Variations of the ionospheric TEC using simultaneous measurements from the China Crustal Movement Observation Network

Variations of the ionospheric Total Electron Content (TEC) over China are investigated using the TEC data obtained from China Crustal Movement Observation Network in the year 2004. The results show a single-peak occurred in post-noon for the diurnal variation and two peaks exit around two equinox points, respectively, for the seasonal variation. Overall, the values of TEC increased from the north to the south of China. There were small but clear longitudinal differences in both sides of the longitudes with zero magnetic declination. The intensity of the day-to-day variation of TEC was not a monotonic change along the latitudes. It was usually weaker in the middle of China than that in the north or south. Comparing with the maximum F-layer electron density ( NmF2) derived from the ionosonde stations in China, it is found that the day-to-day variation of TEC was less significant than that of NmF2, and that the northern crest of the equatorial anomaly identified from the NmF2 data can reach Guangzhou-region. While, the TEC crest was hardly observed in the same location. This is probably caused by the tilt of topside ionosphere near the northern anomaly crest region at lower latitudes.


Introduction
The ionospheric Total Electron Content (TEC), which has both temporal and spatial variations, is one of the most important ionospheric parameters.The variation of the iono-spheric TEC, for example, can have an apparent influence on radio propagation, especially on earth-space communication and satellite navigation.With observations from the Faraday rotation of a moon-reflected radar (Browne et al., 1956;Evans, 1956Evans, , 1974Evans, , 1977) ) and the first satellite, Sputnik 1, significant information on the ionospheric TEC has been obtained (Nisbet, 1960;Nelson, 1968;Evans, 1977).
Recently, due to the wide use of GPS receivers both on satellites at low earth orbit and on the ground, continuous and long-time ionospheric TEC data with increasing accuracy have been obtained and used to analyse the diurnal, seasonal and solar cycle variations of the ionospheric TEC over the Earth (e.g., Breed et al., 1998;Xiao and Zhang, 2000;Stamatis et al., 2004;Yu et al., 2006;Liu and Chen, 2009;Perevalova et al., 2010).Some new parameters, such as the Global Electron Content (GEC) (Afraimovich et al., 2008;Astafyeva et al., 2008) and the global mean TEC (Hocke, 2008(Hocke, , 2009;;Liu et al., 2009), have also been derived to track the global characteristics of ionospheric dynamics.It is shown that there are annual variations, semiannual variations and regional variations in terms of TEC, and there is a strong dependence on the solar and geomagnetic activities for the TEC.At the same time, detailed analysis of temporal-spatial TEC variations have been done based on GPS receivers on the ground in some regions, such as in Victoria (Wu et al., 2006), India (Bagiya et al., 2009;Rama Rao et al., 2006) and Salisbury (Breed et al., 1998).Up to now, however, there are only a few papers for the variations of the ionospheric TEC over the whole China region based on direct observations (Huo et al., 2005;Xiao and Zhang, 2000)."*" stands for the GPS site where there is an ionosonde nearby.G: Geographical co-ordinate; M: Geomagnetic co-ordinate; Lat: Latitude; Lon: Longitude Since V. Appleton (1946) and Liang (1947) found an equatorial anomaly, based on data of ionospheric critical frequencies, much work about this phenomenon has been done.Compared to the data of foF2, the latitudinal distribution of the ionospheric TEC shows a similar structure to the maximum electron density of the ionosphere (Huang et al., 1987).This means that the equatorial anomaly can also be found using TEC data.Huang et al. (1989) also pointed out that the location of the equatorial anomaly crest determined by either TEC or foF2 always varies with not only season, but also the time of the day.Stamatis et al. (2004) showed that the day-today variation for TEC lies between those of foF2 and NmF2, depending on the level of solar activity.
In this paper, the diurnal, seasonal and day-to-day variations of the ionospheric TEC over China were investigated, based on the vertical TEC (VTEC) data from 25 GPS receiver sites from China Crustal Movement Observation Network in 2004.Combining the foF2 data, the day-to-day variations are examined for the TEC and the NmF2, and the location of the equatorial anomaly crest identified by the TEC is compared with that by the NmF2 data.Possible mechanisms for the observed behaviours are discussed.

Data
Table 1 shows the locations of the sites we used, including 25 GPS receiver sites in China Crustal Movement Observation Network and 5 ionosondes stations, where the "*" stands for the GPS site with an ionosonde nearby.All sites run routinely.From Table 1, we can see that the GPS sites cover the whole China region.Therefore, the variable characteristics of the ionospheric TEC over China can be described based on the TEC data here.
The TEC data are calculated using the method as described by Zhang et al. (2008) and Xue et al. (2012).The equations of the GPS P code observation are given as follows: where P i 1j and P i 2j are the dual-frequency P-code pseudorange observational quantities, ρ i j is the distance from the ith satellite to the j -th station, c is the velocity of light in the vacuum, dt i and dt j are, respectively, the clock bias of the satellite and receiver, d i tropj is the tropospheric delay, d i ionkj , dq kj and dq i k are the ionospheric delay, the GPS receiver hardware delay and the satellite hardware delay at two different frequencies (k = 1, 2), respectively.For a high-frequency system, such as GPS, we just consider 1st-order series ex-pression of the ionospheric delay as Eq.(2).
in which, TEC is the ionospheric total electron content along the signal path, in units of el m −2 , f is the signal frequency in units of Hz and d ion is expressed in units of metres.Making the difference between two observed values of P-code pseudorange in Eq. ( 1) and introducing Eq. ( 2), we can obtain the TEC along the path from the station to the satellite as follows: The TEC in Eq. ( 3) is slant TEC (STEC).The elevation angles are limited in a range larger than 10 • in data handling.
Adopting the multinominal expansion model of single-layer ionosphere (450 km) and the projection function of cosine, we can obtain the vertical TEC (VTEC) at the station and the hardware delays.Applying this method for Chinese GPS regional network, the relative accuracies and RMS of the quasi-realtime VTEC can reach 2.0 TECu and 1.925 TECu, respectively (Xue et al., 2012).The sample rate of the STEC, namely the TEC in Eq. ( 3), is 30 s.The VTEC at GPS site is averaged every 15 min and then used in this paper.
The TEC 2004 data are considered, when it is in the declining phase of the 22th solar cycle.This can be seen clearly from Fig. 1, showing the variation of the F107 index from 1996 to 2007.The blue line denotes the daily F107 index, and the red line is the smoothed F107 index with a window of 81 days.The F107 index in 2004 is larger than 90 and less than 120.

Variations of ionospheric TEC
Based on the averaged TEC data above, the TEC contour maps over China were derived using the corrected Kriging method (Liu et al., 2008).As an example, TEC contours at four different times on 20 March 2004 (a: 02:00 BST; b: 08:00 BST; c: 14:00 BST and d: 20:00 BST) are shown in Fig. 2, where "*" stands for the GPS site, and the numbers are the TEC values in units of TECu (1 TECu = 1×10 16 el m −2 ).The F107 index is 112.7 and the Ap index is 6, 12, 4 and 7 for a, b, c and d, respectively.BST stands for Beijing Standard Time, which equals to UT (universal time) added by 8 h.It can be seen that there are sunrise and sunset effects shown in Fig. 2b and Fig. 2d, respectively.As a whole, the TEC have the largest value at 14:00 BST and the smallest value at 02:00 BST.The TEC values increase from the north to the south of China.The diurnal variation, the seasonal variation and the day-to-day variation of the ionospheric TEC over China are studied as follows.

Diurnal variation of TEC
The characteristics of diurnal variation of TEC are investigated in terms of median, upper-quartile (uq) and lowerquartile (lq) of TEC for one month.As an example, Fig. 3 shows the diurnal variation of TEC at Guangzhou (Guan) in March.It can be clearly seen that there is an obvious singlepeak with a value of about 80 TECu around 14:00 LT and a short-lived pre-dawn minimum of about 10 TECu.This pattern of diurnal variation is similar to those over the same region both in 1998 (Xiao and Zhang, 2000) and 2001 (Huo et al., 2005) as well as over the other places (e.g., Chauhan et al., 2011;Gupta and Singh, 2000).Furthermore, the diurnal variations of the TEC data from 6 GPS sites along 120 • E are also investigated.The results are shown in Fig. 4, where the columns show the diurnal variation in spring, summer, autumn and winter, and rows present the TEC data from Hlar, Bjsh, Tain, Whjf, Guan and Qion, respectively.These sites are sorted in turn by their geographical latitude from higher to lower.We take Guan as an example again, which is displayed in the fifth row of Fig. 4 from the top, to analyse the diurnal variation.It can be seen that an obvious peak exists in each month.The TEC peaked at ∼60 TECu in September and ∼50 TECu both in June and December.These values are lower than the peak values in March, says near 80 TECu.At the other sites, the TEC also shows a single peak in the post noon, with the peak value in vernal and autumnal equinoxes are generally higher than in summer and winter.The single peak structures at higher latitudes are not as sharp as those at low latitudes, especially in June.In addition, the peaks at lower latitudes appeared later than those at the higher latitudes in the same month.While for the same site, the peaks in June occur later (∼16:00 LT) than in other months (∼14:00 LT).
Furthermore, it is worth noting that the peak value in winter was a little lower than in summer, which means that there was no winter anomaly in 2004.This is different from the observations in 2001 shown by Huo et al. (2005), which was interpreted to be caused by the difference of solar activity between these two years.Solar activity in 2004 was at a declining phase, while at the solar maximum phase in 2001, as shown in Fig. 1.
From the east to the west of China, the diurnal variations of the TEC over 4 GPS receiver sites along ∼40 • N are investigated, as shown in Fig. 5.The 4 GPS receiver sites are Mizu, Bjsh, Dxin and Wush, respectively.Here, data from one Japanese GPS site at Mizusawa (Mizu, 141.1 • E, 39.1 • N) are used.Generally speaking, it can be found from Fig. 5 that the diurnal variations are very similar among the four stations, indicating that the longitudinal variation of TEC could be described as LT variation in the first order.Examining Fig. 5, we do notice clear longitudinal differences.The times of the daily peaks occurred later on the east-most sites than on the west sites.The daytime density is normally lower in the east-most site than in the west site, and this turns opposite during the nighttime.These results agree well with the recent studies (Zhang et al., 2011(Zhang et al., , 2012)).These longitudinal differences are caused by the difference in magnetic declination which gives rise to upward and downward ion drifts across the zero declination for a given thermospheric zonal wind direction.In China region, the magnetic declination is westward (negative) on the east side of China and eastward (positive) on the west side, with zero value along ∼100 • E longitude.Both Mizu and Bjsh are located east of 100 • E, while Dxin is 100 • E and Wush is located west of 100 • E. Here, the TEC observations confirm that the longitudinal differences occurred over China region, too.

Seasonal variation of TEC
Some seasonal variations of TEC have been discussed in the above section using the monthly median values.In this section, the monthly grand-mean (TEC mean ) and the monthly grand-variation intensity (TEC intensity ) are introduced to study overall properties of TEC seasonal variation.TEC mean and TEC intensity are computed by Eqs. ( 4) and ( 5 where D is the number of the day in one month, 96 is the number of the TEC data in one day. As shown in Sect.3.1, we choose the six sites along 120 • E and the four sites along 40 • N to investigate the seasonal variation of TEC over the China region.Figure 6 shows both the TEC mean and the TEC intensity at the six sites (same as in Fig. 4) along 120 • E. It can be found that, for the two parameters, there are two peaks at the equinoxes at all sites, which are the same as the results of the previous researches (e.g., Huang et al., 1989;Tsai et al., 2001).The values were higher and the double peak structure was more obvious at the low latitudes, which were similar to the seasonal variation of the maximum electron density (e.g., Zhang et al., 2005;Liu et al., 2009;and references therein).
Figure 7 shows both the TEC mean and the TEC intensity at the four sites (same as in Fig. 5) along 40 • N. We can see that they are almost the same values and the same variations, which indicates that the longitudinal differences are much smaller than latitudinal differences.This trend implies that  not only the TEC mean , but also the TEC intensity as a whole mainly varied with the latitude.The lower latitude, the larger values of these two parameters.

Day-to-day variation
Four parameters are introduced to describe the day-to-day variation of TEC.They are relative upper-quartile (U ), lower-quartile (L), relative range of quartiles (R) and ratio of upper and lower quartiles (K), and are defined by Eqs. ( 6)-( 9), respectively.
Here the median (i) , uq (i) and lq (i) denote the median value, the upper quartile and the lower quartile of TEC at time i in one month, respectively.The mean values of these parameters for all times in 2004 are shown in Table 2 for 5 sites (Hlar (49.
It can be seen that the U , L and R parameters had smaller values in the middle region of China, such as Bjsh and Urum.It means that the intensity of the TEC dayto-day variation was not a monotonic change along the latitudes.It was usually weaker in the middle of China than that in the north or south.In addition, the mean of the parameter K was larger than 1, which indicates that the distribution of TEC was asymmetric and the upper half of distribution is more dispersed than the lower half.

Disscussion
In China, there are 11 ionosonde sites all together, from which we can get continuous and accurate ionospheric parameters, such as foF2.Among them, 5 ionosonde sites are close to the GPS sites, which are shown in Table 1 with "*".From foF2, the NmF2 can be computed using Eq. ( 10).
where the units of foF2 and NmF2 are Hz and el m −3 , respectively.
The discussion below is based on TEC and NmF2 data, focusing on the ionospheric day-to-day variation and the location of equatorial anomaly crest.

Comparison of day-to-day variation between TEC and NmF2
The day-to-day variation of TEC has been investigated with the parameters of U , L, R and K. Inserting the uq, lq and median value of NmF2 into Eqs.( 6)-( 9), these parameters for the day-to-day variation of NmF2 were also obtained and shown in Table 3.It can be seen that the values of TEC's U , L and R are usually smaller than those of NmF2, respectively.This is because the TEC is an integral parameter of electron density at all altitudes, while the NmF2 is just the electron density at a particular altitude.The variation of TEC, subsequently, is usually weaker than that of NmF2.For the  parameter of K, however, there is no distinct relationship between TEC and NmF2.
Besides, some detail comparisons of the day-to-day variation between TEC and NmF2 were also investigated by taking the R value as an example.Figure 8 shows the variation of R with LT at Beijing in 2004.It can be found that the mean of R of TEC (0.24) is less than that of NmF2 (0.32), and the R of TEC fluctuates around their mean value more weakly than the R of NmF2 does.At other sites, the R varies in a similar way. Figure 9 shows the seasonal variation of R at the 5 sites.It shows clearly that, except a few months at Hlar, the R values of TEC are always smaller than those of NmF2, especially at Guan and Qion.It is worth noting that there are two peaks of R of TEC occurring in March and July, respectively.

Location of equatorial anomaly crest
It is well known that the low-latitude region of China is influenced by the equatorial anomaly (e.g., Liang, 1947).In the past, however, the location of the equatorial anomaly crest was determined based on foF2 or NmF2.In the south of China, there are 2 ionosondes at Guangzhou and Haikou, respectively, and Haikou (20.2 • N, 110.3 E) is very close to  Qiongzhong (19 • N, 109.8 • E) where there are GPS observations.Therefore, the location of equatorial anomaly crests in the south of China can be analysed in terms of the differences of TEC values between Guangzhou and Qiongzhong and the differences of NmF2 values between Guangzhou and Haikou, respectively.Since it is believed that Guangzhou, with a magnetic latitude of 17.5, is very close to the northern crest of the equatorial ionization anomaly, it is assumed that if the difference is positive, the crest may reach Guangzhou or somewhere toward the north of Guangzhou; if the difference is negative, the crest may just reach Qiongzhong or somewhere around Qiongzhong, but further away from Guangzhou.
The differences for different months and local times are shown in Fig. 10. Figure 10a shows that the TEC values at Guangzhou are larger than that in Qiongzhong only around 14:00 LT in March and April, while in the other months in the post noon, TEC values at Guangzhou are smaller than that at Qiongzhong, especially in winter and equinoxes.However, the NmF2 values at Guangzhou are usually lager than that at Haikou except for the period of time around 16:00 LT in July (see Fig. 10b).It suggests that the NmF2 crest could reach the region around Guangzhou usually, but the TEC crest can be hardly observed in Guangzhou.
This phenomenon could be explained by using the theoretical model named Sami2 (Huba et al., 2000), which is a low-latitude model of ionosphere, developed by the Naval Research Laboratory (NRL).It is the first low-latitude ionospheric model including the ion inertial terms in the momentum equations.The neutral species are specified using the Mass Spectrometer Incoherent Scatter model (MSIS86) and the Horizontal Wind Model (HWM93).The daytime photoionization model uses the solar EUV flux model for aeronomic calculations (EUVAC) developed by Richards et al. (1994).The nighttime photoionization model uses the so-lar EUV flux prescribed by Strobel et al. (1974) and the photoionization cross-sections are obtained from Oran et al. (1974).The electron density in the ionosphere at 15:00 LT on 1 October 2004 is calculated by Sami2, with the Ap index 3 and F107 index 88.1, and the average of F107 index for the 3 months is 113.8.Based on the simulated results from Sami2, the ionospheric TEC, NmF2 and the peak height of the F2 layer (hmF2) were also calculated.The first two parameters are shown in Fig. 11b, while the last one is shown in Fig. 11a by the heavy line.Here, we take Guan as an example again to check the Sami2 model over the China region.The TEC value from the Sami2 model is 62.9 TECu at 23 • N at 15:00 LT in March 2004.It is very close to the TEC value of the observation, which is 63.1 TECu at Guan (23.1 • N) at the same time.This indicates that the Sami2 model is valid for this particular simulation.
The results show that the location of the crest identified by TEC (the solid vertical line in Fig. 11b) is different from that determined by NmF2 (the dotted vertical line in Fig. 11b).The location of TEC crest is closer to the equator.This phenomenon can be interpreted to the tilt of the topside ionosphere near the anomaly crest region.At the altitudes above hmF2 (given by the heavy line in Fig. 11a), the distribution of ionospheric electron density tilts down with latitude, so that, at the same altitude, the electron density at the latitude where NmF2 crest locates (A) is smaller than that at latitudes closer to equator.So the TEC which is integral along the altitude at the latitude of "A" is smaller than the TEC at the lower latitudes.
The tilt of the topside ionosphere is caused by the fountain effects (Martyn, 1947(Martyn, , 1953)).With an eastward electric field, the plasma in the equator drifts upward to a higher altitude, then diffuses and descends to higher latitudes.At the same time, the magnetic field lines are tilted down from the equator to higher latitudes.This results in the equatorial ionospheric anomaly being tilted down as the plasma moves from lower latitudes to higher ones.As a result, the latitude of the TEC crest is lower than that of the NmF2 crest.

Summary
Based on the TEC data from China Crustal Movement Observation Network during declining solar activity phase of 2004, the seasonal, local time, latitudinal and longitudinal variations of the TEC as well as its day-to-day variation were studied.As a whole, the seasonal variation of the ionospheric TEC shows a double-peak structure at the vernal and autumnal equinoxes, and the diurnal variation of the ionospheric TEC shows an obvious peak structure in the post noon.These peak values increased from the north to the south of China.The peak time of the diurnal variation was earlier at higher latitudes than at lower latitudes.There are small, but clear longitudinal difference in both sides of the longitudes with zero magnetic declination, which is probably caused by the

Fig. 3 .
Fig. 3. Diurnal variations of TEC at Guangzhou (Guan) in March 2004 (the solid line is the median TEC value, the dotted one is the UQ value and the dashed one is the LQ).

Fig. 4 .
Fig. 4. Diurnal variations of TEC (median) in four months (March, June, September and December) in 2004 at 6 GPS sites along 120 • E.

Fig. 5 .
Fig. 5.Diurnal variations of TEC (median) in four months (March, June, September and December) in 2004 at 4 GPS sites along 40 • N.

Fig. 6 .
Fig. 6.Seasonal variations of the monthly grand-mean (a) and the monthly grand-variation intensity (b) of TEC at six GPS sites along 120 • E.

Fig. 7 .
Fig. 7. Seasonal variations of the monthly grand-mean (a) and the monthly grand-variation intensity (b) of TEC at four GPS sites along 40 • N.

Fig. 8 .
Fig. 8. Relative ranges of day-to-day variations of both TEC and NmF2 with LT at Beijing.

Fig. 9 .
Fig. 9. Relative ranges of day-to-day variations with month for both TEC and NmF2.

Table 1 .
Information on the GPS sites and ionosondes used in this work.

Table 2 .
The mean value of U , L, R and K of TEC in 2004.

Table 3 .
The mean value of U , L, R and K of NmF2 in 2004.