Using global positioning system (GPS) data taken from the International GNSS Service (IGS), we investigated total electron content (TEC) perturbations associated with medium-scale traveling ionospheric disturbances (MSTIDs). We analyzed TEC data taken from four or five GPS receivers in each of six regions (Japan, Europe, eastern United States, western United States, Australia, and South America) in 3 years (1998, 2000, and 2001). To derive perturbation components of TEC (I′), we subtracted the 1-hour running average form the time sequence of TEC for each satellite–receiver pair. Standard deviation of I′ within 1 hour, δI, was calculated every hour, and MSTIDs activity were defined as δI/, where is the 1-hour average of absolute vertical TEC. We found that MSTIDs activity during daytime is different from that during nighttime with respect to seasonal, solar activity, longitudinal, and latitudinal dependences. Daytime MSTIDs activity are high in winter in all six regions. On the other hand, seasonal variation of nighttime MSTIDs activity is coupled with its longitudinal variation. In the Japanese and Australian longitudinal sector, nighttime MSTIDs are most active near the June solstice, whereas it is most active near the December solstice in the European longitudinal sector. Nighttime MSTIDs activity at the Japanese and Australian longitudinal sector shows negative correlation with solar activity, whereas solar activity dependence is not seen in daytime MSTIDs activity. These results suggest that mechanisms causing MSTIDs could be different between daytime and nighttime.
 Medium-scale traveling ionospheric disturbances (MSTIDs) are phenomena of the electron density perturbations in ionosphere with horizontal scale sizes of 100–500 km (see, e.g., Hunsucker  for a review). MSTIDs have been observed by using various techniques, such as ionosonde [e.g., Bowman, 1990], HF Doppler sounding [e.g., Waldock and Jones, 1986, 1987], satellite beacons [e.g., Evans et al., 1983; Jacobson et al., 1995], and incoherent scatter radar [e.g., Fukao et al., 1993; Kirchengast et al., 1996]. From these observational results, MSTIDs have been thought to be caused by atmospheric gravity waves [Hines, 1960; Hooke, 1968]. The unremitting interest in investigations of atmospheric acoustic gravity waves (AGW) over more than 4 decades dating back to Hines pioneering work is dictated by the important role played by these waves in the dynamics of the Earth's atmosphere. These research efforts have been addressed in a large number of publications, including a series of thorough reviews [Hunsucker, 1982; Hocke and Schlegel, 1996].
 Using TEC data obtained at Los Alamos (35.9°N, 106.3°W) from a very high frequency (VHF) radio beacon from two geosynchronous satellites, Jacobson et al.  reported that daytime and nighttime MSTIDs are different in the seasonal variation of their occurrence rate and propagation direction. Kelley and Miller  described that these differences would be attributed to the difference in mechanisms causing MSTIDs. Miller et al.  suggested that electrodynamical forces, such as electric fields, could play an important role in generating nighttime MSTIDs.
Shiokawa et al.  have investigated the statistical characteristics of MSTIDs in 630-nm airglow images on the basis of 2-year observations at two different locations in Japan. They have shown that the occurrence rate of MSTIDs is highest in the summer. On the other hand, Garcia et al.  have reported a high occurrence in winter over Arecibo, using an approximately 1-year long airglow imaging observation. These results suggest that seasonal variation of MSTIDs occurrence are different between the Japanese and American longitudinal sectors. Otsuka et al.  and Shiokawa et al.  have shown that MSTIDs structures in the 630-nm airglow images are mirrored between Japan (northern hemisphere) and Australia (southern hemisphere). These results suggest that electric fields in the ionosphere are mapped to the other hemisphere along the geomagnetic field.
 Since the latter half of the 1990s, the Global Positioning System (GPS) has been used to measure the total electron content (TEC) along a ray path between the satellite and the receiver. Saito et al.  have first shown two-dimensional maps of TEC perturbations caused by MSTIDs over Japan using a dense GPS network, which consists of about 1000 GPS receivers. Afraimovich et al.  investigated power spectra of TEC variations caused by MSTIDs using the data taken from 100 to 300 GPS stations for 10 days. Ogawa et al.  have shown fairly good correspondence of MSTIDs structures in the TEC map and 630-nm airglow images of the F region ionosphere. Because these previous observations have been conducted only at a few locations, the global behavior of MSTIDs occurrence has not been clarified.
 In this paper we report a statistical study of MSTIDs using GPS-TEC data obtained for the first time at different longitudes and latitudes in 1998, 2000, and 2001. We found that seasonal variation of MSTIDs activity is different between daytime and nighttime. Longitudinal variation of MSTIDs activity are also reported in this paper. Possible generating mechanisms of MSTIDs are discussed on the basis of these results.
2. Data and Method of Analysis
 An international GNSS service (IGS) provides TEC data obtained with dual-frequency GPS receivers installed all over the world. The number of GPS receivers was 385 as of April 2005. GPS data include carrier-phase and group delays (P-code pseudoranges) of dual-frequency (f1 = 1.57542 and f2 = 1.22760 GHz) GPS signals every 30 s. TEC along a ray path from the GPS satellite to the receiver is precisely obtained from carrier-phase delays. Slant TEC, Is, measured along each GPS satellite-receiver path can be described using the following equation [Mannucci et al., 1999]:
where L1 and L2 are the recorded carrier phases of the signal (converted to distance units), λ1 and λ2 are wavelength of the radiowave, n1 and n2 are integer cycle ambiguities, and br and bs are satellite and receiver instrumental biases terms. Because of the ambiguity in phase measurements, the level of the TEC is unknown. The level is adjusted to that of TEC derived from the pseudoranges for each satellite-receiver pair. TEC obtained by the above procedure still contains biases inherent in satellite and receiver hardware. To get absolute TEC, these biases must be removed. The method reported by Otsuka et al.  is used in the present study to get the absolute TEC.
 To derive perturbation components of TEC (I′), we subtracted the 1-hour running average form the time sequence of TEC for each satellite–receiver pair. To convert the perturbation of the slant TEC to that of the vertical TEC, I′ is multiplied by the slant factor. The slant factor is defined as τ0/τ1, where τ1 is the length of the ray path between 250 and 450 km altitudes and τ0 is the thickness of the ionosphere (200 km) for the zenith path. I′ obtained from this method clearly represents TEC perturbations caused by MSTIDs [Saito et al., 1998]. Figure 1 shows temporal variation of I′ along a ray path between a GPS receiver at (35.4°N, 133.1°E) and GPS satellite PRN09 between 2100 and 2300 LT on 17 May 2001. Wave-like perturbation of I′ with a period of about 40 min is observed. Since this perturbation is similar to that reported by Saito et al. , this I′ variation could be attributed to MSTIDs. Since the typical periods of MSTIDs are between 10 and 60 min [Hunsucker, 1982], TEC variations caused by MSTIDs can be detected by GPS data with a temporal resolution of 30 s [Saito et al., 1998, 2002].
 To study I′ statistically, we calculate the standard deviation of I′ within 1 hour, δI, for each satellite-receiver path every hour. The δI obtained by different satellites whose elevation is more than 40° is averaged. We define δI/ as MSTIDs activity, where is the 1-hour average absolute TEC [Saito et al., 2002]. MSTIDs activities (δI/) obtained from four or five GPS receivers in each of six regions are averaged. Monthly averages of MSTIDs activities are calculated to study seasonal variation of MSTIDs activity.
 In the present study, we selected four or five GPS receivers in each of six regions (Japan, Europe, eastern and western United States, Australia, and South America) to investigate longitudinal and latitudinal variations of MSTIDs. Table 1 and Figure 2 show locations of the GPS receivers used in this study. The numbers of the days when the GPS data are used are also shown in Table 1. It is known that F10.7 indicates the solar activity. The yearly average F10.7 in 1998, 2000, and 2001 is 118, 180, and 181, respectively. This data set allows us to investigate the solar-activity dependence of the MSTIDs.
Table 1. Locations of GPS Receiver and Numbers of Days in Which GPS Data Are Available
Number of Data
Western United States
Santa Susana Mountains
San Gabriel Mountains
Eastern United States
Figure 3 shows local time and seasonal variations of MSTIDs activity in six regions (Japan, Europe, eastern and western United States, Australia, and South America) in 1998 (Figure 3a), 2000 (Figure 3b), and 2001 (Figure 3c). Figure 3 shows that MSTIDs activity varies from ∼0 to ∼3% depending on local time, season, and location. The difference in the activity between daytime and nighttime is especially distinct. MSTIDs activity at night is greater than in daytime. At dawn (0400–0700 LT), MSTIDs activity is enhanced, especially during equinoxes and winter. This enhancement seems to be caused by rapid increase of TEC in the morning due to the plasma production by the solar radiation.
 We have averaged the activity over daytime (0900–1500 LT) and nighttime (2100–0300 LT) for each year to investigate seasonal, solar activity, longitudinal, and latitudinal variations (Figures 3 and 4) . Figure 4 shows the seasonal variation of MSTIDs activity averaged over daytime (0900–1500 LT) for 1998, 2000, and 2001. Daytime MSTIDs activity in Japan and Europe shows an annual variation, with a peak around the December solstice every year. In South America and Australia, MSTIDs activity shows an annual variation with a peak around the June solstice with a large year-to-year variation. On the other hand, MSTIDs activity in eastern United States shows a semiannual variation, with peaks around the June and December solstices. MSTIDs activity in the western United States is quite smaller than in other regions, less than 1% for all seasons, whereas it is slightly enhanced near the December solstice. In all six regions, the solar activity dependence of MSTIDs activity cannot be seen, although year-to-year variations exist.
Figure 5 shows the seasonal variation of MSTIDs activity averaged over the nighttime (2100–0300 LT) for 1998, 2000, and 2001. Nighttime MSTIDs activity for the most part is more pronounced than during the daytime. The seasonal variation of nighttime MSTIDs activity in Japan is similar to that in Australia for all years. Especially, for Japan in 1998 and Australia in 1998 and 2000, they have a primary peak around the June solstice and a secondary peak around the December solstice. These two peaks around the June and December solstices are larger in 1998 than those in 2000 and 2001. Since solar activity in 2001 is higher than in 1998, the MSTIDs activity in Japan and Australia have a negative correlation with the solar activity. The MSTIDs activity in other regions do not show such a clear solar activity dependence.
 MSTIDs activity in Europe shows an annual variation, with a peak around the December solstice in every year. In eastern and western United States, the MSTIDs activity show a primary peak around June solstice, whereas the amplitudes in its seasonal variations are small and vary year to year. The MSTIDs activity in eastern United States and South America show a primary peak around June solstice and a secondary peak around December solstice, except for that in South America for 1998. Month of the primary peak in the MSTIDs activity vary from May to August by the year.
 We have investigated statistically TEC variations observed by GPS receivers in six regions at midlatitudes to reveal local time, seasonal, longitudinal, and latitudinal variations of MSTIDs activity. In this study, MSTIDs activity is defined as δI/, where δI is the standard deviation of TEC within 1 hour, and is the TEC average over 1 hour. Statistical results show a distinct difference between MSTIDs activities during daytime and nighttime with respect to their seasonal, solar activity, longitudinal, and latitudinal dependences.
 Using TEC data at Los Alamos (35.9°N, 106.3°W) obtained from a VHF radio beacon from two geosynchronous satellites, Jacobson et al.  have reported that the seasonal variations of MSTIDs occurrence and propagation direction are different in daytime from nighttime. They have shown that daytime MSTIDs occur mainly during the winter equinox and propagate southeastward; however, nighttime MSTIDs occur mainly during the summer solstice through the autumn equinox and propagate west-northwest. Kelley and Miller  have pointed out that the difference between the propagation directions of daytime and nighttime MSTIDs would be responsible for the difference in mechanisms causing MSTIDs. Neutral particle oscillation due to atmospheric gravity waves in the F region moves ions along geomagnetic field lines through neutral-ion collisions. The ion motion across the magnetic field line is restricted because the ion gyrofrequency is much higher than the ion-neutral collision frequency. This ion motion could cause plasma density perturbations. During nighttime, plasma density in the E region recombines, and thus E region conductivity is significantly reduced. As a result, electric fields generated in the F region are not short-circuited by the E region and can be maintained. The eastward (westward) component of the electric field moves the plasma in the F region upward (downward) by E × B drift. Such electric fields generated in the F region could play an important role in generating nighttime MSTIDs.
 Our results show that daytime MSTIDs activity is high in winter in both the northern and southern hemispheres, although its seasonal variation has a longitudinal dependence. These results are consistent with previous observations at high latitude [Ogawa et al., 1987; Bristow et al., 1996]. Using the U.S. Navy Navigation Satellite System (NNSS) in polar orbits at a 1000-km altitude, Ogawa et al.  have shown that daytime MSTIDs occur most frequently in winter in Antarctica. Bristow et al.  have shown that MSTIDs occurrence is highest in winter using Super Dual Auroral Radar Network (SuperDARN) data. They have suggested that this seasonal variation of the MSTIDs occurrence is attributed to the reflection of gravity waves near the mesopause. Because the altitude gradient of the neutral temperature is steep near the mesopause in summer, some of the gravity waves cannot propagate upward from the middle atmosphere through the region of the temperature gradient. Our results suggest that this theory could also be applicable to midlatitudes.
Afraimovich et al.  identified a specific class of midlatitude MSTIDs, namely traveling wave packets (TWPs), which was quasi-monochromatic oscillations of TEC with a period of around 10–20 min, and reported morphology of TWPs using the global GPS network data obtained at California, the Caribbean basin, and Southeast Asia for 105 days in 1998–2001. Most of TWPs were observed during daytime winter, and amplitude of TEC perturbations due to TWPs is about 4% of the background TEC. These features are consistent with those of the daytime MSTIDs activity reported in this study, suggesting that TWPs could mostly contribute to the MSTIDs activity. However, the MSTIDs activity shown in the present study are high even during nighttime. This difference could be caused by difference of the periods of TEC perturbations between daytime and nighttime. Afraimovich et al.  filtered from the original TEC date over the periods of 2–20 min to derive perturbation component of TEC, whereas we subtracted 1-hour running average from the original TEC data in the present study. Comparing the data processing methods and results in the two papers, we speculate that the daytime MSTIDs would have periods less than about 20 min and nighttime MSTIDs more than about 20 min. On the other hand, TWPs occurrence rate was high in autumn, whereas the MSTIDs activity are low in equinoxes. This difference may be attributed to the background TEC. TEC is higher in equinoxes than in solstice. The high TEC values in autumn could probably cause the low MSTIDs activity, even when TWPs cause TEC perturbations.
 The activity of nighttime MSTIDs in Japan and Australia shows semiannual variations, with two peaks near the summer and winter solstices. The primary maximum appears near the June solstice, and a secondary maximum appears near the December solstice. This result indicates that MSTIDs may occur almost simultaneously in both the northern and southern hemispheres in this longitudinal sector. Otsuka et al.  have simultaneously observed OI (630-nm) airglow perturbations caused by MSTIDs at Sata, Japan, and Darwin, Australia, at around midnight of 9 August 2002, using two all-sky CCD imagers. The observed MSTIDs structures were mirrored in the northern and southern hemispheres connected by geomagnetic field lines. They have explained such a mirrored MSTIDs structure with the following scenario. Polarization electric fields are generated to maintain a divergence free of electric currents in the F region, where Pedersen conductivity is perturbed by the MSTIDs. The electric fields are mapped along geomagnetic field lines and push the F region plasma upward or downward by E × B drifts, causing perturbations of the plasma density in both hemispheres. Statistical results obtained in the present study suggest that nighttime MSTIDs would almost have these mirrored structures. Only in the European longitudinal sector, a primary maximum of nighttime MSTIDs activity is near the December solstice; however, in the other region, the peak of nighttime MSTIDs activity are around June solstice. Our results show that the nighttime MSTIDs activity are high in June and/or December solstice and that it is low in equinoxes. This seasonal variation is consistent with that observed for the spread-F occurrence rate [Bowman, 1992]. Midlatitude spread-F is probably caused by off-vertical reflections of radio waves from tilted isoionic surfaces produced by the passage of MSTIDs [Bowman, 1992]. He suggested that the double peaks in MSTIDs occurrence at the solstices may be explained by the seasonal variation of the neutral density in the upper atmosphere. The neutral density shows semiannual variation, with maxima at equinoxes and minima at solstices. The amplitude of gravity waves becomes larger for a smaller neutral density [e.g., Hines, 1960]. A large amplitude of gravity waves could cause large plasma density perturbations, i.e., MSTIDs activity. On the other hand, the linear growth rate of the Perkins instability is also inversely proportional to the neutral density [Perkins, 1973]. The Perkins instability is a likely source of nighttime MSTIDs, although the linear growth rate is very small. These facts suggest that seasonal variation of the nighttime MSTIDs at Japan and Australia could be caused by that of the neutral density. Furthermore, our observational results show that nighttime MSTIDs activity in Japan and Australia tends to increase with decreasing solar activity. Because the neutral density decreases with decreasing solar activity, this result supports the idea that nighttime MSTIDs activity may be controlled by the neutral density in the thermosphere. However, the MSTIDs activity in other regions do not show such a solar activity dependence.
 Our results show that the nighttime MSTIDs activity have a longitudinal dependence. The nighttime MSTIDs activity at Japan-Australian longitudinal sector have a major peak in June solstice. This seasonal variation is consistent with that of spread-F occurrence reported by Bowman . At the American longitudinal sectors, however, the seasonal variations of the nighttime MSTIDs activity are different from those of spread-F occurrence. The nighttime MSTIDs activity at European longitudinal sector have a major peak around June solstice, whereas spread-F occurrence have a major peak around December solstice. At American longitudinal sector, on the other hand, a major peak of the nighttime MSTIDs activity are around December solstice but that of spread-F occurrence is around June solstice. These results indicate that latitudinal difference in the nighttime MSTIDs activity also would exist. Discrepancy between the MSTIDs activity and spread-F occurrence would be due to the difference between the spatial scale sizes of MSTIDs detected by GPS-TEC and those that cause spread-F.
 At high-latitude of American longitudinal sector, MSTIDs occurrence coincide with the spread-F occurrence [Hunsucker and Hargreaves, 1988]. At the midlatitude, however, the seasonal variations of the nighttime MSTIDs activity are different from those of spread-F occurrence. These results would be responsible for the difference in mechanisms causing spread-F between high-latitude and midlatitude.
Afraimovich et al.  investigated an unusual class of MSTIDs of the nonwave type, isolated ionospheric disturbances (IIDs) that manifest themselves in TEC variations in the form of single aperiodic negative TEC disturbances with a duration of about 10 min (the total electron content spikes, TECS). Amplitude of TEC perturbation due to IIDs is about 0.3 TEC. This is comparable to amplitude of TEC perturbation investigated in this study. IIDs occur mainly in the nighttime during spring and autumn. However, our study shows that the MSTIDs activity in equinox are not pronounced in all regions. Since occurrence rate of IID is small (approximately 1%), IID-induced TEC perturbations does not contribute to the MSTIDs activity defined in this study.
 We have statistically investigated TEC variations observed by GPS receivers in six regions (Japan, Europe, eastern and western United States, Australia, and South America) at midlatitudes in 3 years (1998, 2000, and 2001) to reveal local time, seasonal, longitudinal, and latitudinal variations of MSTIDs activity. The results can be summarized as follows.
 1. The difference between MSTIDs activity during daytime and nighttime can be seen with respect to their seasonal, solar activity, longitudinal, and latitudinal dependences.
 2. Daytime MSTIDs activity are high in winter in all six regions. This result suggests that the daytime MSTIDs are caused by acoustic gravity waves propagating upward from middle atmosphere and that the seasonal variation of daytime MSTIDs occurrence is controlled by the altitude gradient of the neutral temperature near the mesopause. In summer, some gravity waves cannot propagate through the region of the steep temperature gradient near the mesopause.
 3. Seasonal variation of nighttime MSTIDs activity are coupled with its longitudinal variation. At the Japanese and Australian longitudinal sector, nighttime MSTIDs are most active near the June solstice, whereas it is most active near the December solstice at the European longitudinal sector. These longitudinal difference in seasonal variations of ionospheric variations is not consistent with the spread-F occurrence rate obtained from ionosonde measurements. Seasonal variation of nighttime MSTIDs activity in Japan is almost identical to that in Australia. This result indicates the geomagnetic conjugate occurrence of nighttime MSTIDs.
 4. Nighttime MSTIDs activity only in Japan and Australia shows a negative correlation with solar activity, whereas solar activity dependence is not seen in daytime MSTIDs activity.
 These results suggest that the mechanisms causing MSTIDs are different between daytime and nighttime.
 GPS data are provided from the International GNSS Service (IGS). This work is supported by a grant from the Ministry of Education, Culture, Sports, Science and Technology, Japan (Dynamics of the Sun-Earth-Life Interactive System, No. G-4, the 21st Century COE Program). N. Kotake is supported by a grant from the Ministry of Education, Culture, Sports, Science and Technology, Japan (Information Nano-Devices Based on Advanced Plasma Science, the 21st Century COE Program).
 Arthur Richmond thanks E. L. Afraimovich and Robert D. Hunsucker for their assistance in evaluating this paper.