Ionospheric solar flare effects monitored by the ground-based GPS receivers: Theory and observation

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109,, doi: /2003ja009931, 2004 Ionospheric solar flare effects monitored by the ground-based GPS receivers: Theory and observation J. Y. Liu 1 and C. H. Lin 2 Institute of Space Science, National Central University, Jungli City, Taoyuan, Taiwan H. F. Tsai 3 Radio Science Institute for Space and Atmosphere, Kyoto University, Uji, Japan Y. A. Liou 4 Center for Space and Remote Sensing Research, National Central University, Jungli City, Taoyuan, Taiwan Received 11 March 2003; revised 4 August 2003; accepted 7 October 2003; published 24 January [1] The ionosphere responses to a solar flare observed by using ground-based receivers of the global positioning system (GPS) are investigated in this paper. Two quantities, the total electron content (TEC) and its time rate of change (rtec), can be derived from the receivers. The theoretical studies show that the rtec is related to the frequency deviation of the GPS signals. Meanwhile, worldwide ground-based GPS receivers are employed to derive the TEC and associated rtec to monitor the ionospheric solar flare effect on 14 July (Bastille Day) It is found that ionospheric solar flare effects can be observed from predawn to postdusk regions, and the most pronounced signatures appear in the midday area. The agreement between theoretical predications and observations demonstrates that the TEC is suitable to monitor the overall variations of flare radiations while the rtec is capable to detect sudden changes in the flare radiations. INDEX TERMS: 2435 Ionosphere: Ionospheric disturbances; 2479 Ionosphere: Solar radiation and cosmic ray effects; 7519 Solar Physics, Astrophysics, and Astronomy: Flares; 2423 Ionosphere: Ionization mechanisms; 2494 Ionosphere: Instruments and techniques; KEYWORDS: ionosphere, solar flare, GPS, TEC, Bastille Day Citation: Liu, J. Y., C. H. Lin, H. F. Tsai, and Y. A. Liou (2004), Ionospheric solar flare effects monitored by the ground-based GPS receivers: Theory and observation, J. Geophys. Res., 109,, doi: /2003ja Introduction [2] A solar flare is a sudden brightening in an active region usually near a complex group of sunspots of the photosphere, which produces immediate increases in the ionospheric ionization of varying degrees at different heights, together called the Sudden Ionospheric Disturbances (SIDs) or the ionospheric solar flare effects [Dellinger, 1937]. The disturbances have important effects on radio communications and navigations over the entire radio spectrum [Davies, 1990]. Davies [1990] reviewed that SIDs were generally recorded as the short wave fadeout [Stonehocker, 1970], sudden phase anomaly [Jones, 1971; Ohshio, 1971] sudden frequency deviation (or frequency shift; Doppler shifts) [Donnelley, 1971; Liu et al., 1996a], sudden cosmic noise absorption [Deshpande and Mitra, 1 Also at Center for Space and Remote Sensing Research, National Central University, Jungli City, Taoyuan, Taiwan. 2 Also at High Altitude Observatory, National Center for Atmospheric Research, Boulder, Colorado, USA. 3 Now at National Space Program Office, Hsinchu, Taiwan. 4 Also at Institute of Space Science, National Central University, Jungli City, Taoyuan, Taiwan. Copyright 2004 by the American Geophysical Union /04/2003JA009931$ ], sudden enhancement/decrease of atmospherics [Sao et al., 1970], and sudden increase in total electron content (TEC) [Mendillo et al., 1974; Davies, 1980]. Meanwhile, Ohshio [1964] studied geomagnetic field strengths response to solar flares, which are termed the geomagnetic solar flare effects, by global ground-based magnetometers. Liu et al. [1996a] further estimated ionospheric electron density changes at about 90 km altitude by examining simultaneous measurements of ground-based geomagnetic field strengths and space based flare X-ray radiations during solar flares. [3] To simultaneously monitor a large area of the ionosphere, the global positioning system (GPS) is ideal to be employed. The system consists of more than 24 satellites, distributed in six orbital planes around the globe at an altitude of about 20,200 km. Each satellite transmits two frequencies of signals ( f 1 = MHz and f 2 = MHz). Since the ionosphere is a dispersive medium, scientists are able to evaluate the ionospheric effects with measurements of the modulations on carrier phases and phase codes recorded by dual-frequency receivers [Sardón et al., 1994; Leick, 1995; Liu et al., 1996b]. From recorded broadcast ephemeris and given subionospheric height, the slant TEC along the ray path can be converted into the vertical TEC at its associated longitude and latitude [cf., Tsai and Liu, 1999]. 1of12

2 [4] In this paper, in addition to the vertical TEC derived from ground-based GPS receivers, we introduce the time rate of its change (rtec) as a new quantity to simultaneously monitor ionospheric responses to solar flares. The TEC and rtec derived from worldwide GPS receivers are employed to monitor the ionosphere response to a large solar flare event on 14 July (Bastille Day) 2000 and some other smaller events (see Appendix A). Finally, the flare signatures of the two quantities in dawn, daytime (midday), dusk, and nighttime (midnight) regions are examined and discussed. 2. Theory and Model [5] In this section, we not only examine the physical presentations and meanings of the GPS TEC and rtec but also develop links between the two quantities and previous observations. In the ionosphere, the rates of change of the electron density N can be expressed by the continuity equation [Davies, 1990, ¼ Q L rðnvþ; ð1þ where Q and L represent the rates of the electron production and loss of the photochemical processes, v is the electron velocity, and therefore the divergent term is due to the transport. The rate of production is mainly a function of the solar X-ray and EUV radiations [Ratcliffe, 1972]. The rate of loss is determined by the recombination constant, which can be generally given by L ¼ an 2 þ bn; where a and b are the recombination constants in the ionospheric E and F 1 regions, respectively. Since the photochemical process is much faster than the transport during the occurrence of a solar flare, the divergent term in equation (1) can be neglected and the change of the electron density at a certain altitude can be expressed as N ¼ Z t t 0 ð2þ ðq LÞdt; ð3þ where t 0 denote a certain time before the solar flare. It can be seen that equation (2) is a second-order polynomial, which results in the integration of equation (3) being nonlinear and very complex [Liu et al., 1996a]. [6] Although the change of electron density in equation (3) is the most direct index showing the ionospheric solar flare effect, very limited observation instruments can be routinely employed and operated [Mitra, 1974]. For instance, incoherent scatter soundings are suitable to monitor both lower and upper ionospheric electron density variations, however, most of these observations are not continuously operated. Thus three observations, magnetic field fluctuations, total electron content (TEC) changes, and radio wave frequency shifts, have often been used to evaluate ionospheric electron density variations during solar flares. For the lower ionosphere, ground-based magnetometer measurements have been employed to estimate the N at about 90 km altitude [Liu et al., 1996a]. However, this estimation often suffers from other geophysical disturbances [Ohshio, 1964; Matsushita and Campbell, 1967; Liu et al., 1996a]. On the other hand, TEC derived from ground-based satellite receivers have been used to estimate the upper and/ or integrated ionospheric N changes during soar flares. Owing to limited satellites and their receivers available, in early years the most common technique to study the ionospheric solar flare effects is to examine Doppler shift f in signals transmitted by Doppler sounding systems. On the basis of a collisionless Appleton formula [e.g., Budden, 1985] and neglecting the contribution of the geomagnetic field, and dropping the transport term, during the flare occurrence the shift f can be approximated by [cf., Liu et al., 1996a] f ¼ f Z Rx cm Q L Þds; ð4þ where Tx and Rx denote the transmitter and receiver antennas, f is the transmitted radio wave frequency, c is light speed in free space, m is the refractive index in the ionosphere, and s denotes the integration along the radio wave path. Note from equations (1) and (4) that the f is proportional to the time rate of change of electron density N/t. Owing to the high frequency (HF) used, the Doppler sounding system observation, however, suffers from the short wave fadeout, and often no data can be recorded even during the midway of the flare occurrence [e.g., Davies, 1990]. [7] The transmitted frequencies in UHF (f 1 = MHz and f 2 = MHz) are much greater than the ionospheric collision frequencies, and therefore the ionospheric absorption (signal fadeout) effects for the GPS signals are minor. Note that if let Tx be the satellite onboard transmitting antenna, equation (4) can be fully adopted by the GPS observation. Thus, scientists can use the TEC and f obtained from ground-based GPS receivers to continuously monitor the ionosphere response to solar flares. To further understand the two quantities, we examine their physical meanings. [8] The TEC between a GPS satellite (Tx) and a receiver (Rx) can be expressed as STEC ¼ Z Rx Tx Nds; ð5aþ where s denotes the integration path along the ray, and therefore the TEC change STEC during a solar flare event is given by STEC ¼ Z Rx Sat Nds: ð5bþ For practical data reduction, a simple way detecting the vertical TEC change during a solar flare is to use STEC(t 0 ), which is observed slightly before the flare onset at time t 0, as a reference to offset its after. It thus can be written as TECðÞ¼ t ½STECðÞ STEC t ðt 0 ÞŠM ð6aþ where M = h/s is the projection factor or mapping function, and h and s are the altitude of the ionospheric 2of12

3 Figure 1. Locations of the subsolar point (star symbol) and 60 GPS receivers (solid triangle). point and distance between the ionospheric point and the ground-based GPS receiver, respectively. Since the Earth s rotation and GPS satellite orbit are one sidereal day (1 sd = 23 hours 56 min) and one-half sidereal day (1/2 sd), respectively, Hernandez-Pajares et al. [1997] developed the high-resolution TEC monitoring method to derive the difference in TEC along the same geometric path. Thus an alternative way obtaining the vertical TEC change can be written as TECðÞ¼ t ½STECðÞ STEC t ðt þ k sdþšm: ð6bþ where k is an integer. [9] Liu et al. [1996a] study the ionospheric solar flare effects observed by a Doppler sounding system and find that the change rate of the flare radiations dramatically affects the ionospheric frequency deviation (i.e., Doppler shift) f. For the GPS signals, the Doppler shift f is made up of two parts: (1) a part due to the motion of the satellite with respect to the receiver, and (2) a part due to the rate of change of the total electron content dtec/dt (or rtec) along the path, which can be expressed as (for detail see Davies [1990]) v l f ¼ f m s c þ 40:3 dtec ffi 40:3 dtec ffi 40:3 TEC cf dt cf dt cf t ¼ 40:3 rtec; cf i.e., ð7aþ cf f rtec ffi 40:3 ; ð7bþ where m s is the refractive index at the satellite and v l is the line-of-sight component of the GPS satellite velocity. With a relatively slow and constant speed of the GPS satellite, the first term in equation (7a) is about a small constant and a sudden Doppler shift is mainly caused by temporal changes in TEC. Since c and f are constant, the rtec and f should be well correlated to each other. Meanwhile, for data processing, the rtec has been defined by subtracting each TEC (or TEC in equation (6a) or (6b)) from its previous 30-s value, which is a simple 2-point differentiation. The simple 2-point differentiation meets physical sprit of equation (7a), which builds a connection (physical equivalence) between current rtec study and previous f works (see the papers listed in the works of Mitra [1974] and Davies [1990]). Based on equations (6) and (7), the TEC monitors the overall time variations of the flare X ray and EUV radiations (or integrated ionospheric electron density) and the rtec instantaneously registers their time rate changes (or sudden changes). 3. Observation [10] The solar flare originated near the center of the solar disk, and its brightness started at 1003 UT, peaked at 1024 UT, and ended at 1043 UT on 14 July (the Bastille Day) The X57 flare has been categorized as an X-class flare, a classification reserved for the most powerful flares. Sixty ground-based receivers of the international GPS service (IGS) are subdivided into four tracking networks to globally observe the ionospheric TEC variations in the dawn ( 4 42 N, E), daytime (17 62 N, E), dusk (14 36 N, E), and nighttime ( N, E) regions (Figure 1). The subsolar point of the Earth s ionosphere (denoted by the star symbol) is located at about (23.5 N, 22.5 E geographic). Note that all the GPS quantities TEC (or TEC) and rtec in this study have been properly converted into their vertical component and location [cf. Liu et al., 1996b; Tsai and Liu, 1999]. Since no time series data of EUV radiations are available, solar X-ray radiations in the 1-min time resolution recorded by the geosynchronous operational environmental satellite GOES-10 are examined. Figure 2a illustrates the solar X-ray flux intensities in 1 8Åon 13 and 14 July 2000, the one on the Bastille Day reaches a maximum at 1024 UT. To visualize the global flare responses, Figures 2b and 2c display the sum of the all 3of12

4 Figure 2. The X-radiations 1 8Årecorded by the GOES-10 on 13 (the reference day), and 14 (the Bastille Day) July 2000 (a), and the sum of the TEC (b), and rtec (c) on each day and their differences from the sixty GPS receivers. recorded (about 5 10 satellites 60 receivers) total electron content TEC(7/13) on 13 and TEC(7/14) on 14 July 2000, and the difference between the two days, TEC(7/ 14 7/13) as well as the associated rtec, respectively (1 TECu = el m 2 ). Recall that the TEC is deduced from GPS signals recorded by a receiver every 30 s, while the rtec is obtained by subtracting each TEC from its previous value, i.e., a simple 2-point differentiation. It can be seen that during UT the flare signatures in the TEC(7/14) and TEC(7/14 713) are slightly different in shape but generally have similar tendencies which reach their maxima around 1029 UT. Note that the flare signatures in the two TEC measurements are relatively small and somewhat difficult to be identified. By contrast, the associated rtec(7/14) and rtec(7/14 713) yield nearly identical and rather clear spike flare signatures at about UT. Thus to avoid unwanted features from 13 July, we simply use rtec(7/14) to monitor the ionosphere response to the solar flare. [11] To understand the solar flare signatures in TEC in further detail, we investigate the vertical TEC(7/14) and TEC(7/14 7/13) (defined by equations (6a) and (6b), respectively) observed by the dayside receivers from five GPS satellites, PRN 11, 15, 20, 29, and 31. Figure 3 displays traces of the ionospheric points of the five GPS satellites observed by the dayside receivers. To remove the background contributions, each vertical TEC has been offset at 1000 UT (see equation (6a) and let t 0 = 1000 UT). For example, each high resolution TEC(7/14 7/13) is obtained by offsetting TEC(7/14) and TEC(7/13) first, and then carry out the one-to-one subtraction. Figures 4 and 5 illustrate the TEC(7/14) and TEC(7/14-7/13) together with their associated averages over all the receivers from the five GPS satellites during UT, respectively. It is found that the averaged TEC(7/14) of satellite PRN 20 and 29 (PRN 11, 15, and 31) reach their maximum (saturated or flatting) values at about 1029 UT (Figure 4), while almost all the averaged TEC(7/14-7/13) of the five satellites yield clear maximum features at about 1029 UT. Since the five curves yield similar and consist features, TEC(7/14-7/13) obtained by applying the high resolution method of Hernandez-Pajares et al. [1997] has a better chance than the simple offset TEC(7/14) to remove background contributions for monitoring flare features in the vertical GPS TEC variations. [12] Figure 6 presents the solar X-ray flux intensities, and the averaged TEC(7/14-7/13) and rtec(7/14) per satellite and per receiver (for simplicity, TEC and rtec, hereafter) of the four regions during UT. The most pronounced solar flare effects in TEC and rtec appear in the daytime region but no signature in the nighttime region. In the dawn, daytime, and dusk regions, TECs yield only one peak at about 1029 UT, while each associated rtec reveals three spikes at 1019, 1024, and 1027 UT, respectively. Variations in the daytime TEC are similar to those of the solar X-ray flux intensity, while the dawn and dusk TECs yield a ledge and a maximum at about 1029 UT and have increase and decrease trends, respectively. Based on equations (5) and (7), the rtec value is related to the associated TEC/t (or integrated N/t). Thus due to the increase and decrease of the flare radiations, the daytime rtecs yield positive and negative values before and after 4of12

5 Figure 3. The dayside satellite traces of PRN 15, 29, 31, 11, and 20 during UT on the reference day or the Bastille Day. The circles show the ionospheric points of satellite-receiver ray paths at 1000 UT UT, respectively. Similarly, gradual increases and decreases of the solar radiations during the dawn and dusk period result in rtec values being positive and negative during UT, respectively. Note that slow and small oscillations near zero values of both TEC and rtec indicate no solar (flare) radiation contribution in the nighttime region. Nevertheless, it is obvious that the very pronounced flare features in the daytime rtec appear at 1019, 1024, and 1027 UT. [13] The global distributions of the GPS receivers allow us to further examine the TEC and rtec as well as their associated percent changes at various latitudes, longitudes, local time, and zenith angles. The percent changes of the two quantities are defined as TEC/TECo and rtec/tect, where TECo and TECt denote the TEC value observed at 1000 UT before the solar flare occurrence and the instant TEC value when rtec is derived, respectively. The comparison between the two quantities and their associated percent changes allows us to further understand the contributions from the ionospheric ambient (or background) condition. Figures 7a, 7b, and 7c (7d, 7e, and 7f) display TEC and (TEC/TECo) distributions at 1029 UT at various latitudes/longitudes hour angles, and zenith angles, respectively. Figure 8 illustrates the same distributions of the rtec and associated rtec/tect at 1024 UT. Note that TEC and rtec reach their greatest values at 1029 and 1024 UT, respectively. Figures 7a and 7d reveal that the signatures in TEC and TEC/TECo are rather complex. Figures 7b and 7e yield a similar feature that the noontime, dawn and dusk TEC yield pronounced flare signatures. Although the signatures have the greatest values in the noontime region, we find no clear and simple relationship between them and the hour or zenith angle (see Figures 7b and 7e or Figures 7c and 7f). The complex relationships imply that the ionospheric background electron density and/ or other geomagnetic variabilities can heavily and easily disturb the flare TEC observations. In contrast, Figure 8 displays that the solar flare effects can be seen even in the predawn (0400 LT) and postdusk (2000 LT) regions and the most pronounced feature appears near the subsolar point (1200 LT), which is in the midday area (Figure 8a). Moreover, it is interesting to find that the rtecs flare signatures are symmetry to the hour angle (Figure 8b) while their percent changes yield the greatest values during predawn (0500 LT) (Figure 8e). The quasi-cosine relations shown in Figures 8b and 8c confirm that the rtec flare signatures are functions of the solar hour angle and zenith angle. The comparisons between Figures 8b 8c and 8e 8f 5of12

6 Figure 4. The offset TECs and the associated averages of the GPS satellite PRN 11, 15, 20, 29, and 31 to the dayside receivers on the Bastille Day. show that it is the rtec but not its percentages to be functions of hour angles and zenith angles. 4. Discussion and Conclusion [14] Numerous techniques have been employed to monitor the ionospheric solar flare effects (see the papers listed in the works of Mitra [1974] and Davies [1990]). However, most of the techniques simply observe flare features at certain altitudes. For instance, magnetic filed strengths recorded by ground-based magnetometers for studying the geomagnetic solar flare effect [Ohshio, 1964] can be employed to evaluate ionospheric electron density changes at about 90 km, and frequency shifts in radio signals probed by HF Doppler sounding systems for studying the ionospheric solar flare effects [Donnelley, 1971] can be used to examine the time rate of changes of ionospheric electron densities at certain altitudes below the F-peak [Liu et al., 1996a]. In this paper based on equations (5a) and (7a), we introduce the TEC and rtec (or TEC) derived from measurements of the ground-based GPS receivers to simultaneously investigate variations of electron densities and the associated time rate of their changes in the whole ionosphere ranging from 90 to 20,200 km altitude during occurrences of solar flares. [15] Liu et al. [1996a] found that the temporal variations of the ionospheric electron density N derived from ground based magnetometer data and those of the solar X-ray radiations yield similar tendencies. Equations (3) and (5a) 6of12

7 Figure 5. The high-resolution TECs and the associated averages of the GPS satellite PRN 11, 15, 20, 29, and 31 to the dayside receivers derived with the Bastille Day and the reference day. also show that the variations of the TEC are proportional to those of the electron production rate of the flare radiations. Based on Liu et al. [1996a] and the theoretical derivations of this paper, we conjecture that the temporal variations of TEC and those of the flare radiations yield similar tendencies. Similar tendencies in the daytime TEC and the flare X-ray radiations shown in Figure 6 (for more examples also see Appendix A) confirm that TEC is suitable to monitor the overall temporal variations of the solar flare radiations. [16] Liu et al. [1996a] demonstrate that not only the magnitude but also the time rate changes of flare radiations affect the frequency shifts f observed by Doppler sounding systems. The theoretical derivation of equation (7) shows the rtec and f derived from ground based GPS receivers to be nearly proportional. Therefore we expect to observe rtec spike features appearing when the flare radiations sharply (or suddenly) increase. Figure 6 illustrates three rtec maxima appearing around 1019, 1024, and 1027 UT but one maximum in the flare radiations at 1027 UT. To search the moments of sharp increases, we differentiate the X-ray 1 8 Å radiations of the GOES-10 shown in Figure 6a. Figure 9 illustrates that the rtec and the differentiated X-ray radiations simultaneously appear at 1019 UT. Recently, Masuda et al. [2000] report that the hard X-ray observation of the Yohkoh satellite starts the early 7of12

8 Figure 6. The solar 1 8 Å X-ray radiations (a), as well as the averaged high-resolution TECs (b), and their associated rtecs (c) in the dawn, daytime, dusk, and nighttime regions observed during UT on the Bastille Day. phase around UT, which has some of the impulsive phase at about 1019 UT, yields sudden increase at 1024 UT, reaches its peak at 1027 UT, and finally goes back to the previous intensity level at 1030 UT. The three sudden changes in the hard X-ray shown by Masuda et al. [2000] coinciding with the three maxima of the rtec observed in this paper at 1019, 1024, and 1027 UT (see Figure 6c) confirms that the rtec variation is highly sensitive to the change rate of Figure 7. The spatial, solar hour angle, and zenith angle distributions of the high-resolution TECs and their percent changes at 1029 UT derived with the Bastille Day and the reference day. 8of12

9 Figure 8. The spatial, solar hour angle, and zenith angle distributions of the rtecs and their percent changes at 1024 UT on the Bastille Day. the solar flare radiations. It can be found that there are no obvious X-ray radiation increases when the rtec maxima appear at 1024 and 1027 UT (Figure 9). We consider that the discrepancies in the sharp increases between the GOES-10 and Yohkoh X-ray radiations might result from that the flare X-ray flux has different temporal features depending on wavelengths. Although no EUV data is available and presented in this study, it has been well known by ionospheric scientists that not only the solar X-ray but also EUV radiations are responsible to the ionospheric ionizations [see, e.g., Ratcliffe, 1972]. Therefore the rtec can be used to detect the sudden increases in the X-ray and EUV flare radiations. [17] Figure 2 illustrates that the flare feature of the TEC (or TEC) is not as obvious as that of the rtec. It can be seen that many large wave-like fluctuations in the sum of TEC(7/13) and TEC(7/14) shown in Figure 2, which might be caused some other geophysical effects, such as geomagnetic storms, traveling ionospheric disturbances, atmospheric gravity waves, etc., result in the flare features in the TEC (or TEC) being relatively difficult to be observed. There are two types of the flare radiations, which are the sudden and the gradual. When the duration of a solar flare is relative long, variations of other geophysical effects start to contaminate and burry the TEC flare features (for more examples see Appendix A). Nevertheless, the theoretical derivations and observational results demonstrate the suitability and detectability of the TEC and rtec to be different. [18] To reduce the contaminations from other geophysical effects, the high resolution TEC monitoring method proposed by Hernandez-Pajares et al. [1997] are adopted and tested. Relatively observable features in TEC(7/14 7/13) (see Figures 2 and 5) indicate the high-resolution method to be a better way to monitor the ionospheric TEC response to the solar flare. Figure 4 reveals that TEC(7/14) of satellite PRN 20 and 29 reach maximum while those of PRN 11, 15, Figure 9. The differentiated solar 1 8ÅX-ray radiations and the sum of the rtecs during UT on the Bastille Day. 9of12

10 Figure A1. The flare radiations, the averaged rtec, and the averaged TEC of the three M solar flares occurred on 1/10, 3/8, and 4/ The three downward arrows in the third panel respectively denote the starting, max, and the ending times of the 1 8ÅX-ray flare radiations. and 31 yield saturated values at about 1029 UT. Note that the footprints (i.e., ionospheric point traces) of the satellites in the daytime region are around south of the ionospheric midlatitude trough, 52 N geomagnetic, where electron densities have minimum values [Ratcliffe, 1974]. Thus the ionospheric points of GPS satellites moving away from the trough latitudes result in the TEC increase. It is interesting to see that when the satellite PRN 20 and 29 move toward the trough, their TEC(7/14) differences between 1000 and 1029 UT are relatively small, and the two TECs reach maximum values at 1029 UT. Meanwhile, for the ionospheric points of satellites PRN 11, 15, and 31 moving away from the trough, their TEC(7/14) differences between 1000 and 1029 UT are relatively large and right after 1029UT tend to have saturated features. This indicates latitudinal (or ambient) effects in the TEC(7/14) to be significant. By contrast, the flare features in all five averaged TEC(7/14 7/13) shown in Figure 5 are rather similar and reveal maximum values at 1029 UT. This once again confirms that the high-resolution TEC monitoring method developed by Hernandez-Pajares et al. [1997] partially remove the ambient effect and is more suitable to observe solar flare TEC features. [19] The TEC observations show no obvious flare signature observed in the nighttime region, and clear flare signatures at dusk but relatively unclear at dawn. It might be interesting to find possible mechanisms causing the difference between the dawn and dusk regions. Figure 6b displays at dawn that the ambient TECs yield an increase tendency due to the sunrise ionization, which superimposes with the TECs increase due to the flare radiations around 1030 UT. The mixture of the two increases causes the flare features in the TECs (or TECs) to be difficultly identified at dawn. By contrast, at dusk, the ambient TEC has a decreasing trend while the TEC temporarily increase due to the flare radiations the TECs increase. The opposite variations in the TECs in practice enhance the visibility of flare signatures (see 10 of 12

11 Figure A2. The flare radiations, the averaged rtec, and the averaged TEC of the three X solar flares occurred on 4/21, 7/3, and 7/ Figure 6b). Nevertheless, the dawn and dusk results indicate day-to-day variations and background of the ionospheric electron densities can easily affect observations of the ionospheric solar effects in the GPS TECs. [20] Figure 7 shows the relationships between the associated percentage changes in the TEC and the geographic latitude/longitude, local time (or solar hour angle), and zenith angle are rather complex. It can be seen from Figures 2 and 6 Table A1. Parameters of the Six Flare Events Date Start, hhmm Maximum, hhmm End Flare Class Increase Rate, Watt/m 2 -min Duration, min 1/10, M /8, M /5, M /21, X /3, X /23, X that the duration of the flare radiation on the 2000 Bastille Day is rather long. During such a long period, many other geophysical viariabilities, such as the wave-like variations in TEC(7/13) and TEC(7/14) shown in Figure 2b, could affect the TEC flare observation (for more examples see Appendix A). By contrast, rise times between the start and the maximum of flare radiations usually are rather short, often less than 10 min (see Figures 2, 6, A1, and A2). Owing to a short time interval, other geophysical effects become less important, and therefore the rtec flare signatures are generally obvious. This explains that the rtec is a nice function Table A2. GPS Receivers Used for the M Class Flares Receiver Geographic Latitude Geographic Longitude BRUS N 4.36 E GOPE N E MATE N E WTZR N E 11 of 12

12 Table A3. GPS Receivers Used for the X Class Flares Receiver Geographic Latitude Geographic Longitude AUCK N E CHAT N E KSMV N E MIZU N E of the hour angle (or zenith angle) shown in Figures 8b and 8c. Long-term observations show that solar flare radiations generally have short rise times but rather long decay times [see, e.g., Davies, 1990]. Therefore it is the characteristics of flare radiations which result in that the rtec have better chances than the TEC to register flare signatures. [21] Scientists have already examined the ionospheric solar flare effect and published numerous papers in last 7 decades. The theoretical and observational results demonstrate that the TEC and rtec represent two different physical quantities, which are corresponding to the previous ground based magnetometer [Ohshio, 1964] and Doppler sounding system [Donnelley, 1971] observations, and can be employed to monitor overall time evolutions and to detect sudden increases of solar flare radiations. There are worldwide thousands of ground-based GPS sites, which provide an excellent chance to continuously monitor ionospheric solar flare effects at various local times as well as longitudes and latitudes. Appendix A: More Examples [22] In this appendix we show observations of other six solar flare events. Three M and three X class solar flares are arbitrarily chosen from year 2001 and Table A1 summarizes the time, class, time rate of flare radiations increase from the start to the maximum, and the duration from the start to the end of the 1 8ÅX-ray radiations from the GOES-10. It can be seen that the 4/5 and 4/21 yield the smallest increase rates and longest durations among their classes. [23] For simplicity, we analyze the TEC and rtec recorded by GPS receivers in the daytime region only. Tables A2 and A3 listed the locations of the ground based GPS receivers for the M and X class flares, respectively. [24] Figures A1 and A2 illustrate the flare radiations (top panel), the averaged rtec (middle panel) and the averaged TEC (bottom panel) of the three M and X classes, respectively. Table A1 shows that the 4/5 and 4/21 events yield the smallest increase rates and as well as the longest durations 77 and 115 minutes among the M and X classes, respectively. The right column in Figure A1 and the left column in Figure A2 demonstrate that due to the gradual (or small) increase rates in the flare radiations and the long flare durations, no obvious flare features in the TEC and rtec can be observed. For those with the obvious rtec features, the time rate of increase in Table A1 show that the 7/3 event are the greatest value, followed by the 7/23, 3/8, and 1/10 events. Through the sequence of the increase rates and that of the magnitude of the rtec, we discover in Figures A1 and A2 that sudden increases in flare radiations result in obvious flare features of the rtec. Meanwhile, we find for the obvious that the magnitude of the TEC flare features and the flare classes seem to be highly correlated. [25] Acknowledgments. Data used in this paper retrieve from IGS and Ministry of the Interior of Taiwan. This research was partially supported by the Ministry of Education grant 91-N-FA and the Office of Naval Research project N to the National Central University. The authors wish to thank A. D. Richmond at the High Altitude Observatory for useful comments and suggestions. The manuscript was originally submitted to Journal of Geophysical Research-Space Physics for publication in March 2001 (2001JA007519). [26] Lou-Chuang Lee thanks M. J. Keskinen and another reviewer for their assistance in evaluating this paper. References Budden, K. G. (1985), The Propagation of Radio Waves: The Theory of Radio Waves of Low Power in the Ionosphere and Magnetosphere, Cambridge Univ. Press, New York. Davies, K. (1980), Recent progress in satellite radio beacon studies with particular emphasis on the ATS-6 radio beacon experiment, Space Sci. Rev., 25, 357. Davies, K. (1990), Ionospheric Radio, 580 pp., Peter Peregrinus, London. Dellinger, J. H. (1937), Sudden disturbances of the ionosphere, Proc. IRE, 25, Deshpande, S. D., and A. P. Mitra (1972), Ionospheric effects of solar flares-iv. Electron density profiles deduced from measurements of SCNA s and VLF phase and amplitude, J. Atmos. Terr. Phys., 34, 255. Donnelley, R. F. (1971), Extreme ultraviolet flash of solar flare observed via sudden frequency deviation: Experimental results, Sol. Phys., 20, 188. Hernandez-Pajares, M., J. M. Juan, and J. Sanz (1997), High resolution TEC monitoring method using permanent ground GPS receivers, Geophys. Res. Lett., 24, Jones, T. B. (1971), VLF phase anomalies due to a solar X-ray flare, J. Atmos. Terr. Phys., 33, 963. Leick, A. (1995), GPS Satellite Surveying, 560 pp., John Wiley, New York. Liu, J. Y., C. S. Chiu, and C. H. Lin (1996a), The solar flare radiation responsible for sudden frequency deviation and geomagnetic fluctuation, J. Geophys. Res., 101, 10,855. Liu, J. Y., H. F. Tsai, and T. K. Jung (1996b), Total electron content obtained by using the global positioning system, Terr. Atmos. Ocean. Sci., 7, 107. Masuda, S., T. Kosugi, and H. S. Hudson (2000), Hard X-ray two-ribbon flare observed with Yohkoh/HXT, paper presented at Fall Meeting, AGU, San Francisco, Calif. Matsushita, S., and W. H. Campbell (1967), Physics of Geomagnetic Phenomena, Academic, San Diego, Calif. Mendillo, M., et al. (1974), Behavior of the ionospheric F region during the greatest solar flare of August 7, 1972, J. Geophys. Res., 79, 665. Mitra, A. P. (1974), Ionospheric Effects of Solar Flares, 294 pp., D. Reidel, Norwell, Mass. Ohshio, M. (1964), Solar flare effect on geomagnetic variations, J. Radio Res. Lab. Jpn., 11, Ohshio, M. (1971), Negative sudden phase anomaly, Nature, 229, 239. Ratcliffe, J. A. (1972), An Introduction to the Ionosphere and Magnetosphere, Cambridge Univ. Press, New York. Sao, K., M. Yamashita, S. Tanahashi, H. Jindoh, and K. Ohta (1970), Sudden enhancements (SEA) and decreases (SDA) of atmospherics, J. Atmos. Terr. Phys., 32, Sardón, E., A. Rius, and N. Zarraoa (1994), Estimation of the transmitter and receiver differential biases and the ionospheric total electron content from global positioning system observation, Radio Sci., 29, 577. Stonehocker, G. H. (1970), Advanced telecommunication forecasting technique, in Ionospheric Forecasting, AGARD Conf. Proc., vol. 29, edited by V. Agy, Advis. Group for Aerosp. Res. and Dev., North Atl. Treaty Organ., Brussels. Tsai, H. F., and J. Y. Liu (1999), Ionospheric total electron content response to solar eclipses, J. Geophys. Res., 104, 12,657. C. H. Lin, High Altitude Observatory, National Center for Atmospheric Research, Boulder, CO 80301, USA. (clin@ucar.edu) Y. A. Liou, Center for Space and Remote Sensing Research, National Central University, 300 Jungda Road, Jungli City, Taoyuan 320, Taiwan. (yue-ian@csrsr.nce.edu.tw) J. Y. Liu, Institute of Space Science, National Central University, 300 Jungda Road, Jungli City, Taoyuan 320, Taiwan. (jyliu@jupiter.ss.nce. edu.tw) H. F. Tsai, Radio Science Center for Space and Atmosphere, Kyoto University, Gokasho, Uji, Kyoto , Japan. (hftsai@ss.nce. edu.tw) 12 of 12