic east-west direction. The GPS data chosen were limited to zenith angle smaller than 40° [see Muella et al., 2008] in order to minimize smearing of the scintillation source position. The spaced receivers are designed to detect the amplitude scintillation caused by the passages of the GPS signal diffraction patterns over them [Yeh and Liu, 1982]. The electron density irregularities causing these scintillations are assumed to occupy a relatively thin layer on the “walls” of the field-aligned bubbles. More details on the calculation of the drift velocity by cross correlation of the scintillation patterns as observed by spaced receiver technique has been described by Costa et al. , Kil et al. , and de Paula et al. .
3. Theoretical Calculation of the Zonal Velocities
 Zonal velocities of the ambient ionospheric plasma over BV/CG were calculated using the detailed theoretical formulations for ionospheric vertical electric field presented by Haerendel et al.  and Eccles ; the vertical electric field being responsible for the zonal velocity. The former work concerned the development of a general expression for the vertical electric field in a field line and the latter introduced a formalism to simplify the calculation of such electric field. In the description of the methodology of the calculations we have adopted the same notation as those used by these authors.
 The zonal velocity of the F region ionospheric plasma, vs, for orthogonal electric and magnetic fields is given by [see Haerendel et al., 1992]
where B is the intensity of the total magnetic field and Eq is the vertical electric field at latitude λ. In order to calculate the zonal velocity over BV/CG, we first calculate Eq. Using the dipole approximation for the geomagnetic field, Eq at any given geomagnetic (dipole) latitude λ is given by [Haerendel et al., 1992]
where ζ = sin λ and EL is the vertical electric field at the geomagnetic equator and positive upward. During nighttime when the ionospheric vertical currents become negligible, Eq and EL become field line mappings of each other [Haerendel et al., 1992]. A simplified expression for EL [Eccles, 1998] is given by
Here B is the total magnetic field intensity at the apex of the field line. ΣP and ΣH are the field line integrated Pedersen and Hall conductivities, respectively, U is the average zonal wind velocity weighted by Pedersen conductivity along the field line, ULH is the average vertical-meridional neutral wind velocity perpendicular to B weighted by Hall conductivity, and VL is the plasma vertical drift velocity in the equatorial plane.
 Substituting equation (2a) in equation (1) we get the expression to calculate the zonal velocity vs at latitude, λ, given by
respectively, where us denotes zonal wind, and uq denotes vertical-meridional wind velocity in the magnetic meridional plane perpendicular to B. The subscript i denotes ith segment, δζi being the increment in ζi, and L the field line apex in terms of earth radius, RE. The local Pedersen (σP) and Hall (σH) conductivities are given by
The conductance is in units of Siemens, S (former Mhos), ki and ke are the ion and electron mobilities given by ωi/νin and ωe/νen, respectively. Ne is the electron density. ΣP and ΣH are given by
The field line integrations in the present calculations were done in the following way: The part of the field line situated above h = 100 km was divided into 52 segments with equal geomagnetic latitude ranges. The E region conductivity was neglected. The dipole field line representation is given by r = L cos2λ, where r is the apex height in Earth radius RE units and L is the invariant latitude. The magnetic latitude λ of CG at 250 km is 11.64° S according to the International Geomagnetic Reference Field (IGRF) model. This model provides the northern conjugate point of CG, also at 250 km, as located over BV. The inputs of zonal wind velocities (u), electron densities and temperatures (Ne, Te), neutrals densities (NO2, NN2, NO, NN) and temperature (T), collision frequencies (νen,νin), gyrofrequencies (ωe and ωi), conductivities (σP and σH), total geomagnetic field intensity (B) and inclination (I), at each of the 52 midpoints were considered in the 52 partial calculations. A data bank of all those parameters was generated through the IDL software that was automatically accessed during the calculations.
 The IGRF model magnetic field intensities B was used to calculate the ion gyrofrequency, ωi = qB/mi [Hz] and electron gyrofrequency ωe = qB/me [Hz] at each segment midpoint. The MSIS model (Mass-Spectrometer-Incoherent-Scatter version E-90) was used to obtain the densities of O2, N2, O, and N [cm−3]. The electron density [Ne, cm−3] and electron temperatures Te at each midpoint were obtained from the empirical ionospheric model IRI (International Reference Ionosphere version 2001). The empirical model HWM (Horizontal Wind Model version 1993) was used for the neutral winds. The following expressions [Kelley, 1989] were used for the ion-to-neutral and electron-to-neutral collision frequencies:
is the mean molecular weight in atomic units, Nn is the neutral number density [cm−3], and Mi is either molecular or atomic mass in atomic units and Ni are number densities of species i (e, O2, N2, O, and N [cm−3]). The vertical velocities VL at the apex were obtained from the Scherliess and Fejer  empirical model.
 Extensive previous studies have shown that under quiet conditions the zonal plasma bubble velocities travel eastward as observed by airglow, and VHF and GPS scintillation, techniques [Sobral et al., 1985, 1999; Sobral and Abdu, 1990, 1991; de Paula et al., 2002; Pimenta et al. 2003; Terra et al., 2004; Abalde et al., 2004; Arruda et al., 2006]. During magnetically disturbed conditions the bubble zonal velocity may slow down and/or reverse westward which is attributed to the disturbance winds and to the associated disturbance dynamo electric field as explained by Abdu et al.  [see also Richmond et al., 2003]. The velocities studied here are found to be all eastward. As regards the seasonal occurrence of the bubbles Sobral et al.  showed, on the basis of 22 years of OI 630 nm airglow data, that they occur over the Brazilian region mostly from October to March with maximum occurrence in the November–January time frame [see also Abdu et al., 1998]. The observations presented here from 2 October to 1 November 2002 represent the beginning of the season of maximum occurrence probability. The frequency of occurrence is also dependent on solar flux and increases with solar activity as will be shown later.
 Although extensive studies of zonal velocities have been conducted from low latitudes little is known about the latitudinal variation in the zonal drift velocity which is important for a better understanding of the thermospheric zonal wind and the conductivity distributions. In this context it is useful to take into consideration that during nighttime the ionospheric vertical currents Jq become negligible, and the field line mapped vertical electric field drives the zonal plasma motion [Haerendel et al., 1992; Eccles, 1998].
 Another important aspect of the zonal velocities of the ionospheric plasma concerns the relationship between the thermospheric zonal wind velocities and the plasma zonal drift velocities. The general expression for the zonal plasma drift velocity is given by equation (3). During nighttime the contribution from VL and ULH to the vertical electric field (EL) of equation (3), however, may be neglected according to Eccles , and in this case the plasma zonal velocity at the apex location reduces to
That is, the plasma zonal velocity is equal to the Pedersen conductivity weighted zonal wind. [see also Martinis et al., 2003]. Consequently the zonal plasma velocity tends to be close to that of the neutral wind at altitudes where Pedersen conductivity is higher and with practically null ion drag. In the present theoretical calculations, for completeness, we have included both VL and ULH, however. With the above comments we will now discuss the important features of the zonal drift velocities inferred from optical and GPS techniques, including a comparison with the theoretical calculations and a possible dependence of the observed velocities on solar flux and magnetic activity. In the remaining text the zonal velocities obtained through airglow measurements, GPS techniques and theory will be referred to as “airglow velocities,” “GPS velocities,” and “theoretical velocities,” respectively.
4.1. General Characteristics of the Airglow Bubble Zonal Velocity
 It is pertinent to make a comment here on the method of calculation of the plasma bubble zonal velocity using the airglow depletion images: the OI 630 nm airglow is emitted by neutral oxygen atoms, O(1D). The lifetime of the O(1D) to O(3P) transition that leads to a 630 nm emission is 110 s and therefore the emission lags behind the dissociative recombination reaction (O2+ + e → O + O + hν) by 110 s. But the emission intensity from the depletions is determined by the plasma (that is, the O+ ion) density in them that is driven by EL × B zonal drifts. The velocity that we calculate, therefore, corresponds to the displacement rate of the depletions within the successive images taken at 7-min cadence, each lagging by 110 s with respect to the plasma density depletions. This time lag has no consequence on the deduced zonal drift pattern if the velocity varies smoothly within the 7-min observational cadence, which we assume to be the case under quiet conditions. Thus the airglow emission minima (the center of the depletion) can be considered as a reference for calculating the zonal drift velocity of the depletions. The practically identical zonal velocities over the two conjugate stations apparent in Figure 3 reveal the rigid alignment of the bubbles which is expected because of the high electrical conductivity of the geomagnetic field lines in the low-latitude region [Farley, 1960]. Otsuka et al.  reported analogous finding in the Asian sector. Previous airglow studies carried out at a single station also showed evidence of field alignment of the bubbles [Weber et al., 1978; Sobral et al., 1980; Tsunoda, 1980]. Such alignment is subjected to the integrated effects of the zonal winds that blow across the field line/flux tube [Haerendel et al., 1992].
 In Figure 3 we further note that the zonal velocities presented distinct time variations on different nights. For example, in the top two plots (2 and 4 October) the zonal velocity showed a brief initial increase which was followed by a decrease toward later hours of the night. In the third plot from the top (8 October), the zonal velocity presented an increase from the beginning of the night up to a maximum around 2220 and rapidly decreases from then on. The plot for 1 November shows that the zonal velocity remains nearly steady around ∼75 m s−1 during most of the premidnight period. The possible causes of these different time variation patterns may be the variabilities in conductivities and wind velocities along the field line, on a day-to-day time scale, as mentioned before.
Figures 4a and 4b show the bubble zonal velocities over one of the conjugate sites plotted against those over the other site on individual days and collectively for all the days, respectively. Statistical parameter, such as, the standard deviation SD, cross correlation CC coefficients, ratio of the mean velocities /, etc. are indicated in Figures 4a and 4b. We note in Figures 4a and 4b that the zonal drifts at the conjugate points follow each other extremely well on individual days as also when looked collectively. The CC is a measure of how closely the two variables vary jointly and the ratio / is a measure of the difference in their average magnitudes. Overall, a striking match between the conjugate point velocities is revealed by the SD/CC of 3.8 m s−1 and 0.99, respectively, in Figure 4b. The CC and the SD were calculated according the expressions indicated in Figure 4b. The ratio of the average velocities over CG and BV is equal to only 67.8 m s−1/66.6 m s−1 = 1.018 or ∼2% of difference in the velocities. These numbers clearly indicate a very high overall match of the airglow velocities over BV and CG both in terms of SD and CC.
4.2. Comparison Between the Airglow and GPS Zonal Velocities
Figure 5 shows a comparison of the airglow velocities (open squares) over CG with the GPS velocities (filled circles) as a function of local time. We recall that the GPS velocities correspond to approximately to the peak height hm of the F layer [see Kil et al., 2000] where as the airglow velocities refers to the airglow emission layer which is situated ∼40 km below the F layer peak height. Therefore the GPS velocity being larger than that of the airglow velocity represents the presence of a positive velocity shear in the plasma zonal velocities with height. Note that such altitude gradient (shear) could correspond to a latitudinal gradient in plasma zonal velocities (zonal velocities increasing with increasing geomagnetic latitude). On days 2 and 8 October and 1 November the GPS velocities are greater than the airglow velocities all the time. On 8 October at the start of the observation a large difference in the two velocities can be noted, the GPS velocity being large (∼150 m s−1) and the airglow velocity being very small (∼20 m s−1) and in the time interval up to 2145 the latter velocity increased and the GPS velocities decreased somewhat. The airglow velocity however, remained below the GPS velocity for the rest of the observational period. This represents an acceleration of the zonal velocity at the altitude of the airglow emission layer and deceleration near the F layer peak. This shows that within the height interval sensitive to the two techniques (of the order of 50 km) the zonal velocity/vertical E field regime may vary considerably, with the lower-altitude electric fields over CG increasing with local time while the electric fields at higher altitudes decreasing at the same time. On 10 October the two velocities are very close to each other in the interval of 2200 LT until 2345 LT which means very low or no height gradient of the zonal velocity. The same occurred on 4 October between 2245 and next day 0030. There are no data outside this time range to compare. On 11 October there are hardly any data to be compared. Since these velocities are functions of zonal neutral wind systems acting across different flux tubes, the two velocities are certainly not necessarily equal to each other. Figure 5 clearly shows, however, the high variability of the plasma shear patterns from day to day. It is perhaps important to remark that the shears mentioned above can be related to upward propagating waves, a fact that could possibly be detected by using two FPIs looking at the same spot in the sky to measure nocturnal F region vertical winds/waves.
Figure 6a shows the airglow velocities plotted against GPS velocities for the days indicated. The SD of the fit ranges from 5 m s−1 to 19 m s−1 that are smaller than the magnitude of the corresponding average velocities and . The CCs, however, are not as high and vary from one day to the other, except for 2 and 4 October when they were 0.81. The poor correlations between the airglow and GPS velocities are indicative of the day-to-day variabilities of the velocity shear with height as observed in Figure 5. The ratios / < 1 for all 5 days indicate that the average airglow velocities were smaller than the average GPS velocities for all days. It is interesting to note that the intense storm that occurred on 2 October did not cause any decrease in the CC compared to the quieter days 10 October and 1 November. There was not enough simultaneous data from GPS and airglow to make a plot for 11 October. On the three quieter days, 8 and 10 October and 1 November, the variables x and y were uncorrelated as seen by the CC: 0.36, −0.22 and −0.23, respectively. In view of the general scatter in the data on these three quiet days the considerably higher CC for the two very disturbed days is a surprising result. More such cases need to be studied to establish any possible connection between CC and disturbance conditions. One may speculate here that the substantially higher correlation coefficients and better (smaller) SDs for the disturbed days in Figure 6a might indicate that the vertical electric fields/zonal motions on these disturbed days might be under control of a disturbance dynamo electric fields that is relatively uniform in a larger height and latitude ranges than those of the quiet time electric fields.
 A mass plot of all the data points of Figure 6a is shown in Figure 6b. As to be expected significant scatter is present in this plot, mostly contributed from the quietest days, 8, 10, and 11 October. The ratio / = 75.7/114.6 ≅ 0.66 statistically represent an overall positive altitude gradient of the zonal velocity.
4.3. Observed Velocities Versus Theoretical Velocities
4.3.1. Airglow Velocity Versus Theoretical Velocity
Figures 7a and 7b show local time plots of the airglow velocities at ∼7 min cadence over BV and CG, respectively, and the corresponding theoretical zonal velocities of the ambient plasma calculated using equation (3). According to equation (3) the lower total magnetic field intensity over CG (21,300 nT at h = 250 km, source IGRF) compared with that over BV (26,400 nT at h = 250 km, source IGRF) will result in smaller velocity for BV compared with CG. In fact we see in Figures 7a and 7b that the theoretical velocities over CG are ∼24% (26,400/21,300 = 1.24) larger than those over BV, as calculated using IGRF values. The observed airglow velocities, over CG are just about 2% higher than those over BV, as discussed previously. The figure of 24% may arise because of the limitation of the IGRF to realistically reproduce the total magnetic field strength at h = 250 km over those two stations possibly owing to local factors such as the presence of the geomagnetic anomaly and very abrupt declination variation in this region. It is important to notice that the total magnetic field intensity decreases continuously toward the southern end along the direction BV-CG because of the presence of the South Atlantic geomagnetic anomaly. In general, the airglow velocities are almost always smaller than the calculated values. The statistical parameters for these two velocities will be discussed later. The theoretical and experimental velocities over CG (Figure 7b) matched each other on 2 October between 0130 and 0245 and on 1 November between 0215 and 0315. Approximately the same trend occurred over BV (Figure 7a).
Figures 8a and 8b show theoretical velocities versus 30-min average airglow velocities over BV and CG for six days, 2, 4, 8, 10, and 11 October and 1 November 2002 from Figures 7a and 7b, respectively. The SDs/CCs for BV shown in Figure 8a are 13.6 m s−1/0.84, 7.4 m s−1/0.95, 21.4 m s−1/0.61, 3.3 m s−1/0.99, 3.7 m s−1/0.98 and 11.9 m s−1/0.01, respectively. The ratios / in Figure 8a varied between 1.05 and 2.08. The higher ratios on 2 and 4 October are due to a decrease in the observed average airglow velocity, as the theoretical calculations that do not include disturbance electric fields do not yield a velocity decrease. This decrease in the observed velocity appears to be caused mainly by the presence of a disturbance dynamo electric field and possibly also by disturbance zonal winds (as pointed out by Abdu et al. ) that dominate the recovery phase of the storm. The F10.7 index (135.8) being the smallest of the six days discussed here might also contribute to a relative decrease in the observed zonal velocity. Thus a comparison of theoretical and observed velocity ratios between disturbed and quiet days shows evidence of disturbance dynamo effects. In spite of the presence of an intense magnetic storm with a maximum SYM-H excursion of −160 nT on 2 October the CC on that day stood as the second best of the set of six days. On all the days studies here the average theoretical velocity was in general larger than the average airglow velocity yielding ratio / > 1. The SDs/CCs of Figure 8b are in general similar to those of Figure 8a except some differences arising because of the number of data points at the two sites not being exactly equal. Such differences may be caused by interruptions in the airglow images related to minor differences in the cloudiness over the two sites.
Figure 9a shows a scatterplot of the theoretical versus observational velocities for BV using all 46 data points of Figure 8a. Figure 9b shows the corresponding scatterplot for Campo Grande. The corresponding statistical parameters SD and CC of the linear fit are 17.6 m s−1 and 0.79, respectively. The individual SDs of the variables x and y are 27.8 m s−1 and 28.4 m s−1 and 119.7 m s−1/38.1 m s−1, respectively. The GPS velocities match better with theoretical velocity than with airglow velocity. The average theoretical velocity 91.6 m s−1 is 51% larger than the average airglow velocity of 60.8 m s−1.
4.3.2. GPS Velocity Versus Theoretical Velocity
Figure 10a shows the temporal variation of the GPS velocities (open circles with vertical variance bars) over CG and theoretical zonal velocities over BV/CG (solid circles) as calculated by equation (3). Many of the open circles that represent GPS velocities seem to be full circles owing to the overlap of a large number of open circles. The GPS velocities are seen to be closer to the theoretical velocities than are the airglow velocities shown in Figures 7a and 7b. Only during the disturbed days (days 2 and 4 October) the magnitudes of the theoretical velocities of this plot appear to be consistently larger than the GPS velocities, as it was largely the case for the airglow velocities as well. We shall come to this point later. The match between the theoretical velocities and the GPS velocities can be verified by their linear fit parameters SD and CC in Figure 10b, wherein the velocities from GPS are averages in 30 min time bins centered at each whole hour, with horizontal variance bars. A statistical analysis shows that the best fit with least SD are for the days 4 and 2 October and 1 November (with SD = 2.8 m s−1, 6.2 m s−1, 7 m s−1, respectively), followed by the less matching days, 10, 8, and 11 October (with SD = 9.1 m s−1, 15.3 m s−1, 22.3 m s−1, respectively). The best match in terms of the CC follow practically that same order of days, except for 1 November, that is, 4, 2, 10, 8, and 11 October 2002 and 1 November 2002 (with CC = 0.99, 0.95, 0.87, 0.85. 0.79, and 0.54, respectively). The / ratios were 1.39, 1.61, 1.02, 1.20, 1.12, 1.07 for the days 2, 4, 8, 10, and 11 October and 1 November 2002, respectively. The amplitude ratios between theoretical and GPS velocities are found to be significantly lower, for the latter four days compared to the former two days.
Figure 10c shows a plot of all the theoretical velocities versus GPS zonal velocities. The overall SD and CC of the fit are 18.5 m s−1 and 0.83, respectively. A comparison of this result with those of Figure 9a and 9b shows that the theoretical velocities match the GPS velocity better than the airglow velocities in terms of the magnitudes and the CC between the two sets of velocities. Such better matching may be attributed to the large number of data points involved in the GPS technique as also to the fact the GPS and theoretical velocities refer to the same altitude region, that is, around F region peak, different from the airglow velocity that represents ∼40 km below the F region peak. The ratios / of mass plots of Figures 9a, 9b and 10c are 1.51, 1.76, and 1.16, respectively, which confirm the best match between the GPS and the theoretical velocities.
4.4. Magnetic Activity Effects on the Bubble Zonal Velocity
 We have used the AE and SYM-H indices to interpret the disturbance time low-latitude zonal velocities. The AE index is related to the auroral electrojet activity and the SYM-H is equivalent to the Dst index [Iyemori, 1990; Sugiura and Kamei, 1991; Iyemori and Rao, 1996; Wanliss and Showalter, 2006]. The AE index may become very intense and sustained during many hours, even in the absence of storms, referred to as High-Intensity Long-Duration Continuous AE Activity or HILDCAAs [Tsurutani and Gonzalez, 1987; Sobral et al., 2006]. During the occurrence of very intense AE activity, disturbance dynamo electric fields may be set up at low latitudes causing a reduction in the postsunset plasma eastward velocity [Richmond and Lu, 2000; Abdu, 1997; Abdu et al., 2003; Sobral et al., 2006, and references therein]. As shown in Figure 11a an intense storm started on 1 October which resulted in the large decrease of SYM-H/Dst to about −160 nT around 1600 LT on 1 October. The storm main phase continued till ∼1000 LT of 2 October followed by recovery continuing into the night. The disturbance dynamo expected to be active in the recovery phase seems to be responsible for the reduced zonal velocity on the night of 2 October as seen in the airglow as well as the GPS velocities in Figures 3 and 10a. Thus the geomagnetic storms that occurred on 1 and 2 October contributed to a perceivable decrease in the average velocities over BV and CG. On 3 October another less intense storm started shortly after 1200 LT with SYM-H index gradually progressing more negative up to a minimum of ∼120 nT on the next day at ∼1100 LT. Sustained AE activity is observed during this storm event. The Dst recovery continued into the night of 4 October. There is an indication in Figure 3 that the bubble zonal velocity suffered some degree of decrease after 2200 LT of this night. On 7 October a weak storm started around 0500 LT and the SYM-H index stabilizes around 70 nT until 1200 LT next day. The remaining days of this plot, 10 and 11 October and 1 November, can be considered to be magnetically quiet. The behavior of the SYM-H and AE indexes during these last three days appear to characterize HILDCAA events.
 In Figure 11b, we present a plot of /, that is, the average theoretical/GPS velocity ratio taken from Figure 10b versus the Kp index, which shows approximately a linear relationship. The data point for 8 October is rather outside the trend of the points for the other days in this plot. We have no immediate explanation for that singularity but it may stem from different global distribution of ionospheric conductivities which are responsible for the DD efficiency.
 During disturbed days the disturbance dynamo upward electric fields in the postsunset low-latitude F region may cause considerable decreases in the magnitudes of the postsunset ionospheric zonal drift velocities [Blanc and Richmond, 1980; Abdu et al., 2003; Huang and Chen, 2008]. The velocities calculated by Haerendel-Eccles methodology and the IRI, MSIS and HWM models do not include disturbance dynamo electric field effects. Therefore the general increase of the ratio / with Kp is indicative of the observed zonal velocities decrease with increasing magnetic activity as to be expected from the action of disturbance dynamo electric field that could dominate in later phases of a storm event [see also Richmond et al., 2003]. The theoretical estimate of the velocity that does not fully include the DD effect is in fact an overestimation of the velocity for the quiet conditions. We may note further that the day by day match between GPS velocities and theoretical velocities of Figure 10b is better than that between the airglow velocities and theoretical velocities of Figures 8a and 8b. This may have to do with the day-to-day variation in the height gradient in the zonal velocity depending on a corresponding variability in the integrated conductivity parameter.
4.5. Solar Flux Effects on the Bubble Zonal Velocity
Figure 12 shows the airglow velocities over BV and CG as a function of the F10.7 cm solar radio flux in units of 10−22 W m−2 s−1. The vertical bars are standard deviations. We note here a tendency of the zonal velocities to increase almost linearly with increasing F10.7 solar flux units. The and data points of Figure 12 are the same as those of Figure 4a. Notice that in Figure 12 except for the night of 10 October, the average velocity over BV (full circle) tends to be slightly smaller than that of CG (open circles) which is due to the lower magnetic field intensity over CG compared with BV, as discussed before. Similar tendency (of slightly larger zonal drift over CG) was more apparent in the plasma irregularity zonal drift obtained from VHF and GPS techniques during COPEX as pointed out by Muella et al. .
 The velocities for 11 October in Figure 4a appear too small and to deviate from the linear trends of the other nights because during this night the data were available only within the limited local time interval of 2315 to 0130 when the zonal velocities have smaller magnitudes (see also Figure 3). For higher F10.7 values the bubble vertical electric field tends to be larger owing to a larger thermospheric zonal wind that drives bubble zonal motion. The larger intensity of the zonal wind must be driven by an enhanced solar thermal tide due to the larger solar flux. It may be pointed out that the smaller airglow velocities for 2 and 4 October might have contributions to their slower motion both from disturbed dynamo effects as well as from lower F10.7 solar flux.
 The data analyzed in this paper were obtained from the COPEX campaign which represents the first conjugate point ionospheric observations conducted in Brazil. We have analyzed zonal velocities of ionospheric plasma bubbles over the two low-latitude conjugate stations as inferred from ionospheric bubbles signatures in OI 630 nm airglow images. The plasma bubble zonal velocities were also obtained using a GPS spaced-receiver technique over CG. The observed velocities were compared with theoretical calculations of the ambient plasma zonal velocities. The main conclusions of this study may be summarized as follows:
 1. The airglow depletion representing the medium to large-scale plasma bubble structures are highly symmetric around the magnetic equator, during their developing and developed phases.
 2. The airglow velocities over the two conjugate stations turned out to be strikingly symmetric with respect to the geomagnetic equator, as seen by their overall standard deviation (SD) of only 3.8 m s−1 and high correlation coefficient (CC) of 0.99, and differences in their average velocities of just 1.8%, the velocity being larger over Campo Grande.
 3. The tendency for larger velocity over CG is more pronounced for velocities measured by VHF and GPS techniques (as quoted from a study of a different COPEX data set). Considering a dipole configuration for the magnetic field, the difference between the plasma zonal velocities over the two conjugate stations, which is consistent with the formulation by Haerendel et al. , is attributed to the local total magnetic field intensity being weaker over Campo Grande located close to the South Atlantic anomaly.
 4. These practically coincident conjugate airglow velocities show near-perfect alignment of the plasma depletions/bubbles along the magnetic flux tubes and that the neutral wind dynamo operates along the entire flux tube.
 5. Both GPS and airglow velocities were eastward for the six nights analyzed. But the airglow velocities were found to be ∼ 66% smaller that the GPS velocities, which is attributed to the difference of altitudes sensitive to the two techniques, the airglow layer being located ∼ 40 km below the F region peak (hmax) sensitive to the GPS scintillation altitude. The larger GPS velocities occurring at higher altitudes compared with those of the airglow occurring at lower altitudes clearly confirm the existence of a shear in the zonal (eastward) plasma flow velocity at these heights.
 6. Our results further show that on some nights GPS velocity decreased while the airglow velocity increased in rather short time interval thereby providing evidence of rapid variation in the degree of shear in the eastward plasma flow. The results also show clear evidence of remarkable day-to-day variation on the plasma shear with height.
 7. Calculations of the zonal velocities based on the formalisms developed by Haerendel et al.  and Eccles  with the use of the IRI, HWM and IGRF models, were found to match very well the experimental results from GPS that refers to height near the F layer peak but less so with the airglow velocities that corresponds to a lower height region.
 8. The theoretical velocities are on an average ∼16% higher than the GPS velocities and 51% –76% higher than the airglow velocities.
 9. The good match (in particular in terms of correlation coefficient (CC)) between the calculated ambient plasma and the observed plasma bubbles velocities appears to confirms that these velocities are coincident after cessation of the ionospheric vertical currents in the evening, as suggested by Eccles  and Martinis et al. .
 10. The model implementation introduced in this work functioned very well under magnetically quieter conditions but during disturbed periods the observed velocities are in general smaller than the modeled values which do not include storm time disturbed dynamo effects in the vertical electric field.
 11. There is a tendency for the zonal velocity to decrease with increase in magnetic activity (represented by Kp) and to increase with increase in the solar flux (represented by the F10.7).
 A speculative hypothesis can be advanced for the observed higher correlations coefficients for the theoretical velocities against the airglow and GPS velocities under disturbed conditions (as in the examples of Figures 6a and 10b) as due to the disturbance dynamo electric fields being relatively more uniform in a wider latitude/height extension than is the case for the quiet time F layer dynamo electric field. In the present study, however, we could not carry out a detailed investigation of such a hypothesis because of experimental limitations. Further studies on the related aspects of the zonal plasma drift over Brazilian low-latitude region are in progress.
 The authors are thankful to the reviewers. The authors are grateful for the efforts of all North American and Brazilian participants who made possible the carrying out of the Conjugate Point Equatorial Experiment (COPEX). AFRL participation was through the support of AFOSR task 2311AS. We thank the Universidade para o Desenvolvimento do Estado e da Região do Pantanal – UNIDERP, the Centro Técnico Aeroespacial – CTA, and Base Aérea de Campo Grande for their logistical support in carrying out the experiments concerned here. This work was supported by the following grants: Fundação de Amparo a Pesquisa do Estado de São Paulo - FAPESP grant 1999/00437-0 and Conselho Nacional de Desenvolvimento Cientifico e Tecnológico-CNPq grants 520185/95-1, 500003/91-2, 521980/94-1, and 305028/2006-5.
 Amitava Bhattacharjee thanks R. Sridharan and another reviewer for their assistance in evaluating this paper.