We present simultaneous observations of daytime medium-scale traveling ionospheric disturbances (MSTIDs) over North America, using measurements of the Total Electron Content (TEC) by the US GPS network and ionospheric plasma data from the DEMETER micro-satellite. Several events show latitudinal variations of the plasma parameters at satellite altitude corresponding to the MSTID structures revealed on 2D TEC maps. In a case study with a very well defined MSTID, quasi periodic variations of the plasma density and of the ion velocity parallel to the Earth's magnetic field are observed along the satellite orbit that match the signature of the MSTID on the TEC maps. We believe it is the first simultaneous observation of parallel plasma motion in the topside ionosphere and propagating MSTID structures in the F-region. An initial analysis of this event is performed in the light of results from simple model of ionospheric disturbances associated with Atmospheric Gravity Waves (AGWs).
 Medium-scale traveling ionospheric disturbances (MSTIDs) are quasi-periodic ionospheric plasma perturbations with typical horizontal wavelengths of several hundreds km and periods between 15 and 60 minutes. Daytime MSTIDs are produced by atmospheric gravity waves (AGWs) that are most often generated at high latitude and propagate towards the equator (see reviews by Hunsucker  and Hocke and Schlegel ). Following the pioneering work of Hines , many theoretical, observational and model studies have been conducted over the last 30 years to understand the complex chain of physical processes that take place during the development and propagation of MSTIDs.
 Large Scale TIDs (LSTIDs) associated with strong magnetic activity move at large speeds of several hundreds m/s with long wavelengths (>1000 km) and periodicity (0.5 ∼ 3.0 hours). The intensified auroral electrojets dissipate large amounts of energy in the lower thermosphere, giving rise to large scale AGWs that propagate away from the auroral zone. The AGW structure is then imparted to the ionospheric plasma, generating LSTIDs.
 Since the mid 1990s, high resolution two-dimensional Total Electron Content (TEC) maps have been made available over Japan and North America, where dense GPS networks are installed and provide continuous data [e.g., Tsugawa et al., 2007, and references therein]. These maps constitute a unique tool to reveal the detailed features of the spatial structure and evolution of MSTIDs.
 There are very few in-situ satellite observations of MSTIDs in the ionosphere. Using data from Dynamics Explorer 2 satellite, Johnson et al.  reported quasi-sinusoidal patterns of vertical velocity of both ions and neutrals over the polar caps. Shiokawa et al.  discussed a simultaneous ground-satellite observation of a nighttime MSTID event using DMSP F15 satellite data and airglow images from ground-based observation. From the DEMETER initial satellite data, quasi-periodic variations of plasma parameters were also observed at sub-auroral latitudes. It was suggested that these variations were associated with AGW/TID [Berthelier et al., 2006]. The present paper deals with the first simultaneous observations of daytime MSTID using GPS-TEC and DEMETER plasma parameters over North America.
 In the first section the GPS-TEC and DEMETER plasma data are briefly described, then, in the second section we present a few events for which a comparison between the two sets of data was possible. In conclusion, we briefly discuss the results obtained for one of these events in the frame of a simple model of the propagation of the MSTIDs in the F-region.
2. Description of Measurements
 The TEC data obtained from the GPS receiver network in North America have been selected for this study because they offer a dense coverage over a wide area, allowing at least one comparison with satellite data per day. As of January 2008, more than 1700 receivers are available, each of which calculates and records the slant TEC values every 30 seconds from 5 ∼ 8 visible GPS satellites with elevation angle larger than 30 degrees. In the TEC map construction, we have employed the same method as that of Tsugawa et al.  as follows. Integer cycle ambiguities and instrumental biases contained in each receiver-satellite data set are removed by detrending the data with a 60-minute running average, which is longer than a typical MSTID periodicity (10 ∼ 60 minutes) [Hunsucker, 1982]. The vertical TEC value, or simply the TEC value, is then obtained by multiplying the slant TEC by a factor defined by the ratio of the “ionospheric thickness” taken equal to 200 km to the line-of-sight (LOS) length. The position assigned to every TEC measurement is defined by the point on the LOS at an altitude of 300 km. These simplifying assumptions are usually made in TEC studies since the F-region of the ionosphere, which provides the main contribution to the TEC, peaks at an average altitude of 300 km and has a typical thickness of about 200 km (for more details on TEC determination see, e.g., Mannucci et al. ). The resulting values are then binned and averaged in pixels of 0.5° × 0.5° resolution in latitude and longitude. The map of the detrended TEC, ΔTEC, is then constructed with a spatial resolution of 1.5° × 1.5° by smoothing over 3 × 3 pixels and with a temporal resolution of 10 minutes, by averaging over 20 30-sec time frames. Examples of ΔTEC maps are shown in Figure 1 (top).
 DEMETER is the first micro-satellite in the CNES MYRIAD program, currently orbiting on a nearly circular sun-synchronous orbit at 650 km altitude with a descending node at ∼10:30 LT on the day-time part of the orbits. The IAP (for Instrument Analyseur Plasma) Plasma Analyzer makes use of two analyzers: APR (for Analyseur à Potentiel Retardateur) performs the energy analysis of the rammed ionospheric ions and ADV (for Analyseur de Direction de Vitesse) determines the direction of arrival of the rammed ions [Berthelier et al., 2006]. We use the ion velocity data from APR and ADV to compute the component of the ion velocity parallel to the Earth's magnetic field (referred to hereafter as the parallel ion velocity) which is counted positive for an upward ion motion. Electron density data from the ISL (for Instrument Sonde Langmuir) Langmuir probe [Lebreton et al., 2006] have been used in the case study discussed in the last section. As explained in more detail below, satellite data are similarly detrended and averaged to filter out quasi DC shifts and fast fluctuations due to the small scale ionospheric irregularities and/or instrument noise.
Figure 1 (top) shows three GPS ΔTEC maps obtained during a time interval of 40 minutes, from 17:10:00 UT to 17:50:00 UT with a time-step of 20 minutes. This interval of time encompasses the DEMETER pass over the central region of the USA on January 22, 2008 in the day-time part of orbit 19011 (referred to hereafter as sub-orbit 19011.0). On the map in Figure 1b, the variations of the parallel ion velocity are displayed colour coded along the orbit track. From the bands of enhanced ΔTEC values seen on the maps, the distances between bands and their temporal evolution, an east-west elongated MSTID structure can be recognized. It extends between the western and the central regions of the USA (Long: −120° ∼ −90° and Lat: 36°N ∼ 44°N) with a wavelength of about 400 km and a southward (equatorward) propagation velocity of about 100 m/sec.
Figure 1b at 17:30:00(UT) corresponds to the average time when the satellite flies over the region of interest. The satellite crosses the US northern boarder at 17:28:00 UT and leaves the region covered by the GPS network at ∼17:36:00 UT, corresponding to a latitude interval from ∼48°N to ∼28°N. For the purpose of comparison with ΔTEC observations, the positions of measurements shown in Figure 1b are those of the satellite northern conjugate point at 300 km. This allows us to compare ΔTEC and in-situ plasma measurements on the same magnetic field line. The parallel ion velocity data are actually subject to small scale fluctuations. To reduce the fluctuation level, the data are smoothed over a length equivalent to 2 degrees of latitude and detrended with 10-degree running averages in latitude which provides the detrended values ΔV∥ of the parallel ion velocity. The overlapped images of ΔV∥, and ΔTEC show an excellent correspondence, with a phase difference between the latitudinal profiles of the two parameters along the satellite orbit. Both ΔV∥, and ΔTEC display 3 peaks at 46°N, 39°N and 34°N and at 47°N, 42°N and 36.5°N, respectively.
 In Figure 2, several other cases of simultaneous observations of MSTID by the GPS network and DEMETER are presented. In Figures 2a–2d, the ΔTEC map at the moment indicated and the ΔV∥ variation along the orbit are shown as in Figure 1. All cases show a good correlation of ΔV∥ in the top-side ionosphere with the MSTID structures in the denser ionosphere close to the F-region peak.
 To allow for an easier quantitative comparison, curves of ΔTEC and ΔV∥ values have been plotted in Figure 1 (bottom) below the colour maps. In these plots the ΔTEC values have been calculated slightly differently by taking, for a given time and coordinates of the satellite, the 5-minute running average of ΔTEC values. Time and coordinates indicated in the horizontal axis are those from the satellite's orbit and the ΔTEC value at each point corresponds to the value at the satellite northern conjugate point at 300 km altitude. The ΔV∥ is shown by a solid line and ΔTEC values by a dashed line. Since we are interested in a relative correlation of the parameters, a spatial detrending with 10-degree running average in latitude is applied to ΔTEC as it is applied on ΔV∥.
 The satellite velocity (7.5 km/sec) is much faster than the MSTID propagation speed (100 m/sec). Therefore the satellite observations provide, at the time of the satellite pass, a nearly instantaneous snapshot of the ionosphere over the whole area where the MSTID structure is revealed by the ΔTEC geographical maps.
 When the angle between the satellite orbital velocity and the MSTID propagation direction normal to the wavefront is large, the apparent wavelength of the MSTID on satellite data can be larger than its actual value. Since the satellite orbital velocity in day-time sub-orbit is inclined ∼8° towards the west, the wavelength of an MSTID with a wave front inclined from south-west to north-east, thus propagating south-eastward would appear larger and this effect has to be corrected. Among the various cases that have been studied, the ΔV∥ variations range from 5 to 50 m/sec per 0.1 TECU of ΔTEC. From the comparison of several MSTID events (not shown here), we have noted that the amplitude of ΔV∥ becomes smaller, when the MSTID propagates at a larger angle with respect to the geomagnetic plane. In such cases, the AGW component of the neutral gas velocity along the Earth's magnetic field is smaller. Since the component of the neutral velocity along the Earth's magnetic field is transferred to the ions through neutral-ion collisions, the resulting parallel ion velocity is likewise smaller in accordance with observations.
 ΔTEC and ΔV∥ from Orbit 19011.0 are shown in Figure 1 (bottom) together with the satellite measurement of the detrended electron density variations ΔNe (dotted line). To get the ΔNe variations, the electron density data are processed similarly as the parallel ion velocity data with a 10-degree detrending and 2-degree smoothing. In most of the events analyzed thus far, the electron density data display quite large fluctuations which often mask the presence of well defined regular variations but sub-orbit 19011.0 is one case in which ΔNe shows clear periodic variations that are in very good agreement with those of ΔTEC and ΔV∥. The MSTID structure is recognized between 30°N and 45°N on all parameters. The local maxima of ΔV∥ at 46°N, 39°N and 34°N latitudes correspond to three maxima of ΔTEC at 47°N, 42°N and 36.5°N latitudes, respectively. Three maxima of ΔNe can also be recognized at 45°N, 38°N and 32°N latitudes, corresponding to the three maxima of two other parameters.
 The observations displayed in Figure 1 (bottom) can be summarized as follows.
 1. ΔTEC and ΔNe. With the above mentioned simplifying assumptions on TEC measurements, ΔTEC and ΔNe represent the electron density variations at two different altitudes, the peak of the F-layer near 300 km and the satellite altitude at 650 km. (a) The maxima in ΔNe are shifted toward the equator with respect to the maxima in ΔTEC. (b) The latitude shift becomes larger at lower latitude: the shift in latitude between two peaks is ∼3° for the ΔTEC maximum at ∼42° while it increases to ∼5° at lower latitude (ΔNe peak at 32°N and ΔTEC peak at 37°N).
 2. ΔV∥ and ΔNe. (a) Peaks in ΔV∥ (namely at 34°N, 39°N and 45.5°N) are always found on a slope of the ΔNe that increases from a trough to a peak (at 32°N, 38°N and 45°N, respectively) toward the equator, implying that the phase difference is in the range from 0° to 180°. (b) The phase difference also becomes larger at lower latitudes.
4. Discussion and Conclusion
 Since the ion density in the upper F-region is much smaller than the neutral density, the ion motion has little feedback on the neutral motion and the ionospheric ions may be considered as a passive tracer of the neutral gas disturbances. The ions are magnetized and, in presence of an AGW, can only move along the magnetic field lines with a velocity equal to the component of the neutral gas velocity parallel to the Earth's magnetic field. Hooke  has analyzed the temporal and spatial variations of the plasma density and ion velocity in TIDs in the altitude range where ion and neutrals are effectively coupled by collisions. For a neutral density nN ∼ 1015m−3, an ion density ni ∼ 1011m−3 and temperature T ∼ 1000 K at the peak of the F-layer near 300 km, the ion-neutral collision frequency is ∼1.0 Hz. This number reduces to 10−4 Hz at the satellite altitude, thus much smaller than the AGW frequency. The collisional/non-collisional boundary is therefore between 300 km and the satellite altitude, at about 450 km altitude as shown by Hooke . Hooke  used a plane wave approximation for the AGW and a simple model based on the continuity equation for the ionospheric plasma. His results show that, in this altitude range, the disturbances of the plasma density ΔNe are organized along phase fronts that reproduce the propagating phase fronts of the neutral gas density and move at the same velocity as the AGW. The variations of ΔV∥ are organized along a similar pattern with moving phase fronts and there is a phase difference, ΔΦ, which was derived by Hooke  as a function of two parameters of the AGW propagation vector, the component along the magnetic field and the vertical damping rate, as well as of the local scale height of the undisturbed plasma. Thus the phase difference, ΔΦ, between the plasma velocity and density disturbances varies with altitude.
 At altitude above ∼450 km in the top-side ionosphere, neutral-ions collisions become negligible and the dynamics of the ionospheric plasma is controlled by its transport along magnetic field lines. To describe the propagation of the plasma disturbances we may refer to the work of Marchand and Berthelier . Using the two-dimensional ionospheric model SAMI2, these authors have studied the ionospheric disturbances following an earthquake and associated with the acoustic pulse which is generated by the main shock and propagate in the neutral atmosphere above the epicenter region. Their results show that, in the top-side ionosphere up to the DEMETER altitude, thus above our collisional boundary, plasma disturbances propagate along the magnetic field at the ion acoustic velocity, of the order of 1 to 2 km/s, without significant variation in the phase difference between the plasma density and parallel velocity disturbances. The distance that the MSTID travels at 100 m/sec, while an ion acoustic wave travel up to the DEMETER's altitude, is a few tens of kilometers and is therefore negligible with respect to the size of the MSTID structure at 300 km altitude. We can thus neglect in practice the propagation delay in the non-collisional altitude when comparing the DEMETER plasma observations and the TEC data.
 In Figure 3, we have given a schematic interpretation of our observations by assuming for simplicity an AGW and its associated MSTID propagating in a geomagnetic plane and considering two separate regions in the upper atmosphere separated by the collisional/non-collisional boundary at ∼450 km. The variations of the neutral gas density have been represented by a grey-scale contour with the dense grey depicting the position of the maximum density. At high altitudes, in the tenuous non collisional upper atmosphere, the AGW become evanescent and, due to the higher atmospheric viscosity [Pitteway and Hines, 1965], the neutral density structures are organized along the vertical direction. At lower altitudes the velocity of the neutral gas, represented by arrows in Figure 3, oscillates in the direction parallel to the phase front of the AGW. In the collisional regime studied by Hooke, ΔV∥ variations coincide with the AGW variations of the neutral gas velocity. The ΔNe variations are organized along a similar phase front structure with a phase difference ΔΦ between the ΔNe and the ΔV∥ maxima as shown in Figure 3 where ΔΦ has been chosen in accordance with the DEMETER observations in sub-orbit 19011.0. According to Hooke , ΔΦ varies with altitude, thus giving rise to ΔΦA ≠ ΔΦB in Figure 3. In the collisional zone, when one follows the equatorward propagating MSTID/AGW, the ΔV∥ maximum lags behind the ΔNe maximum and thus occurs at higher latitude. Above the collisional boundary ΔNe and ΔV∥ map along the magnetic field lines until the altitude of DEMETER with, according to the abovementioned article, negligible change in phase difference, i.e. ΔΦB ≈ ΔΦC. Since DEMETER flies towards the equator, it observes the ΔV∥ maximum prior to the ΔNe maximum i.e. at higher latitude. Due to the respective inclination of the AGW phase front and magnetic field lines, DEMETER encounters the magnetic field line on which the ΔTEC maximum (assumed to correspond to the ΔNe maximum at 300 km) at a higher latitude than both the ΔNe and ΔV∥ maxima. Some cases actually occur when the phase difference between ΔTEC maximum and ΔV∥ maximum, Δθ, is observed to be small and even negligible as in Figure 2b which appears to correspond to an LSTID with a much larger wavelength than the MSTID shown in Figure 1. A simplified interpretative drawing (not shown here) similar to Figure 3 with a longer wavelength AGW shows that the larger ΔΦ results in a shift of the ΔTEC maxima close to ΔV∥ maxima. In this letter, we focus on TIDs with moderate wavelengths that can be readily identified on the US TEC map. Further studies on LSTIDs may even show cases with negative Δθ.
 As shown in Figure 1 the phase differences between the maxima of the three parameters ΔTEC, ΔNe and ΔV∥ increase along the path of the MSTID towards the equator. We believe that this is a phenomenon observed for the first time and may give a fruitful information on the AGW propagation mechanism over a long latitudinal range. A detailed discussion of this effect is beyond the scope of this paper but a simple idea may be suggested based on an increasing inclination of the propagation vector with respect to the horizontal along the path of the AGW. This would induce simultaneously a larger phase difference between the maximum of ΔTEC and ΔNe and between the maxima of ΔNe and ΔV∥. However the reality is certainly much more complex and the detailed understanding of the observations will require a more thorough analysis and modeling effort currently in progress.