Low latitude ionosphere-thermosphere dynamics studies with inosonde chain in Southeast Asia

. An ionosonde network consisting of a meridional chain and an equatorial pair was established in the Southeast Asian area. Three of four ionosondes are along the magnetic meridian of 100 ◦ E; two are close to the magnetic conjugate points in Northern Thailand and West Sumatra, Indonesia, and the other is near the magnetic equator in the Malay Peninsula, Thailand. The fourth ionosonde is also near the magnetic equator in Vietnam but separated by about 6.3 ◦ towards east from the meridional chain. For a preliminary data analysis, nighttime ionospheric height variations at the three stations of the meridional chain were examined. The results demonstrate that the coordination of the network has a great potential for studying ionosphere/thermosphere dynamics. Through the assistance of model calculations, thermospheric neutral winds were inferred and compared with the HWM93 empirical thermospheric wind model. Higher-order wind variations that are not represented in the empirical model were found. from strong north-south coupling and equally strong thermospheric wind and zonal electric ﬁeld effects. We established an ionosonde network in Southeast Asia that consists of four stations. Three of them form a meridional chain at 100 ◦ E, at the magnetic conjugate points and the magnetic equator. Another station is near the magnetic equator and is about 730 km east of the meridional chain. The study demonstrated that the meridional chain of ionosonde stations is advantageous for studying the ionosphere-thermosphere coupling. By scaling the virtual height from ionograms, nighttime ionospheric height variations were analyzed. The seasonal trend of the inferred transequatorial-mode wind is generally reasonable, show-ing that the wind is directed from the summer to the winter hemispheres. The results were compared with theoretically calculated height variations incorporating an empirical thermospheric wind model, HWM93. The observed meridional wind exhibits a variation with a 5–6 h period, which is not


Introduction
Two major parameters that specify the ionospheric status are the peak electron density (NmF2) and the height of the electron density peak (hmF2). The principal mechanism of the ionospheric peak formation was first presented by Chapman (1931) for an idealized atmosphere, in which atmospheric density decays in an exponential way with altitude. The de-Correspondence to: T. Maruyama (tmaru@nict.go.jp) cay constant of the atmospheric species, or the scale height, determines hmF2 to attain equilibrium between chemical and diffusion processes (Yonezawa, 1966;Rishbeth et al., 1978). The actual ionosphere is not as simple as such a static picture suggests. At midlatitudes, the ionospheric height varies dynamically through the action of an E×B drift due to zonal electric fields and the field aligned motions induced by the drag of thermospheric neutral winds. The peak electron density and the height of the density peak are closely related to each other such that the upwelling of the layer height causes an increase in the electron density because of the reduced recombination rates at higher altitudes (Strobel and McElroy, 1970;Tanaka and Hirao, 1973). Variations in NmF2 at low latitudes are more complex than they are at midlatitudes. At the magnetic equator, the upwelling of the layer caused by the E×B drift due to an eastward electric field results in a decrease in NmF2 because of the field aligned diffusion of the ionization to higher latitudes (Hanson and Moffett, 1966;Anderson, 1973), which is called the fountain effect. The fountain effect results in higher density at low latitudes and forms two density peaks (crests) in the Northern and Southern Hemispheres or equatorial anomaly (Namba and Maeda, 1939;Rishbeth, 2000). Another complexity comes from the transequatorial component of the neutral wind. The transequatorial ionization flux can decrease the electron density in the windward hemisphere and increase it in the leeward hemisphere. The total effect of the wind on the density variations is a compromise between the above fieldaligned interhemispheric flow and the change in the recombination loss rate due to the layer upwelling/downwelling induced by ion drag (Hanson and Moffett, 1966). Sharma and Hewens (1976) show that foF2 at the equatorial anomaly crest is generally higher in the winter hemisphere in solstice seasons, during which transequatorial neutral wind blows from the summer to the winter hemispheres. Similar hemispheric asymmetry can be found in Fig. 9b of Maruyama and Matuura (1984), in which the peak electron density is higher in the Northern (winter) Hemisphere, nevertheless, the peak heights are lower there. Walker and Chan (1989) give model calculations for NmF2 and hmF2 and their Fig. 4 demonstrates the above mentioned complexity of the transequatorial neutral wind effect on the ionosphere.
Compared with the complexity of the peak electron density, the variations in its height are straightforward in response to external forces, i.e., field perpendicular motions of the E×B drift due to zonal electric fields and field aligned motions induced by the drag of thermospheric neutral winds (Rishbeth et al., 1978;Titheridge, 1995;de Medeiros et al., 1997).
One of the important issues of ionospheric height variations is the connection with nighttime plasma instabilities at low latitudes. At the magnetic equator, the lower part of the ionosphere rapidly recombines after sunset, and the resultant steep vertical-density gradient leads to an onset of plasma irregularities through the Rayleigh-Taylor (R-T) instability. The growth rate of the R-T instability is high when the ionospheric layer height is high. A high layer height is attained through upward E×B drift due to enhanced eastward zonal electric field through the F-region dynamo process (Matuura, 1974;Heelis et al., 1974;Farley et al., 1986). A large transequatorial neutral wind, say larger than 100 m/s, also causes an apparent layer uplift, because the plasma at lower altitudes inside the short flux tube is blown away, however, this does not lead to plasma instabilities (Maruyama, 1996). A possible effect of the transequatorial wind is suppression of the R-T instability through increased Pedersen conductivity at the leeside foot that is electrically connected with the equatorial ionosphere by a magnetic field line with a large conductivity (Maruyama and Matuura, 1984;Maruyama, 1988;Mendillo et al., 1992;Saito and Maruyama, 2006). Thus, it is important to separate the zonal electric field and meridional neutral wind effects on the ionospheric height changes. However, in general, this separation is not easy to do with ionosonde observations taken at a limited number of locations.
Several authors have proposed and elaborated methods for obtaining neutral wind characteristics from measured ionospheric peak heights at midlatitudes (Buonsanto, 1986;Miller et al., 1986Miller et al., , 1997Richards, 1991;Titheridge, 1995;Liu et al., 2003). Miller et al. (1986) obtained a numerical factor that linearly relates the horizontal component of the neutral wind along the magnetic meridian and the resulting deviation in the F layer peak height from the balance height. The result is used for determining the neutral wind velocity from measurements of hmF2. This derivation of the neutral winds ignores the contribution of the E×B drift to the layer height changes. However, by using a dataset from the Millstone Hill incoherent scatter radar (magnetic latitude ∼51 • ) for comparison, Miller et al. (1987) show that uncertainties introduced into the derivation of meridional neutral winds from hmF2 by the exclusion of electric fields are within the uncertainty of the density peak and ion drift measurements. Igi et al. (1995) also compare the neutral winds derived from hmF2 observed at a lower midlatitude station, Kokubunji (magnetic latitude ∼30 • ), based on the method proposed by Miller et al. (1986), with those observed by the MU radar at Shigaraki, Japan, and demonstrate that neglecting the E×B drift does not yield a large error.
All the above methods neglect the E×B drift effect on the peak height, and thus, they can not be applied to the equatorial and low latitude ionosphere, where magnetic dip angles are small and the E×B drift due to the zonal electric field contributes as well as the neutral wind to changing the ionospheric height. De Medeiros et al. (1997) proposed an extension of the method used to derive meridional neutral winds at midlatitudes that would be applicable to those at low latitudes. They explicitly included the effect of the E×B vertical drift of the ionosphere for deriving neutral wind from hmF2 at a low latitude station, Cachoeira Paulista (magnetic latitude ∼19 • ). The E×B drift velocity was deduced as dh ′ F/dt from 15-min interval ionograms over an equatorial station, Fortaleza, and it was mapped to the low latitude station by using an electrodynamic model. Thus obtained neutral winds were compared with the HWM-90 empirical thermospheric wind model (Hedin et al., 1991) and Fabri-Perot measurements. The results were in reasonable agreement (de Medeiros et al., 1997). Bittencourt and Sahai (1978) calculated neutral winds by comparing hmF2 observed at either end of a low-latitude field line. Their method does not need the zero-wind reference height information required in the above methods for the midlatitude condition. Thus obtained neutral winds are the transequatorial component averaged over the two locations. The meridional winds at the conjugate points, however, are not necessarily equal. In other words, the wind field would be decomposed into the transequatorial and divergence/convergence modes. The divergence/convergence wind can not be inferred from the height variations at the conjugate points even if zero-wind reference heights are assumed, because the E×B drift largely changes the height in the same direction at the conjugate points. That is, the upward E×B drift and convergence wind effects yield similar height changes at both locations. To overcome this problem so that ionosphere/thermosphere at low-latitudes can be studied, we established a meridional ionosonde network that consists of one equatorial station and two low-latitude conjugate stations along the meridian of 100 • E. A fourth station was installed near the magnetic equator but separated in longitude to detect possible fluctuations that do not migrate with the motion of the sub-solar longitude (Kelley et al., 1981). The network is similar to one constructed on the Brazilian longitude for an observation campaign, COPEX, Abdu et al., 2004), but our network is permanently operated.   Bittencourt and Sahai (1978). Boa Vista, Cachimbo, and Campo Grande constitute the meridional chain of the COPEX campaign Abdu et al., 2004).
The second section presents the Southeast Asia lowlatitude ionospheric network (SEALION). The third section is devoted to model prediction of the ionospheric height variations during nighttime responding to external forcing and the four one to the data analysis. We summarize the results to conclude the paper in the fifth section.

Ionosonde network
The magnetic equator runs mostly over the ocean as shown in Fig. 1, and it is not very easy to find suitable locations to observe the ionosphere along the magnetic meridian. In the Southeast Asian area, the Indochinese Peninsula lies to the north of the magnetic equator and the Indonesian Archipelago to the south. We established an ionospheric observation network in this area (SEALION: Southeast Asia low-latitude ionospheric network). Three ionosondes composing a meridional chain along the 100 • E meridian were installed at Chiang Mai, a city in northern Thailand, at Chumphon, in the middle of the Malay Peninsula, and at Kototabang in West Sumatra, Indonesia (Fig. 2). Chumphon is slightly north of the magnetic equator (magnetic latitude is 3.3 • N). Kototabang is almost at the geographic equator (0.2 • S), and an atmospheric observation facility including the Equatorial Atmosphere Radar (EAR) is in the same observation field. A combined use of the ionosonde and EAR data in the ionospheric irregularity study is described by Fukao et al. (2006). The magnetic latitude of Kototabang is 10.0 • S, and its magnetic conjugate point is at (100.1 • E, 16.6 • N). Chiang Mai was selected as a site close to the magnetic conjugate of Kototabang. All stations are within 1.4 • in longitude, and two magnetic meridians passing over Chiang Mai and Kototabang are separated by 140 km at the F region height. The fourth station was installed in the south- ernmost area of Vietnam, Bac Lieu, near the magnetic equator (magnetic latitude is 1.7 • N). The two meridians passing over Chumphon and Bac Lieu are separated by 740 km at the F region height. The parameters for the stations are summarized in Table 1.
In order to conduct continuous observations, we developed a frequency modulated-continuous wave (FM-CW) ionosonde, which operates from 2 to 30 MHz with low electric power. A commercial power amplifier module delivers a maximum power of 150 W. The prototype of the ionosonde is described by Nozaki and Kikuchi (1987).

Expected height variation at equatorial latitudes
Ionospheric height variations are often discussed in terms of the F2-layer peak height (Rishbeth et al., 1978;Bittencourt and Sahai, 1978;Miller et al., 1986;Richards, 1991). The peak height can be calculated through a real height analysis (e.g., Titheridge, 1988) or formulations using the ionospheric transmission factor M(3000)F2 (Shimazaki, 1955;Berkey and Stonehocker, 1989). However, it was sometimes difficult to scale the whole ionogram trace accurately, especially near the critical frequency. Another parameter representing the ionospheric height is the ionospheric virtual height, h ′ F , which can be directly and easily scaled from an ionogram. This is a good indicator of the bottom height of the ionosphere in the night when solar radiation does not cause new ionization. In this paper, we discuss the ionospheric height in terms of the virtual height. However, we must be careful in using h ′ F as a height parameter; when the layer height is largely depressed, the recombination effect overcomes the dynamical effect on the bottom height, and changes in h ′ F do not necessarily reflect actual layer movements correctly. Although ionospheric height variations strongly reflect external forces such as thermospheric neutral winds and zonal electric fields, these forces are difficult to separate from ionosonde observations alone. We examined the individual effects through model calculations for idealized situations. The model calculations solve the continuity and momentum equations (e.g., Maruyama, 1988) for equinox, medium solar activity (F10.7 solar flux is 100) conditions with a centered dipole magnetic field geometry. As described below, the assumption of forcing is overly simplified and the calculated features do not provide qualitative new information, but are expected theoretically. However, they should be helpful to demonstrate the ionospheric response to external forces for interpreting the observational results.
We calculated the ionospheric bottom height corresponding to the electron density of 7.75×10 4 cm −3 (plasma frequency of 2.5 MHz) at 20:00 LT. The first run contained no external forcing, and the height was determined by chemical and diffusion processes alone. The result is shown in Fig. 3 by dotted lines, and is used as a reference curve in the all panels. The bottom height appears to be uplifted at the magnetic equator, and it gradually reaches a constant height as it leaves the equator. This is owing to the changes in the geometry of the magnetic flux tube. At low latitudes, plasma would be supplied from higher altitudes along the magnetic field line and the recombination losses would be somewhat compensated. However, at the magnetic equator, such replenishment across the magnetic field lines does not exist.
The second run examined the effect of an upward E×B drift, and the result is shown by the thick solid line in Fig. 3a. In this run, an electric field corresponding to 30 m/s at the magnetic equator was imposed at 19:00 LT and persisted for one hour. Through this upward E×B drift, the bottom height is raised by about 102 km at the magnetic equator, which is close to the actual displacement of the magnetic flux tube (108 km). On the other hand, at the latitude of 10 • , the uplift is only about 38 km, because of the downward diffusion imposed by the fountain effect. The third run examined the effect of a transequatorial neutral wind, and the result is shown by the thick solid line in Fig. 3b. In this run, a Ann. Geophys., 25, 1569-1577, 2007 www.ann-geophys.net/25/1569/2007/ uniform northward neutral wind at the speed of 100 m/s was imposed at 19:00 LT and persisted for one hour. The bottom height at the magnetic equator is uplifted by about 10 km by the blow away effect (Maruyama, 1996). At 10 • in the windward hemisphere, the bottom height is uplifted by about 60 km, while in the other hemisphere, the bottom height is depressed by about 43 km. The relatively small (negative) perturbation compared with the windward hemisphere is the recombination effect which is significant in the lowered ionosphere. A noticeable feature is that the bottom heights at low latitudes in the windward hemisphere are higher than at the magnetic equator. Figure 3c shows the effect of a convergence neutral wind. In this run, equatorward winds had amplitudes of 30 m/s at 10 • in both hemispheres and linearly decreased with lower latitude. The bottom height at the magnetic equator is the lowest. The last run (Fig. 3d) used the upward E×B drift model used for the second run and the neutral wind model used for the third run (transequatorial mode wind). The bottom height difference between the Northern and Southern Hemispheres at 10 • is about 100 km, which is nearly the same as for run 3 (60+43 km) with only the neutral wind effect (Fig. 3b). This result indicates that the height difference caused by transequatorial winds is not strongly affected by the amplitude of the upward E×B drift.
In the above, external forces were imposed at 19:00 LT, and the bottom height responses at 20:00 LT are described. However, the response is a function of time. Figure 4 shows how quickly the bottom height responds to the force. In this run, northward transequatorial wind turned southward at 20:00 LT, with the same amplitude (50 m/s). The north/south asymmetry of the height disturbance reversed about 30 min after the wind turned direction, and the time constant of the height response to the wind is around one hour.

Preliminary observational results
The first complete set of observations for the solstitial condition from the meridional chain was available in December 2004. Virtual heights at 2.5 MHz were scaled from the ionogram with a 15-min interval, and the results for nighttime are plotted in Figs. 5a to c for Chiang Mai, Chumphon, and Kototabang, respectively, for the period from 7 December 2004 to 5 January 2005 (December solstice). Data for individual nights are shown by thin lines, and the median values of the period are shown by diamonds. The number of data points was greatly reduced after about 04:00 LT, because foF2 is close to or lower than 2.5 MHz, and the reliability of the data after that is low. The layer heights at the low-latitude stations, Chiang Mai and Kototabang, after 23:00 LT exhibit a large day-to-day variability. The values at the equatorial station, Chumphon, are somewhat smaller than the values at the low-latitudes, which suggests that the variability at low latitudes mostly originated from a neutral wind effect, but not from a zonal electric field effect. On the other hand, the Chumphon data shows a moderate day-to-day variability around 20:00 LT, which originated from the variability of the evening enhancement of the zonal electric field.
As the layer height at Chumphon is mostly controlled by the zonal electric field, the high layer height at Kototabang seen in the median plot, which peaked at 21:15 LT, and the low layer height at Chiang Mai are considered to be effects of northward neutral wind, as demonstrated in Fig. 3b. For a more detailed comparison, we obtained no-wind heights at Chiang Mai and Kototabang, with the aid of the model calculation described in the previous section, but for the tilteddipole geometry and the day number of 356. First, the E×B drift velocity was adjusted to reproduce the height change at Chumphon. Then, the electron density distribution in the magnetic meridional plane was obtained. The difference between the observed median values and the modeled heights at Chiang Mai and Kototabang are shown in Fig. 6a. Before 18:30 LT, h ′ F is affected by the propagation retardation in the lower part of the ionosphere, and the observation data are not compared with the modeled heights here. After sunset, the layer height at Kototabang was higher than the no-wind height and the layer at Chiang Mai was lower than the nowind height. That is, transequatorial northward wind is predominant, as mentioned above. As time elapsed, the height at Kototabang increased farther, but the height at Chiang Mai reached the no-wind height, and after 20:00 LT, the heights at Chiang Mai and Kototabang eventually became higher than the no-wind height. There could be two cases in which the layer height increases at both conjugate points: upward E×B drift (Fig. 3a) and convergence wind (Fig. 3c). However, the E×B drift effect was already included in the model calculations by fitting the height variations at Chumphon. Thus, Fig. 6a  gradually intensified while the transequatorial-mode wind stayed the same. After 20:00 LT, the height at Kototabang decreased, but the height at Chiang Mai stayed at a moderately high level. That is, the transequatorial-mode wind weakened and the weak convergence-mode wind remained. As mentioned already, the response of the layer height to the neutral wind is delayed by half to one hour, and the actual times of the wind direction reversal should be earlier than in the above discussion.
It would be worthwhile to compare the observations with an empirical thermospheric wind model, HWM93 (Hedin et al., 1991), which is widely used to predict thermospheric wind fields. For this purpose, we ran another model calculation which incorporated HWM93 and re-adjusted the E×B drift velocity. (The E×B drift velocity is slightly different from that obtained for the no-wind case, because Chumphon is not exactly at the magnetic equator.) The difference between the observed median value and the modeled height is shown in Fig. 6b. The large deviation of h ′ F at Kototabang at around 20:00 LT is reduced to some extent when the HWM93 wind is assumed, but still there is a large discrepancy between the observations and the model results. The second peak of h ′ F for Kototabang at around 01:00 LT does not change very much even if the HWM93 wind field is taken into account. The most noteworthy feature in Fig. 6b is the periodic 5-h variation in h ′ F for both Kototabang and Chiang Mai. Such a higher order variation is not represented in HWM93 (Hedin et al., 1991); yet the higher order variation is quite significant, especially after midnight at Kototabang, the geographic equator. The neutral wind field is subject to variation with the latitude of the sub-solar point. Figures 7 and 8  (crosses) was largely enhanced compared with the other seasons. This is owing to an intense prereversal enhancement of the zonal electric field, which is known to be large during the equinox (Fejer et al., 1995). In contrast, the height at the low latitude stations, Chiang Mai, and Kototabang, was lower than at the equator, as expected from the small seasonal or transequatorial wind. A notable periodic change of about 6 h is clearly seen in the difference between the heights at Chiang Mai and Kototabang, similar to the one in Fig. 6 for the December solstice. That is, the transequatorial neutral wind periodically reversed northward and southward. At the June solstice, the layer height at Chiang Mai was higher than at Kototabang throughout the night, indicating southward transequatorial winds due to northward migration of the sub-solar point. However, the pattern of the height variation was not quite the reverse of the December solstice, nor were there variations with 5-6 h.

Summary
The equatorial and low latitude ionosphere exhibits unique dynamical features stemming from strong north-south coupling and equally strong thermospheric wind and zonal electric field effects. We established an ionosonde network in Southeast Asia that consists of four stations. Three of them form a meridional chain at 100 • E, at the magnetic conjugate points and the magnetic equator. Another station is near the magnetic equator and is about 730 km east of the meridional chain. The study demonstrated that the meridional chain of ionosonde stations is advantageous for studying the ionosphere-thermosphere coupling. By scaling the virtual height from ionograms, nighttime ionospheric height variations were analyzed. The seasonal trend of the inferred transequatorial-mode wind is generally reasonable, showing that the wind is directed from the summer to the winter hemispheres. The results were compared with theoretically calculated height variations incorporating an empirical thermospheric wind model, HWM93. The observed meridional wind exhibits a variation with a 5-6 h period, which is not represented by the existing empirical wind model and the amplitude is significant during the December solstice. The importance of such a variable wind field in the thermosphereionosphere system, including plasma instabilities and onsets of plasma bubbles, should be analyzed further. The day-today variability of ionospheric height and its relationship to onsets of equatorial spread F are given elsewhere . The preliminary analysis presented here is limited to the nighttime because we used directly scaled virtual heights, which are a good indicator of the layer height only in the nighttime. Peak height analysis using the ionospheric transmission factor M(3000)F2 and foF2 (Shimazaki, 1955;Berkey and Stonehocker, 1989) would be a more universal means, and will be a future topic. Also, the present analysis is on one-month median values in which possible local disturbances such as traveling ionospheric disturbances (Kelley et al., 1981) are smoothed out. The equatorial pair of Chumphon and Bac Lieu have the potential for a more detailed analysis based on a day-to-day values.