Two-dimensional imaging of large-scale traveling ionospheric disturbances over China based on GPS data

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 117,, doi: /2012ja017546, 2012 Two-dimensional imaging of large-scale traveling ionospheric disturbances over China based on GPS data Feng Ding, 1 Weixing Wan, 1 Baiqi Ning, 1 Biqiang Zhao, 1 Qiang Li, 2 Rui Zhang, 2 Bo Xiong, 1,3 and Qian Song 1,3 Received 20 January 2012; revised 1 June 2012; accepted 3 July 2012; published 17 August [1] This paper reports the first results of the 2D imaging of large-scale traveling ionospheric disturbances (LSTID) using GPS network data from China, combined with observations of these events using an ionosonde chain. 2D TEC perturbation maps for North America were also constructed to allow the study of LSTIDs on a global scale. During the medium storm on 28 May 2011, the onset of a substorm initiated a slow-speed LSTID over North America just after midnight. Subsequently, an LSTID reached China 1.5 hours later, at dusk. A second LSTID was observed over China before midnight, 6.6 hours after substorm onset. The phase fronts of the China events had a front width of at least 1600 km, and moved southwestwards at a speed of m/s and m/s, respectively. Ionosonde data addressed a downward vertical phase velocity of 75 m/s for the dusk event and 60 m/s for the night event. Although the nighttime LSTID travelled farther south than the earlier dusk event, both disappeared in South China, and this was due to increase of the attenuation at low latitudes. According to the energy dissipation equation of atmospheric gravity waves there is severe dissipation due to viscosity and heat conductivity at low latitudes, since such dissipation increases strongly with time; dissipation due to ion drag is less important but cannot be ignored because of enhancement in background TEC; In addition, uplift of the ionosphere at low latitudes is another factor that results in a reduced amplitude of TEC perturbation at low latitudes. Citation: Ding, F., W. Wan, B. Ning, B. Zhao, Q. Li, R. Zhang, B. Xiong, and Q. Song (2012), Two-dimensional imaging of large-scale traveling ionospheric disturbances over China based on GPS data, J. Geophys. Res., 117,, doi: /2012ja Introduction [2] Large-scale traveling ionospheric disturbances (LSTIDs), with horizontal wavelengths of more than 1000 km and periods of h, are frequently observed at high and middle latitudes. Previous studies demonstrate that LSTIDs act as passive tracers of atmospheric gravity waves (AGWs) [Francis, 1975]. While medium-scale traveling ionospheric disturbances (TIDs) could be excited at any latitude by localized sources such as jet streams [Buss et al., 2004] or meteorological processes [Wan et al., 1998; Boška and Šauli, 2001], it is widely accepted that the enhancement of the auroral electrojet or particle precipitation is the primary sources of LSTIDs. Sources of LSTIDs other than auroral 1 Beijing National Observatory of Space Environment, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China. 2 National Earthquake Infrastructure Service, China Seismic Administration, Beijing, China. 3 Graduate University of Chinese Academy of Sciences, Beijing, China. Corresponding author: F. Ding, Beijing National Observatory of Space Environment, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, , China. (dingf@mail.iggcas.ac.cn) American Geophysical Union. All Rights Reserved /12/2012JA activities have been discussed previously [Chimonas, 1970; Vadas and Liu, 2009]. Recently, Vadas and Liu [2009] found through numerical modeling that the dissipation of mediumscale gravity waves in the thermosphere could generate largescale secondary AGWs, thereby providing new insights into the possible excitation mechanism of LSTIDs at low latitudes. [3] Previous studies have reported two patterns of mechanism that explain the long-distance propagation of AGWs in the upper atmosphere: ducted AGW modes and freely propagating internal gravity waves [Hunsucker, 1982]. Many works have examined these patterns, of which some are considered here. The steep height gradient of atmospheric temperature in the lower thermosphere can support the ducted gravity waves, which propagate horizontally with energy concentrated near the sudden temperature rise [Francis, 1973]. These waves belong to several guided or quasi-guided wave modes with a discrete spectrum [Hunsucker, 1982]. However, Richmond [1978a] argued that a realistic temperature profile might not produce such a ducting mechanism, and an alternative concept of freely propagating waves that was proposed by Hines [1960] appears more practical. The waves propagate obliquely in the atmosphere, some traveling long distances through reflection in the thermosphere due to the 1of11

2 Figure 1. Locations of GPS stations (dots) and ionosondes (stars) in China. temperature gradient or gradients in viscosity and heat conductivity [Richmond, 1978b; Maeda, 1985]. Using a simple numerical model, Richmond [1978a] obtained propagation velocities and attenuation distances of AGWs that are similar to those reported by Francis [1973]. [4] Initially, the propagation features of storm-time LSTIDs were observed by HF Doppler, ionosonde, incoherent scatter radar, and total electronic content. These early observations gave a detailed insight into the excitation and propagation features of LSTIDs at a global-scale. For example, Williams et al. [1988] used the EISCAT incoherent scatter radar, and a chain of ionosondes and HF Dopplers, to monitor the excitation and propagation of LSTIDs over northern Europe and the UK. They found that the periodic variation in the magnetospheric electric field generated atmospheric waves with the same period, which were detected over the UK an hour after excitation. Using ionosonde chains in Japan and Australia, Hajkowicz [1990] investigated the conjugate effects of LSTIDs and found simultaneous trains of LSTIDs over a large range of southern and northern latitudes. Rice et al. [1988] conducted a combined observation of LSTIDs during a moderate storm using ionosondes, HF Doppler, and incoherent scatter radar in North America and Europe. They found that an LSTID excited in the night sector of northern Russia could propagate across the polar region and reach the middle and high latitudes of North America. However, because of the limited number of stations available in these early studies it was often difficult to identify and continuously monitor spatial and temporal variations in individual LSTID events over such large areas. [5] Since the late 1990s, the increased density of the GPS receiver network has made it possible to continuously monitor the 2D propagation of band-like structures in LSTIDs over a wide area. Previously, the observation of LSTIDs using TEC perturbation maps was conducted mainly in Japan [Saito et al., 1998, 2001; Tsugawa et al., 2003, 2004, 2006], North America [Nicolls et al., 2004; Ding et al., 2007, 2008], and Europe [Borries et al., 2009]. Tsugawa et al. [2003] recorded the passage of two consecutive TEC enhancements over Japan caused by a LSTID during a storm on September 22, The statistical study of Tsugawa et al. [2004] shows that, in addition to the LSTIDs that propagate westwards, many LSTIDs observed over Japan propagate eastwards, which is thought to be due to the westward deviation of the geomagnetic declination from north in this region. Ding et al. [2007] produced TEC perturbation maps during the superstorm of October 29, 2003 over North America that showed that the propagation direction of the 2D band like structures altered as they moved from high to middle latitudes, and they relate this to a change in the position of the electrojet enhancement area near the auroral oval. Borries et al. [2009] used the European GNSS data to map TEC perturbation caused by LSTIDs over Europe, and showed that the average wavelength of LSTIDs over Europe is similar to that over Japan; however, LSTIDs over Europe move considerably faster than those observed over Japan. [6] The observation of LSTIDs over China using dense GPS networks has received less attention. Mainland China covers the latitude range from 18 to 53 N (i.e., magnetic latitude: 7 42 N). As the northern boundary of the equatorial anomaly crest is located around 19 N (magnetic latitude), China spans middle latitudes, low latitudes, and the equatorial anomaly crest. In contrast, GPS stations on mainland Japan lie north of the boundary of the equatorial ionospheric anomaly (i.e., magnetic latitude: N [Shiokawa et al., 2002; Tsugawa et al., 2004]), while GPS stations in North America and Europe mainly lie in the middle to high magnetic latitudes at N[Nicolls et al., 2004] and N[Borries et al., 2009], respectively. Consequently, China s location provides an opportunity to study the nature and propagation of LSTIDs in lower latitudes. [7] This paper reports the first results of the 2D imaging of large-scale TIDs using GPS network data from China, combined with the observation of these TIDs using an ionosonde chain. We collected TEC data from 231 GPS stations in the Crustal Movement Observation Network of China (CMONOC) and the IGS. Using these data, we derived the 2D TEC perturbation maps and observed the band-like structures of LSTIDs during a recent geomagnetic storm (April 28, 2011). To facilitate the study of LSTIDs at the global scale, we also created TEC perturbation maps for North America using GPS data from the IGS and the Southern California Integrated GPS Network. We investigated the propagation features of LSTIDs, both near the equatorial ionization anomaly (EIA), and at middle to high latitudes. 2. Data and Methods 2.1. Data [8] The installation of GPS stations across China began in the late 1990s when the first stage of the Crustal Movement Observation Network of China (CMONOC) was installed, and this includes 28 GPS stations that carry out continuous observations [Mao et al., 2008]. The second stage of CMONOC was completed in 2010 and consists of more than 200 GPS stations throughout the country. For this paper, we collected data from 231 GPS stations in and around China (Figure 1), of which 188 stations belonged to CMONOC, and the other 53 stations were part of the IGS (International GNSS service) network. The majority of the stations were located in the east of China. The number of GPS stations in China is less than that in Japan (i.e stations) and North America (i.e. 600 stations in 2003). In order to ensure a sufficient number of data points in the maps, we set the lower limit of the 2of11

3 elevation to be 30. The data show carrier phase and pseudorange measurements in two L-band frequencies with a resolution of 30 seconds. We used the carrier phase data to compute the slant TEC, which is the integral of electron density along the GPS satellite receiver path. The smoothed pseudo-range measurements are used in the estimation of integer ambiguities. [9] In addition, an ionosonde chain tracked the propagation of the LSTIDs over China. This chain was composed of 4 DPS-4 ionosondes located at Mohe (52 N, E), Beijing (40.4 N, E), Wuhan (30.5 N, E), and Sanya (18.3 N, E), respectively (Figure 1), which belong to the Beijing National Observatory of Space Environment. The ionosonde locations extended from the far north of China to the far south, and had a longitudinal range of less than 14. Hence, they are suitable for the detection of the long-distance propagation of LSTIDs in this region GPS Data Processing [10] With the exception of the filter, the TEC perturbation maps were developed, and the propagation parameters determined, according to the method of Ding et al. [2007]. First, we obtained the TEC perturbation time series by filtering out the background trend from the original TEC series observed by each GPS receiver-satellite pair. The filter was based on the following equation: VTEC w ¼ STEC=M h B s r =M h C 0 C 1 ðlat lat 0 Þ C 2 ðlat lat 0 Þ 2 C 3 ðlt LT 0 Þ ð1þ where VTEC w is the TEC perturbations time series corresponding to a GPS receiver-satellite pair; STEC refers to the original slant TEC time series; Lat and LT are the geographic latitude and local time of the ionospheric pierce points, respectively; Ionospheric pierce points are the points where the line-of-sight between the satellite and the ground receiver intersect the ionosphere, under the estimation of a single layer. Lat 0 and LT 0 are the latitude and local time when the LOS reaches its maximum elevation, respectively; B r s is the instrumental bias; and C 0, C 1, C 2, and C 3 are the fitting coefficients to be solved. The instrumental bias and the fitting coefficients were calculated for each GPS receiver satellite pair. Cycle slips and instrumental bias were corrected by comparing the data with those of Global Ionospheric Maps (GIM) from the Crustal Dynamics Data Information System (CDDIS) [Noll, 2012], which were interpolated in both space and time. M h is the mapping function:!1 M h ¼ 1 R e cosðeleþ 2 2 ð2þ R e þ h i where R e is the average radius of the Earth; h i is the height of the ionosphere under the estimation of a single layer; ele is the elevation of the line of sight (LOS) that connects the receivers and the ionospheric pierce points. The mapping function is described in Jakowski et al. [2012]. [11] In equation (1), the background trend of the TEC is expressed as a one-order function of local time, as well as a two-order function of latitude. This differs from the approach of Ding et al. [2007], who express the background TEC as a one-order function of local time and latitude. This amendment was required here because the latitudinal trend of background TEC varies more sharply in China than in North America due to the latitudinal difference between the two regions. Compared with JPL GIM data, there is an average error of TECU for the background TEC derived in the present study (1 TECU = ele/m 2 ). Tsugawa et al. [2004] reported an average amplitude of 1.3 TECU over Japan. The error is much smaller than the amplitudes of LSTIDs at low latitudes. To generate the TEC perturbations time series we input the data from the observed slant TEC series into equation (1), and calculated the value of C 0, C 1, C 2, and C 3 using a least squares method. [12] Then, we divided the area bounded by N, E, into pixels with a size of 1 latitude 1 longitude. The amplitude of TEC perturbation in each pixel was set to be the average of the TEC variations (VTEC w ) for all of the GPS satellite-receiver paths whose ionospheric pierce points crossed the pixel during the time LT 150 seconds. Thus, we obtained the 2D TEC variation map for that time frame. We repeated this calculation to obtain a sequence of maps with a temporal resolution of 5 minutes. [13] During the passage of a disturbance, a number of band-like structures will appear on the maps, which will move at a certain speed in one direction. By monitoring the movement of these band-like structures, we were able to identify LSTID events and calculate their horizontal phase velocities and propagating azimuths. [14] Some authors have reported the limitations of TEC measurements regarding observations of TIDs. As TEC is the line-of-sight integral of electron density, it provides no information on the height distribution of TIDs, because ionospheric disturbances along the ray path do not give rise to disturbances in TEC. Jacobson et al. [1995] found that the TEC signatures of TIDs are sensitive to the orientation of the line of sight relative to both the geomagnetic field and the TID phase velocity. However, the greatest advantage of GPS TEC measurement is the dense distribution of GPS receivers and multiple GPS satellites. It is generally possible to simultaneously observe 2 8 slant TEC time series from a single station, with their lines of sight (LOS) moving in different directions. We set the disturbance value in each pixel to be the average of disturbances for all the lines of sight with their ionospheric pierce points crossing the pixel. We expect this method can offset the observational bias that arose from the orientations of lines of sight. 3. Observations 3.1. The Storm and Substorm [15] A medium-sized geomagnetic storm occurred between May 27 and 31, Figure 2 plots the temporal variation of both the Dst index (Figure 2a), and the AU and AL indices (Figure 2b). The Dst index starts to decrease at around 1500 UT on May 27, followed by a slight increase between 0200 and 0600 UT on May 28, before continuing to decline to a minimum of 80 nt. The value of Dst recovered slowly from this point onwards. This is a typical two-step medium-sized storm, whose main phase undergoes a two-step growth in the ring current because the B z component of the interplanetary magnetic field (IMF) turns southward twice during the storm, causing two injections of 3of11

4 Figure 2. Temporal variations in (a) the Dst index and (b) the AU and AL indices over the period May 27 29, particles into the inner magnetosphere, and hence, two enhancements of the ring current [Kamide et al., 1998]. [16] An intense substorm occurred 2.3 hours after the second decrease in the Dst index (Figure 2b). The AL index began to drop at 0824 UT on May 28, and the substorm then experienced a rapid development of the expansion phase. The value of the AL index fell to a minimum of 2500 nt at 0849 UT, and this was followed by a quick recovery. In the following 4 hours, there were several smaller falls in the AL index, with a minimum value of around 1000 nt Observation of LSTIDs Over China [17] Two LSTID events were identified over China during the storm of May 28, Figures 3 and 4 present the sequence of 2D TEC perturbation maps for these two events. TEC perturbations can be observed in both figures, shown as Figure 3. Sequence of 2D TEC perturbation maps over China between 1000 and 1030 UT on May 28, The color interval depicts the deviation in the TEC (units: TECU). The black line marks the phase front of the LSTID event. 4of11

5 Figure 4. As for Figure 3, but between 1510 and 1600 UT. Black and gray lines mark the first and second phase fronts, respectively. a movement of band-like structures. Based on the study of Georges and Hooke [1970], TEC perturbations in Figures 3 and 4 can be interpreted as enhancements or depletions of electron density caused by atmospheric gravity waves. Some authors have stated that daytime gravity waves carry the charged particles along the magnetic field lines and cause variations in electron density [Jacobson et al., 1995; Beach et al., 1997]. However, Beach et al. [1997] indicated that nighttime gravity waves can produce vertical motion in the ionosphere, which can be measured by the ray-path difference in slant TEC. At night, the vertical force balance among diffusion, gravity, and electric fields dominates the ionosphere due to the absence of photoproduction [Kelley, 1989]. Any disturbances would cause the ionosphere to move up or down to re-establish vertical equilibrium [Beach et al., 1997]. Although vertical motion of the ionosphere can be measured by slant TEC series, due to the sensitivity of slant TEC to the elevation of line of sight, it cannot be measured from DTEC maps. This is not only because the slant TEC series had been converted to vertical TEC series, but also because the DTEC value for each pixel in the maps is the average of VTEC w for all the satellite receiver pairs. As the lines of sight move at different elevations and azimuths, this would offset any disturbance that is sensitive to the elevation of the line of sight. It is noted that height-disturbances in the nighttime ionosphere may cause electron density variations, because uplift/falling of the ionosphere leads to variations in the electron loss coefficient as well as upward/downward flux along the magnetic field lines. Such electron density variations and associated height-disturbances caused by gravity waves were observed simultaneously at Arecibo Observatory (18.34 N, E) [Nicolls et al., 2004]. These studies indicate that, although the vertical movement dominates the nighttime ionosphere, the atmospheric gravity waves can move the plasma up or down along the field lines and possibly lead to variations in electron density through flux and recombination. [18] Analysis of the sequence of maps in Figures 3 and 4 clearly shows the movement of positive/negative structures from northeast to northwest. The first event began 1.5 hours after the onset of the auroral substorm and was observed at dusk between 1000 and 1030 UT (i.e., LT). As illustrated in Figure 3, an area with negative differential TEC (DTEC) emerged at 1000 UT at a geographical latitude of around N (Figure 3a). The front of the negative area, as indicated by the black line, which was obtained through polynomial fits, separates negative and positive areas, moved southwestward for a distance of 500 km before it reached Central China (33 N). The maximum amplitude, phase velocity, and azimuth (clockwise from north) of the LSTID were TECU, m/s, and 198 6, 5of11

6 respectively. The amplitude is derived from temporal variations in DTEC at three fixed points (i.e. (30 N, 105 E), (35 N, 105 E), and (40 N, 105 E)). The errors here are the standard deviation of the parameters for all the grids along the phase fronts. This dusk event was characterized by a small amplitude, a limited latitudinal range, and a short lifetime, when compared with those observed at night, or around midday, over North America [Ding et al., 2007] and Japan [Tsugawa et al., 2003]. This is most likely to be related to significant variations in the background TEC at dusk (local time). Such variations in background TEC influence the steady propagation of phase fronts, and lead to a short lifetime, as well as a limited propagation range, of the LSTIDs. The climatology of the LSTIDs was conducted at a similar latitude to Japan [Tsugawa et al., 2004], but at a lower latitude than North America [Ding et al., 2008]. Statistical results from both regions indicate that there is a minimum in LSTID occurrence around dusk (local time). [19] The second LSTID event occurred in China around midnight (local time). It was observed at UT ( local time), with a maximum amplitude, phase velocity, and azimuth of TECU, m/s, and , respectively. The phase fronts of the LSTID (Figure 4) were 1600 km wide, and moved from northeast to southwest for a distance of >1000 km until they reached 29 N latitude. The phase fronts did not move south of 29 N. Compared with the dusk event, the band-like structures of the nighttime TID were much wider, and moved farther to the south. Figure 4 also shows an area with enhanced TEC in the southwest of China at N, E, which might be caused by ionospheric storms that occurred at low-latitudes. As this area of enhanced TEC remained almost stationary during the observation period, it was not classified as an LSTID. [20] The LSTIDs were observed at the same time by the ionosonde chain in China. Figure 5 presents the temporal variation of the virtual height at various detection frequencies, ranging from 2 to 13 MHz, with a step of 1 MHz. The frequencies and virtual heights were read from the F-layer trace in the ionograms recorded by ionosondes at Mohe (52 N, E), Beijing (40.4 N, E), Wuhan (30.5 N, E), and Sanya (18.3 N, E). [21] Large variations in virtual height occurred between 0600 and 2000 UT on May 28 (Figure 5). In each sub-plot, there is a phase difference among the peaks of variation in virtual height observed at different frequencies (see the red dashed lines). The peaks appear earlier at higher frequencies than at lower ones, indicating a downward vertical phase velocity. This is a typical characteristic of atmospheric gravity waves (AGWs), previously observed by ionosondes in Brazil [Becker-Guedes et al., 2007; Klausner et al., 2009]. [22] Only one ionosonde covered the track of the dusk LSTID (Figure 5c: Wuhan (30.5 N), UT). During this period, variations in virtual height can also be seen in the two ionosonde traces to the north of Wuhan (Figures 5a and 5b), which also seem to be related to the dusk event, although these northerly echoes are not sufficiently strong to unequivocally identify the LSTID. The nighttime LSTID was recorded by the ionosonde chain over a much wider area (Figures 5a 5c), and strong oscillations in virtual height can been seen at Mohe (52 N), Beijing (40.4 N), and Wuhan (30.5 N) between 1300 and 1800 UT ( LT). A downward vertical phase velocity of 75 m/s for the dusk event and 60 m/s for the night event can be estimated. The peak-to-peak time interval in the virtual height series indicates a wave period of 1.75 hours. It took 0.6 hours for the disturbance to travel from Beijing to Wuhan, as estimated by the time delay between the peaks at two ionosonde stations. This is similar to the time taken for the TEC perturbation fronts to travel from Beijing to Wuhan (time delay of 0.66 hours). In addition, strong variation in virtual height also occurred at the most southerly station, Sanya (18.3 N, Figure 5d). However, while perturbations occurred at the three northerly stations for only a few hours after substorm onset, there were intense perturbations at Sanya all day. The downward propagation of the phase front is also evident in Figure 5d, indicating the presence of a TID over this station. However, the variations of virtual heights at Sanya (18.3 N, Figure 5d) are inconsistent with those at Wuhan (30.5 N, Figure 5c). [23] Consequently, the TID observed by the ionosonde at Sanya is not considered to be the same TID recorded by the more northerly stations. TIDs similar to that seen in Figure 5d are frequently observed near the ionospheric crest [Fagundes et al., 2007; Klausner et al., 2009]. These events are likely to be manifestations of atmospheric gravity waves excited by some local source in the EIA region, such as the convective plume [Vadas and Liu, 2009], or the equatorial electrojet [Chimonas, 1970]. However, Fejer et al. [2007] and Sahai et al. [2009] argued that low-latitude ionospheric plasma perturbations during storms are caused mainly by the combined effects of relatively short-lived prompt penetration, and longer-lasting ionospheric disturbance dynamo electric fields. Attempts to identify a direct cause-and-effect relationship between the excitation and propagation of such TIDs in the EIA region are ongoing. [24] While there is a northerly limit of 40 N for LSTIDs observed by GPS measurements in China, the ionosonde data indicate that the LSTIDs may be found as far north as Mohe (52 N). Band-like structures in DTEC maps are not recorded over China to the north of 40 N, due to the low density of the GPS station network there (Figure 1). However, bandlike structures to the north of 40 N have been reported from Japan, with a latitudinal range of N, in the sequence of GEONET perturbation TEC maps ( kyoto-u.ac.jp/figs/map//2011/148_2011/). The Japan data show that events occurred at UT and UT, and these two events coincide well with the events observed in China, both temporally and spatially. It may be deduced that the LSTIDs observed in Japan and China are the same events, and that they may have originated from source regions in the northeast of Japan Observation of the LSTID Over North America [25] To compare the propagation features of LSTIDs in China with those observed at higher latitudes, we used the method of Ding et al. [2007] to generate 2D TEC perturbation maps for North America on May 28, 2011 (Figure 6). We used GPS RINEX data from the Southern California Integrated GPS Network and IGS (ftp://garner.ucsd.edu). Nine minutes before the significant drop of the AL index (just after midnight, local time), one LSTID event was recorded over North America. Figure 6 shows that the LSTID occurred between 0815 and 0930 UT ( local time), in the 6of11

7 Figure 5. Temporal variations in virtual height, at detection frequencies ranging from 2 to 13 MHz, and with a step of 1 MHz, recorded by the ionosonde chain in China on May 28, Frequencies are shown on each curve. The frequencies and virtual heights were read from the F-layer trace in the ionograms recorded by ionosondes at Mohe (52 N), Beijing (40.4 N), Wuhan (30.5 N), and Sanya (18.3 N). The time resolution is 5 min for Sanya, 10 min for Wuhan, and 15 min for Mohe and Beijing. The vertical dotted line marks the substorm onset. The red dashed lines in each plot connect the peaks of variation at different frequencies. northwest of North America. The band-like structures, with a maximum front width of 2300 km, moved slowly to the southwest for a distance of 600 km. The amplitude, horizontal phase velocity, and azimuth for the North America LSTID were TECU, m/s, and , respectively. This pattern is similar to previous observations in this area [Afraimovich et al., 2000; Nicolls et al., 2004; Ding et al., 2007, 2008]; i.e., a substorm over North America around midnight (local time) usually causes LSTIDs around the time of substorm onset. Ding et al. [2007] show that the equatorward expansion of the nightside auroral oval during a severe storm moves the southern boundary of the auroral oval very close to North America. The expansion of southern boundary of the auroral oval is obvious even during moderate storms, as reported by The National Oceanic and Atmospheric Administration (NOAA) on the website Consequently, LSTIDs excited near the nightside oval can quickly reach North America. 7of11

8 Figure 6. TEC perturbation maps for North America. The contour interval depicts the deviation in the TEC (units: TECU). Black and gray lines mark the first and second phase fronts, respectively. [26] Hence, during the day of May 28, 2011, when the substorm occurred at 0824 UT, the local time in North America was after midnight, and it was afternoon in China. A LSTID occurred immediately in North America around the time of substorm onset (Figure 6). Then, 1.5 hours later, a LSTID reached China around dusk (local time) (Figure 3). Finally, 6.6 hours after substorm onset, and before midnight in China, the second LSTID was recorded (Figure 4). Given that the China events were observed at the same time over Japan, the source regions of these events may lie in the northeast of Japan. It seems impossible that the LSTID in North America travelled southwestward for a long distance and reached China, because this would have taken more than 10 hours, and no LSTIDs were observed in China at the expected arrival time. Our study supports the view of Schunk and Nagy [2000] and Klausner et al. [2009], who proposed that atmospheric gravity waves are not global, but have localized sources. However, global propagation is possible for LSTIDs with large phase velocities, because such LSTIDs will experience relatively minor energy attenuation and will travel longer distances [Mayr et al., 1990]. For example, LSTIDs with phase velocities of m/s have been reported to be excited in one region and then propagate globally to different time zones [Rice et al., 1988; 8of11

9 dissipation during time 0 and t can be expressed as follows [Richmond, 1978b]: Z t vdt g k m p c p H þ m C 2 t 3 H 4 x 2 þ s 1B 2 t ð3þ r 0 Figure 7. Latitudinal variations in (a) background TEC and (b) F 2 peak heights along the meridian of 120 E. Solid line and dashed line in Figure 7a present latitudinal variations in TEC averaged over UT and over UT on 28 May 2011, respectively. The two time intervals correspond to the time of two LSTID events occurred in China. Solid line and dashed line in Figure 7b show the variation in h m F 2 averaged over the same time intervals on 28 May Dotted line and dash-dotted line in the plots represent the corresponding mean values during quiet days on May Perevalova et al., 2008; Cai et al., 2011]. LSTIDs propagating across the equatorial region were addressed in a previous study [Bruinsma and Forbes, 2009]. 4. Discussion [27] Although the night event in China travelled farther south than the dusk event, the phase fronts of both events disappeared near the northern boundary of the EIA region (30 N, magnetic latitude: 19 N). Despite the dense GPS network in southern China, no band-like structures were recorded. It is known that, besides TIDs, ionospheric irregularities in EIA region can also cause variations in electron density and thus influence the propagation of LSTIDs. However, in the present study, we calculated the rate of TEC index (ROTI), and found that no ionospheric irregularities occurred between 1500 and 1600 UT in South China, although such irregularities were previously observed there during storms [Li et al., 2006, 2009]. [28] However, the increased attenuation of atmospheric gravity waves at low latitudes may have been the main cause of the dissipation. Attenuation of AGWs in the thermosphere is due mainly to molecular viscosity, heat conductivity, and ion drag [Hunsucker, 1982]. Richmond [1978b] used a numerical model to examine the dissipation of large-scale gravity waves, and defined a wave energy dissipation rate coefficient v, which is the ratio of energy loss over a full wave period to the total energy density. Assuming that the waves were generated at time t = 0, the energy where g is acceleration due to gravity; p and r are the pressure and mass density of the atmosphere, respectively; k m and m are the coefficients of molecular heat conductivity and molecular viscosity, respectively; C p is the specific heat at constant pressure; H is the pressure scale height; x is the horizontal distance from the source; C is the limiting gravity wave speed; s 1 is the Pedersen conductivity; and B is the geomagnetic field that is vertically downward. [29] The equation yields quantity instructions for the dissipation of gravity waves. According to Richmond [1978b], the first term on the right of equation (3) represents molecular dissipation caused by viscosity and heat conductivity, and the second term represents Joule dissipation due to ion drag. Many previous studies have demonstrated that molecular viscosity and heat conductivity are more important than ion drag in the dissipation of gravity wave energy [e.g., Francis, 1973; Richmond, 1978b]. As stated by Richmond [1978b], for waves traveling between the source and a given observation point, the wave attenuation due to viscosity and heat conductivity increases strongly with time (as t 3 ), while dissipation due to ion drag increases linearly with time. Given that LSTIDs in the present study travelled rather slowly and for long distances before reaching low latitudes, severe dissipation due to viscosity and heat conductivity can be expected. The severe dissipation of slowly moving TIDs was also considered in the transfer function modeling work of Mayr et al. [1990]. [30] Given the increase in electron density at low latitudes, dissipation due to ion drag cannot be ignored. Figure 7a shows the latitudinal variation of background TEC along the meridian of 120 E during the time of two LSTID events. The data was from Global Ionospheric Maps (GIM) from the Crustal Dynamics Data Information System (CDDIS) [Noll, 2012]. Compared with quiet-time values, we see an increase in TEC by 30% in the dusk and by 100% at midnight at low latitudes. It is shown in equation (3) that dissipation due to ion drag is proportional to Pederson conductivity and inversely proportional to neutral density. An increase in electron density in the low latitude ionosphere can lead to an increase in Pederson conductivity. However, dissipation due to TEC enhancement may be offset by an increase in neutral density. Although we did not observe neutral density in the present study, an increase in thermospheric neutral density during storm events has been observed previously, due to heating of the auroral atmosphere and its meridional circulation [Forbes et al., 2005; Liu and Lühr, 2005; Sutton et al., 2005]. Indeed, the CHAMP satellite observed an average increase of 75% in thermospheric neutral density during moderate storms [Lei et al., 2010]. Therefore, based on equation (1), there is no significant increase in dissipation due to ion drag at low latitudes. [31] Uplift of the ionosphere at low latitudes is another factor that results in a reduced amplitude of TEC perturbation. Figure 7b shows variations in the average F 2 peak heights (h m F 2 ) around the time of China s dusk TID (the solid line) and night event (the dashed line). For comparison, 9of11

10 we also show the value of h m F 2 during a quiet time. For the time of the two events, a latitudinal uplift of more than 40 km is apparent from Mohe (52 N) to Sanya (18.3 N). At all ionosonde stations, the storm effect causes a peak height uplift of 70 km at dusk and 20 km at midnight, relative to quiet time. The simulation work of Francis [1973] revealed the energy profiles of several gravity wave ducted modes (G modes), whereas for a given wave period 1.75 hours, the peaks of energy range between 150 km and 250 km altitude. The energy of gravity wave decreases quickly above this height range, due to the increase of dissipation with height. Hence, following the uplift of h m F 2 which is high above 250 km altitude at low latitudes, the perturbation in TEC would show small amplitudes. 5. Summary [32] This paper reports the first results of the 2D imaging of large-scale TIDs using GPS network data from China. We collected TEC data from 231 GPS stations in the Crustal Movement Observation Network of China (CMONOC) and IGS. Using these data, we developed TEC perturbation maps and observed the band-like structures of LSTIDs during a medium-scale storm on May 28, The GPS observations were combined with data from 4 DPS-4 ionosondes located at Mohe (geographic latitude: 52 N), Beijing (40.4 N), Wuhan (30.5 N), and Sanya (18.3 N). For comparison, we also constructed 2D TEC perturbation maps for North America. [33] The medium-scale geomagnetic storm of May 28, 2011 was accompanied by an intense substorm. At substorm onset the local time in North America was after midnight, and it was afternoon in China. A slow-speed LSTID occurred immediately in North America around the time of onset. The maximum amplitude, phase velocity, and azimuth (clockwise from north) of the LSTID were TECU, m/s, and , respectively. Then, 1.5 hours after substorm onset, a LSTID reached China at dusk (local time), which was recorded between 1000 and 1030 UT ( local time), with a maximum amplitude, phase velocity, and azimuth (clockwise from north) of TECU, m/s, and 198 6, respectively. Finally, 6.6 hours after substorm onset, and before midnight in China, the second LSTID was recorded. This nighttime event in China occurred between 1510 and 1600 UT ( local time), with a maximum amplitude, phase velocity, and azimuth of TECU, m/s, and The latter two LSTIDs were observed simultaneously by the ionosonde chain in China, and a downward vertical phase velocity of 75 m/s for the dusk event and 60 m/s for the night event can be estimated through ionosonde observation. The phase fronts of these LSTIDs, with a front width of at least 1600 km, moved from northeast to southwest over a distance of more than 1000 km. As the events in both regions were observed at similar geographic latitudes, and over a time period of several hours, they cannot have originated from the same source. [34] Although the nighttime LSTID travelled farther south than the earlier dusk event, both disappeared in South China, and this was due to increase of the attenuation at low latitudes. According to the energy dissipation equation of atmospheric gravity waves [Richmond, 1978b], there is severe dissipation due to viscosity and heat conductivity at low latitudes, since such dissipation increases strongly with time; dissipation due to ion drag is less important but cannot be ignored because of enhancement in background TEC; In addition, uplift of the ionosphere at low latitudes is another factor that results in a reduced amplitude of TEC perturbation at low latitudes. [35] Acknowledgments. We acknowledge the Scripps Orbit and Permanent Array Center (SOPAC) and IGS for providing GPS network data via the Internet. The data of Global Ionospheric Maps used in this study were acquired as part of NASA s Earth Science Data Systems and archived and distributed by the Crustal Dynamics Data Information System (CDDIS). This work was supported by the Chinese Academy of Sciences (KZZD- EW-01-2), the National Natural Science Foundation of China (grants , ), and the National Key Basic Research Program of China (2011CB811405). [36] Robert Lysak thanks the reviewers for their assistance in evaluating this paper. References Afraimovich, E. L., et al. (2000), Observation of large-scale traveling ionospheric disturbances of auroral origin by global GPS networks, Earth Planets Space, 52, Beach, T., M. Kelley, P. Kintner, and C. Miller (1997), Total electron content variations due to nonclassical traveling ionospheric disturbances: Theory and Global Positioning System observations, J. Geophys. Res., 102(A4), , doi: /96ja Becker-Guedes, F., et al. (2007), The ionospheric response in the Brazilian sector during the super geomagnetic storm on 20 November 2003, Ann. Geophys., 25, , doi: /angeo Borries, C., N. Jakowski, and V. Wilken (2009), Storm induced large scale TIDs observed in GPS derived TEC, Ann. Geophys., 27, , doi: /angeo Boška, J. P., and P. Šauli (2001), Observations of gravity waves of meteorological origin in the F-region ionosphere, Phys. Chem. Earth, 26, Bruinsma, S., and J. M. Forbes (2009), Properties of traveling atmospheric disturbances (TADs) inferred from CHAMP accelerometer observations, Adv. Space Res., 43, , doi: /j.asr Buss, S., et al. (2004), Analysis of a jet stream induced gravity wave associated with an observed stratospheric ice cloud over Greenland, Atmos. Chem. Phys., 4, , doi: /acp Cai, H. T., et al. (2011), Observations of AGW/TID propagation across the polar cap: A case study, Ann. Geophys., 29, , doi: / angeo Chimonas, G. (1970), The equatorial electrojet as a source of long period traveling ionospheric disturbances, Planet. Space Sci., 18(4), , doi: / (70) Ding, F., W. Wan, B. Ning, and M. Wang (2007), Large scale traveling ionospheric disturbances observed by GPS TEC during the magnetic storm of October 29 30, 2003, J. Geophys. Res., 112, A06309, doi: /2006ja Ding, F., W. Wan, L. Liu, E. L. Afraimovich, S. V. Voeykov, and N. P. Perevalova (2008), A statistical study of large-scale traveling ionospheric disturbances observed by GPS TEC during major magnetic storms over the years , J. Geophys. Res., 113, A00A01, doi: /2008ja Fagundes, P. R., V. Klausner, Y. Sahai, V. G. Pillat, F. Becker-Guedes, F. C. P. Bertoni, M. J. A. Bolzan, and J. R. Abalde (2007), Observations of daytime F2-layer stratification under the southern crest of the equatorial ionization anomaly region, J. Geophys. Res., 112, A04302, doi: / 2006JA Fejer, B. G., J. W. Jensen, T. Kikuchi, M. A. Abdu, and J. L. Chau (2007), Equatorial ionospheric electric fields during the November 2004 magnetic storm, J. Geophys. Res., 112, A10304, doi: /2007ja Forbes, J. M., G. Lu, S. Bruinsma, S. Nerem, and X. Zhang (2005), Thermosphere density variations due to the April 2002 solar events from CHAMP/STAR accelerometer measurements, J. Geophys. Res., 110, A12S27, doi: /2004ja Francis, S. H. (1973), Acoustic gravity modes and large scale traveling ionospheric disturbances of a realistic dissipative atmosphere, J. Geophys. Res., 78, , doi: /ja078i013p Francis, S. H. (1975), Global propagation of atmospheric gravity waves: A review, J. Atmos. Terr. Phys., 37, , doi: / (75) of 11

11 Georges, T. M., and W. H. Hooke (1970), Wave-induced fluctuations in ionospheric electron content: A model indicating some observational bias, J. Geophys. Res., 75, , doi: /ja075i031p Hajkowicz, L. A. (1990), A global study of large-scale traveling ionospheric disturbances (TIDs) following a step like onset of auroral substorms in both hemispheres, Planet. Space Sci., 38, , doi: / (90)90058-x. Hines, C. O. (1960), Internal atmospheric gravity waves at ionospheric heights, Can. J. Phys., 38, , doi: /p Hunsucker, R. D. (1982), Atmospheric gravity waves generated in the high-latitude ionosphere: A review, Rev. Geophys., 20(2), , doi: /rg020i002p Jacobson, A. R., R. C. Carlos, R. S. Massey, and G. Wu (1995), Observation of traveling ionospheric disturbances with a satellite-beacon radio interferometer: Seasonal and local time behavior, J. Geophys. Res., 100(A2), , doi: /94ja Jakowski, N., C. Borries, and V. Wilken (2012), Introducing a disturbance ionosphere index, Radio Sci., 47, RS0L14, doi: /2011rs Kamide, Y., N. Yokoyama, W. Gonzalez, B. T. Tsurutani, I. A. Daglis, A. Brekke, and S. Masuda (1998), Two-step development of geomagnetic storms, J. Geophys. Res., 103(A4), , doi: /97ja Kelley, M. C. (1989), The Earth s Ionosphere: Plasma Physics and Electrodynamics, Academic, San Diego, Calif. Klausner, V., P. R. Fagundes, Y. Sahai, C. M. Wrasse, V. G. Pillat, and F. Becker-Guedes (2009), Observations of GW/TID oscillations in the F2 layer at low latitude during high and low solar activity, geomagnetic quiet and disturbed periods, J. Geophys. Res., 114, A02313, doi: / 2008JA Lei, J., J. P. Thayer, A. G. Burns, G. Lu, and Y. Deng (2010), Wind and temperature effects on thermosphere mass density response to the November 2004 geomagnetic storm, J. Geophys. Res., 115, A05303, doi: /2009ja Li, G., B. Ning, W. Wan, and B. Zhao (2006), Observations of GPS ionospheric scintillations over Wuhan during geomagnetic storms, Ann. Geophys., 24, , doi: /angeo Li, G., B. Ning, B. Zhao, L. Liu, W. Wan, F. Ding, J. S. Xu, J. Y. Liu, and K. Yumoto (2009), Characterizing the 10 November 2004 storm-time middle-latitude plasma bubble event in Southeast Asia using multi-instrument observations, J. Geophys. Res., 114, A07304, doi: /2009ja Liu, H., and H. Lühr (2005), Strong disturbance of the upper thermospheric density due to magnetic storms: CHAMP observations, J. Geophys. Res., 110, A09S29, doi: /2004ja Maeda, S. (1985), Numerical solutions of the coupled equations for acousticgravity waves in the upper thermosphere, J. Atmos. Terr. Phys., 47, , doi: / (85) Mao, T., W. Wan, X. Yue, L. Sun, B. Zhao, and J. Guo (2008), An empirical orthogonal function model of total electron content over China, Radio Sci., 43, RS2009, doi: /2007rs Mayr, H. G., I. Harris, F. A. Herrero, N. W. Spencer, F. Varosi, and W. D. Pesnell (1990), Thermospheric gravity waves: Observations and interpretation using the transfer function model (TFM), Space Sci. Rev., 54, , doi: /bf Nicolls, M. J., M. C. Kelley, A. J. Coster, S. A. González, and J. J. Makela (2004), Imaging the structure of a large-scale TID using ISR and TEC data, Geophys. Res. Lett., 31, L09812, doi: /2004gl Noll, C. (2012), The Crustal Dynamics Data Information System: A resource to support scientific analysis using space geodesy, Adv. Space Res., 46, , doi: /j.asr Perevalova, N. P., E. L. Afraimovich, S. V. Voeykov, and I. V. Zhivetiev (2008), Parameters of large-scale TEC disturbances during the strong magnetic storm on 29 October 2003, J. Geophys. Res., 113, A00A13, doi: /2008ja Rice, D. D., R. D. Hunsucker, L. J. Lanzerotti, G. Crowley, P. J. S. Williams, J. D. Craven, and L. Frank (1988), An observation of atmospheric gravity wave cause and effect during the October 1985 WAGS campaign, Radio Sci., 23, , doi: /rs023i006p Richmond, A. D. (1978a), The nature of gravity wave ducting in the thermosphere, J. Geophys. Res., 83(A4), , doi: / JA083iA04p Richmond, A. D. (1978b), Gravity wave generation, propagation, and dissipation in the thermosphere, J. Geophys. Res., 83, , doi: /ja083ia09p Sahai, Y., et al. (2009), Effects observed in the Latin American sector ionospheric F region during the intense geomagnetic disturbances in the early part of November 2004, J. Geophys. Res., 114, A00A19, doi: / 2007JA Saito, A., S. Fukao, and S. Mayazaki (1998), High resolution mapping of TEC perturbations with the GSI GPS network over Japan, Geophys. Res. Lett., 25, , doi: /98gl Saito, A., et al. (2001), Traveling ionospheric disturbances detected in the FRONT campaign, Geophys. Res. Lett., 28, , doi: / 2000GL Schunk, R. W., and A. F. Nagy (2000), Ionospheres: Physics, Plasma Physics, and Chemistry, 274 pp., Cambridge Univ. Press, New York, doi: /cbo Shiokawa, K., et al. (2002), Imaging observations of the equatorward limit of midlatitude traveling ionospheric disturbances, Earth Planets Space, 54, Sutton, E. K., J. M. Forbes, and R. S. Nerem (2005), Global thermospheric neutral density and wind response to the severe 2003 geomagnetic storms from CHAMP accelerometer data, J. Geophys. Res., 110, A09S40, doi: /2004ja Tsugawa, T., A. Saito, Y. Otsuka, and M. Yamamoto (2003), Damping of large-scale traveling ionospheric disturbances detected with GPS networks during the geomagnetic storm, J. Geophys. Res., 108(A3), 1127, doi: /2002ja Tsugawa, T., A. Saito, and Y. Otsuka (2004), A statistical study of largescale traveling ionospheric disturbances using the GPS network in Japan, J. Geophys. Res., 109, A06302, doi: /2003ja Tsugawa, T., K. Shiokawa, Y. Otsuka, T. Ogawa, A. Saito, and M. Nishioka (2006), Geomagnetic conjugate observations of large-scale traveling ionospheric disturbances using GPS networks in Japan and Australia, J. Geophys. Res., 111, A02302, doi: /2005ja Vadas, S. L., and H. Liu (2009), Generation of large-scale gravity waves and neutral winds in the thermosphere from the dissipation of convectively generated gravity waves, J. Geophys. Res., 114, A10310, doi: /2009ja Wan, W. X., H. Yuan, B. Ning, and J. Liang (1998), Travelling Ionosphere Disturbances associated with tropospheric vortexes around Qinghai-Tibet Plateau, Geophys. Res. Lett., 25, , doi: /1998gl Williams, P. J. S., et al. (1988), The generation and propagation of atmospheric gravity waves observed during the Worldwide Acoustic-Gravity Wave Study (WAGS), J. Atmos. Terr. Phys., 50, , doi: / (88) of 11