Effects of the major geomagnetic storms of October 2003 on the equatorial and low-latitude F region in two longitudinal sectors

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 110,, doi: /2004ja010999, 2005 Effects of the major geomagnetic storms of October 2003 on the equatorial and low-latitude F region in two longitudinal sectors Y. Sahai, 1 P. R. Fagundes, 1 F. Becker-Guedes, 1 M. J. A. Bolzan, 1 J. R. Abalde, 1 V. G. Pillat, 1 R. de Jesus, 1 W. L. C. Lima, 2 G. Crowley, 3 K. Shiokawa, 4 J. W. MacDougall, 5 H. T. Lan, 6 K. Igarashi, 7 and J. A. Bittencourt 8 Received 29 December 2004; revised 28 September 2005; accepted 11 October 2005; published 31 December [1] The intense modifications in the ionosphere-thermosphere system in the equatorial and low-latitude regions associated with the dynamic and electrodynamic coupling from high to low latitudes and chemical changes during geomagnetic storms are important space weather issues. In the second half of October 2003, the intense solar activity resulted in one intense and two major geomagnetic storms on 29 and 30 October. In this paper we present and discuss the ionospheric sounding observations carried out from Palmas and São José dos Campos, Brazil (the Brazilian sector), and Ho Chi Minh City, Vietnam, and Okinawa, Japan (the East Asian sector), during these storms. The two sectors are separated by about 12 hours in local time (so while one sector is in daytime, the other one is in nighttime) and provide valuable information related to the storm-time longitudinal differences. Copious storm-time changes were observed in both sectors. It should be pointed out that the two longitudinal sectors investigated in the present study clearly show the global nature of the storm-time effects. However, the responses to the storm-time effects are also associated with the local time in the two sectors. The present investigations show that there are both similarities and differences in the storm-time response in the two sectors. During the storm main phases, with sharp decreases of the Dst index, both sectors showed (dusk or dawn periods) fast uplifting of the F layer associated with magnetospheric electric field penetration. Although in the East Asian sector, Ho Chi Minh City and Okinawa are located fairly close in longitude, with only 2 hour difference in local lime, on occasions the storm-time responses have been very different. Some differences in the latitudinal response of the F region were also observed in the two sectors. Both positive and negative storm phases have been observed at all the four stations. A comparison of the ionospheric parameters obtained from the TIMEGCM model runs and the observed ionospheric parameters at the four stations shows a reasonable agreement during the quiet periods. During the geomagnetic disturbance period, when there were sharp decreases in Dst, some of the observed rapid uplifts of the F region peak heights are not reproduced by the model results. Also, sometimes the model fof2 results differ considerably from the observed fof2 variations. The period investigated represents an extreme storm situation for validation of the model and points to ways in which the model might be improved in the future. Citation: Sahai, Y., et al. (2005), Effects of the major geomagnetic storms of October 2003 on the equatorial and low-latitude F region in two longitudinal sectors, J. Geophys. Res., 110,, doi: /2004ja Introduction [2] During the early part of October 2003, with a few small sunspots, it appeared we were moving toward the 1 Universidade do Vale do Paraiba (UNIVAP), São José dos Campos, São Paulo, Brazil. 2 Centro Universidade Luterano de Palmas (CEULP), Universidade Luterana do Brasil (ULBRA), Palmas, Tocantins, Brazil. 3 Southwest Research Institute, San Antonio, Texas, USA. 4 Solar-Terrestrial Environment Laboratory (STELAB), Nagoya University, Toyokawa, Aichi, Japan. Copyright 2005 by the American Geophysical Union /05/2004JA solar cycle minimum. However, with the sudden increase in sunspots in the second half of October 2003, we witnessed extraordinary solar storms during the period of 18 October to 5 November 2003 (44 M-class and 11 X-class solar flares [Woods et al., 2004]), mainly from two large sunspot groups 5 Department of Electrical Engineering, University of Western Ontario, London, Ontario, Canada. 6 Ho Chi Minh City Institute of Physics, Vietnamese Academy of Science and Technology, Ho Chi Minh City, Vietnam. 7 National Institute of Information and Communications Technology (NICT), Koganei, Tokyo, Japan. 8 Instituto Nacional de Pesquisas Espaciais (INPE), São José dos Campos, São Paulo, Brazil. 1of15

2 Figure 1. Variations of the Kp, Dst, and AE geomagnetic indices for the period 29 to 31 October Also, with the Dst geomagnetic index variations (dots), we show the SYM-H index (1-min resolution Dst; black line) and geomagnetic H-component variations (1-min values; gray line) observed during the same period at Vassouras (VAS), Brazil. (484 and 486). The sunspot group 486 on 28 October was near the solar disc center and unleashed a huge H-alpha 4B/ X-17.2 class solar flare at 1110 UT, which was followed by another X10 class solar flare on 29 November at 2049 UT (N. Srivastava, Udaipur Solar Observatory, India, private communication, 2004). The intense solar flares on 28 and 29 October launched fast coronal mass ejections (CMEs) directly toward the Earth with maximum speeds of 1900 km/s (29 October) and 1940 km/s (30 October) (Advanced Composition Explorer (ACE) satellite observations [Zurbuchen et al., 2004]). The impacts of the two fast CMEs on the Earth s magnetosphere resulted in major geomagnetic disturbances on 29 and 30 October. The extraordinary solar activity in October November 2003 attracted rapid [e.g., Woods et al., 2004; Zurbuchen et al., 2004] and widespread (e.g. Violent Sun-Earth Connection Events of October November 2003, special session AGU/CGU/SEG Joint Assembly, Montreal, Canada, May 2004) attention of the scientific community. [3] Figure 1 presents the time variations of the Kp, Dst, and AE geomagnetic indices, during the period 29 to 31 October 2003 (hereafter referred as the October 2003 events). The Dst and AE indices are provisional. The geomagnetic field H-component variations (every minute) observed at Vassouras (hereafter referred as VAS), Brazil, and the high-resolution variation of the SYM-H index (1-min values; which closely follows the Dst index (hourly values), are also shown in Figure 1. The storm-time H-component variations at VAS shown in Figure 1 were obtained after subtracting a quiet day (23 October 2003; P Kp = 11) H-component diurnal variations. A perusal of Figure 1 indicates that the period between 29 and 31 October was having major geomagnetic disturbances on UT days 29 and 30 October with recovery phase starting on 31 October. There were two storm sudden commencements (SSCs) during this period occurring at 0611 UT on 29 October and 1029 UT on 30 October ( The SSC on 29 October was major, whereas the SSC on 30 October was relatively very small. During the period 29 to 30 October, three geomagnetic storms were observed. First an intense storm with jdstj max = 180 nt at 1000 UT (29/10), second a major storm with jdstj max = 363 nt at 0100 UT (30/10), and third another major storm with jdstj max = 401 nt at 2300 UT (30/10). The AE variations, based on high-latitude magnetograms, also show intense substorm activity on 29 and 30 October, with the AE index several times attaining values more than 2500 nt. [4] Reviews on ionospheric storms covering high, middle, and low latitude regions have been recently provided by Schunk and Sojka [1996], Abdu [1997], and Buonsanto [1999]. However, numerous individual studies of magnetosphere-ionosphere interactions at equatorial and low-latitude regions during geomagnetic storms continue to be published. Some of the recent studies use both observational [e.g., Sa. Basu et al., 2001, 2005; Su. Basu et al., 2001; J.J. Lee et al., 2002; C.-C. Lee et al., 2002, 2004; Sastri et al., 2002; Blagoveshchensky et al., 2003; Yizengaw et al., 2004; Lima et al., 2004; Sahai et al., 2004] and modeling [e.g., Fuller- Rowell et al., 2002; Meier et al., 2005] results. The reason for the continued intense interest is the lack of understanding and our inability to predict thermospheric and ionospheric space weather responses. The outstanding questions focus on how variable the storm-time responses can be and how they are driven. There appear to be three major drivers of the ionospheric response: electric fields, winds and composition changes. The electric fields in the equatorial and low-latitude ionospheric F region during geomagnetic storms are affected by the solar wind-magnetosphere dynamo resulting in prompt or direct penetration of the magnetospheric convective electric field [e.g., Senior and Blanc, 1984; Spiro et al., 1988] and the ionospheric disturbance dynamo that results from global thermospheric wind circulation due to Joule heating at high latitude [e.g., Blanc and Richmond, 1980]. A detailed study of the storm-time dependence of the equatorial zonal electric fields has been presented by Scherliess and Fejer [1997] and Fejer and Scherliess [1997]. As pointed out by Nicolls et al. [2004], low-latitude and midlatitude disturbances are created by the transport of energy from high latitudes in the form of traveling atmospheric disturbances (TADs) and changes in global wind pattern through direct heat deposition from precipitating particles via Joule heating or Lorentz forcing. Extension of the negative ionospheric storm or phase (a decrease in F region peak electron density) during geomagnetic storms to low latitude indicates changes in the thermospheric composition (O/N 2 ) ratio [e.g., Greenspan et al., 1991; Zhang et al., 2003; Meier et al., 2005]. 2of15

3 Table 1. Details of the Observing Sites Used in the Study Location Symbol Used Geog. Lat. Geog. Long. Dip Lat. Observation/ Model Results Palmas, Brazil PAL 10.2 S 48.2 W 5.7 S h 0 F, hpf2, fof2 & Sp-F/hmF2 & fof2 Vassouras, Brazil VAS 22.4 S 34.6 W 18.5 S Geomagnetic field H component São José dos Campos, Brazil SJC 23.2 S 45.9 W 17.6 S h 0 F, hpf2, fof2 & Sp-F/hmF2 & fof2 Ho Chi Minh City, Vietnam HCM 10.5 N E 2.9 N h 0 F, hmf2, fof2 & Sp-F/hmF2 & fof2 Okinawa, Japan OKI 26.3 N E 21.2 N h 0 F, hpf2, fof2 & Sp-F/hmF2 & fof2 Sata, Japan SAT 31.0 N E 26.0 N OI 630 nm emission all-sky images Yamagawa, Japan YAM 31.2 N E 26.6 N Indication of sp-f occurrence [5] In this paper we report on ionospheric sounding observations obtained from Palmas and São José dos Campos, Brazil (the Brazilian sector), and Ho Chi Minh City, Vietnam and Okinawa, Japan (the East Asian sector), during the major geomagnetically disturbed period of 29 to 31 October These data allow us to compare and contrast the latitudinal and longitudinal responses in the Brazilian and East Asian sectors. The data are also compared with simulations from a global three-dimensional (3-D) model, the TIMEGCM (Thermosphere Ionosphere Mesosphere Electrodynamics General Circulation Model) [Roble and Ridley, 1994; Crowley et al., 1999]. 2. Observations [6] Establishment of a new ionospheric sounding network in Brazil was initiated by the Universidade do Vale do Paraiba (UNIVAP) in August 2000 with the installation of a Canadian Advanced Digital Ionosonde (CADI) [Grant et al., 1995] at São José dos Campos (hereafter referred as SJC) and was followed by installation of two more CADIs at Palmas (hereafter referred as PAL) and Manaus in collaboration with Universidade Luterana do Brasil (ULBRA) during the year Unfortunately, the CADI at Manaus had a technical problem during the October 2003 events. In this paper we present and discuss the ionospheric sounding observations obtained at PAL (a near-equatorial station) and SJC (a low-latitude station located under the crest of equatorial ionospheric anomaly), in order to investigate the response of the equatorial and low-latitude ionospheric regions in the Brazilian sector during the October 2003 events. The two stations PAL and SJC are nearly meridional (geomagnetic) and in the same time zone (UT = LT + 3 hours). Also, in order to investigate the longitudinal differences in the response of the equatorial and low-latitude ionospheric regions during the October 2003 events, we present and discuss the ionospheric sounding observations obtained at Ho Chi Minh City, Vietnam (a near-equatorial station with UT = LT 7 hours; hereafter referred as HCM) and Okinawa, Japan (a low-latitude station with UT = LT 9 hours; hereafter referred as OKI), in the East Asian sector. The two longitudinal sectors investigated differ by about 12 hours in local time, so while one sector is in daytime the other sector is in nighttime. Table 1 provides the details of all the observing locations from which data have been presented in this study. [7] The storm-time ionospheric response is usually studied primarily in terms of deviations of the F region critical frequency (fof2) from the median value [e.g., Danilov and Morozova, 1985] for the same time of the day (positive ionospheric storm or phase (increase in peak electron density) or negative ionospheric storm or phase (decrease in peak electron density)) and changes in the F region height [e.g., VanZandt et al., 1971; Reddy and Nishida, 1992; Sastri et al., 2002]. These height changes can be represented by the change in the minimum virtual height (h 0 F), the virtual height at any fixed frequency or the peak height (hmf2) of the F region. In the present investigation we have used h 0 F, and we also used hpf2 (virtual height at fof2) [see, e.g., Piggot and Rawer, 1972] to represent the peak height of the F region at PAL, SJC, and OKI. For HCM we have used hmf2. As pointed out by Danilov and Morozova [1985], the determination of the F region peak height using hpf2 is less reliable and overestimated during the daytime compared with the true-height analysis (hmf2). However, during the nighttime hpf2 values are fairly close to hmf2 [Danilov and Morozova, 1985]. Batista et al. [1991] have reported that the parameter hpf2 could depart from the hmf2 obtained by the height ionogram inversion by 50 km during daytime and by 10 km at night. In the present investigation, the principal interest is to study the large height changes from the median values during the major geomagnetic storms. Therefore it appears that the use of hpf2 even during the daytime is reasonable. [8] In order to calculate the median values of the ionospheric parameters (h 0 F, hpf2 or hmf2, and fof2), we have used the measurements obtained on 8 to 9 quiet days (QD) (based on the list of 10 quiet days provided with the Kp index for the months of October November 2003 and available through the website tab/). We have used QD measured values before and after the storm period. Since October is a transition month and medians change significantly from summer to winter (or vice versa), quiet days from both October and November 2003 have been used to obtain the median values. The quiet days (UT) used for the different ionospheric sounding stations are as follows: PAL (every 30 min; 4, 5, 6, 7, 8 October, 5, 7, and 8 November, 2003; spread F on all the nights), SJC (every 30 min; 4, 5, 8, 9, 10, 23 October and 5, 7, and 8 November, 2003; spread F on all the nights except 5 6 October, 7 8 November, and 8 9 November), HCM (every 30 min; 4, 5, 10, 11, 12, 23 October, and 5, 7, and 8 November 2003; spread F on all the nights except October, 5 6, 7 8 November, and 8 9 November), OKI (every 30 min; 5, 6, 8, 10, 11, 12 October and 5, 7, and 8 November 2003; spread F on the nights of 6 7 and October and 5 6 November). The scaled ionospheric parameters for the ionospheric sounding stations HCM and OKI were kindly provided by the respective operating agencies. 3of15

4 Figure 2. Variations of the ionospheric parameters hpf2 (red dots), h 0 F (red dots) and fof2 (red dots) observed at Palmas (PAL), during the period 29 to 31 October The inset figure on 30 October shows the unusual h 0 F variations between 2200 UT and 0030 UT (h 0 F attained 950 km altitude; red dots connected with line). Also, the local time (LT) is shown at the top. The median quiet day variations of hpf2, h 0 F, and fof2 are shown in black line. Variations of the ionospheric parameters hmf2 (blue line) and fof2 (blue line) obtained from the TIMEGCM model during this period are also shown. The hatched portions indicate local nighttime periods (1800 to 0600 LT). The horizontal black bar indicates the presence of spread F. [9] In the present investigation, we also discuss the observations of equatorial spread F (ESF), when present, during the October 2003 events at different locations. 3. Results and Discussion [10] In general, geomagnetic storms show a SSC (sometimes), a main phase, and a recovery phase. As discussed by Schunk and Sojka [1996], during the growth or main phase, the magnetospheric electric field and particle precipitation patterns expand, the electric fields become stronger, and precipitation becomes more intense. During this phase, the Joule and particle heating rates and the electrojet currents increase, providing maximum energy input to the upper atmosphere, while during the recovery phase the geomagnetic activity and energy input decreases. Also, in the main phase, when rapid changes of the Dst index occurs, penetration electric fields of magnetospheric origin are known to affect the ionospheric dynamics at low latitudes [Wygant et al., 1998; Su. Basu et al., 2001; Basu et al., 2005]. The storm-time penetration of magnetospheric electric field to low latitudes must be global and is observed on both the dayside and nightside [Abdu et al., 1991; Fejer and Scherliess, 1997; Sastri et al., 2002]. However, as pointed out by Reddy and Nishida [1992], only large-magnitude disturbance electric fields can cause recognizable simultaneous changes of F region heights over a range of latitudes in daytime because much larger ionospheric conductivity in daytime increases the attenuation of the disturbance electric field Quiet Day Median Variations of the Ionospheric Parameters at PAL, SJC, HCM, and OKI [11] Figures 2, 3, 4, and 5 show the quiet day calculated median values (black lines) of the ionospheric parameters (h 0 F, hpf2 or hmf2, and fof2) at PAL, SJC, HCM, and OKI, respectively. A prominent feature observed in the fof2 variations at the low-latitude stations, SJC and OKI, is the presence of the equatorial ionospheric anomaly, with maximum of fof2 values around the afternoon hours (about LT). At the equatorial stations, PAL and HCM, the fof2 variations show the noontime biteout and enhancement close to the local midnight. Again, the h 0 F variations at 4of15

5 Figure 3. Same as in Figure 2, except for São José dos Campos (SJC). both PAL and HCM show the well-known prereversal enhancement (postsunset) of the vertical drift and another uplifting of the F layer in the early morning hours. The discontinuous fof2 and hpf2 median plots in the premidnight period are due to the presence of strong spread F on the quiet nights at both PAL and SJC. The onset of spread F is normally associated with the rapid upward motion of the F layer ionization that usually occurs just after sunset, near the magnetic equator. After the reversal of the plasma drift velocities, from upward to downward, an inverted fountain effect is produced, which tends to move the F layer downward in the equatorial region, during nighttime, into regions of increased recombination rates, at the same time that ionization is transported from higher latitudes toward the magnetic equator, giving rise to near midnight peak in fof2 close to the magnetic equator region. Thereafter, the values of fof2 continue to decrease because of recombination. The slight rise in the values of h 0 F that occurs in the early morning hours, near the magnetic equator, is probably just an apparent effect, since the heights at this time are usually actually decreasing. This apparent height rise in h 0 F may be the result of recombination when ionization transport from nonequatorial latitudes is already much reduced Storm-Time Variations of the Ionospheric Parameters at PAL, SJC, HCM, and OKI [12] As discussed by Basu et al. [2005], the geomagnetic storms during the period October are complex in nature and have three rapid ring current injection periods, causing the rate of change of Dst to exceed 50 nt/hr ( UT (29/11), UT (29/10), and UT (30/10)). As mentioned earlier, these periods of rapid changes of the Dst index are appropriate for prompt penetration of electric fields of magnetospheric origin to low latitudes [Wygant et al., 1998; Su. Basu et al., 2001; Basu et al., 2005]. During complex storm periods, sometime the penetrating magnetospheric electric and the electric fields generated by the disturbance dynamo (associated with the Joule heating in high-latitude regions and indicated by the enhancement of the auroral electrojet (AE) index) could result in mixing of the two effects [Reddy and Nishida, 1992; Sobral et al., 1997]. Sastri et al. [2002] have pointed out that near the magnetic equator, vertical plasma drift is essentially due to zonal electric fields. Meridional winds are usually ineffective in producing vertical plasma drifts close to the magnetic equator but gain in importance with increase of dip angle (I) with maximum effect at I = 45. The prominent effects observed during these two major geomagnetic storms in the two longitudinal sectors will be the principal focus of the present investigation. [13] Figure 2 also shows the variations of the ionospheric parameters hpf2 (top panel), h 0 F (middle panel), and fof2 (bottom panel), obtained every 15 min during the period 29 to 31 October 2003 at PAL. In each panel, the storm-time observations are indicated by red dots, and for comparison 5of15

6 Figure 4. Same as in Figure 2, except for Ho Chi Minh City (HCM) and at HCM we have hmf2 in place of hpf2. Numbers in the hatched portions inset indicate the respective nights. Some of the red dots are shown connected by line just to indicate rapid changes. typical quiet day median variations are shown in black line. Model predictions for the storm period (every 1 hour values) are shown in blue line. Figures 3, 4, and 5 present similar storm-time results for SJC (every 15 min values), HCM (every 15 min values), and OKI (every 30 min with sometimes 15 min), respectively, with typical quiet day median variations and model results (every 1 hour values). Note that the top panel compares the measured hpf2/hmf2 with the modeled hmf2, and since the model does not predict h 0 F, no model values are shown in the middle panel. Local nighttime periods (1800 LT to 0600 LT) are indicated by hatched portions. The absence of storm-time data and quiet time median values in any of these figures is due to instrumental problems, presence of strong spread F, radio-wave absorption, or any other difficulty in ionogram scaling. [14] Figure 6a shows the variations of virtual heights at six fixed reflection frequencies (electron densities), with measurements every 100 s (isofrequency plots), observed at PAL (top panel) and SJC (bottom panel) on UT days 29 to 31 October The use of multifrequency ionospheric sounding observations is fairly common to study gravity waves and propagation characteristics [e.g., Abdu et al., 1982; MacDougall et al., 1997; Lee et al., 2002; Lima et al., 2004]. [15] A perusal of Figure 6a indicates that soon after the SSC on 29 October at 0611 UT, there was an unusual simultaneous fast uplifting (eastward electric field) of the F region at PAL and SJC at about 0730 UT (marked with line A in Figure 6a). For about half hour during the predawn period 0730 to 0800 UT, the F layer traces were outside the normal 1000 km height range of the ionograms at the two locations. During this period the fof2 values decreased considerably at both stations, in relation to the quiet time observations. The unusual uplifting observed is possibly associated with direct penetration of magnetospheric electric field from high to low latitudes during the period with rapid changes in the Dst index ( UT (29/11) as mentioned earlier). During the predawn period, conductivity is very low so electric field could easily penetrate. The variations in fof2 at SJC after this unusual uplifting resulted in a sharp increase from about 4 to 19 MHz, and some increase was also observed at PAL from about 3 to 9 MHz, which lasted just for a short period of about 2 hours at both the locations. It should be pointed out that the vertical drifts are upward during the daytime and downward during the nighttime in quiet conditions. Therefore it appears that the storm-time electric field lifted the F region in the equatorial region and the plasma diffused down the magnetic field lines due to gravity, creating a short duration ionization 6of15

7 Figure 5. Same as in Figure 2, except for Okinawa (OKI). Numbers in the hatched portions inset indicate the respective nights. cloud over SJC. The process is similar but distinct from the formation of equatorial ionospheric anomaly. [16] During this direct penetration of magnetospheric electric field, the station HCM in the Vietnamese sector was in daytime (0730 UT, 1430 LT), whereas OKI in the Japanese sector was close to dusk period (0730 UT, 1630 LT). Figure 4 shows that both h 0 FandhmF2atHCMstart decreasing at about 0730 UT, possibly associated with westward electric field. This illustrates the opposite polarity of penetration electric fields expected in the two longitude sectors (Brazilian and Vietnamese) so that the F region heights are seen to increase in one sector and the ionosphere collapses in the other sector. However, it appears that the effect of the prompt penetration electric field was different in the Japanese sector compared with the Vietnamese sector. Figure 5 shows an unusual increase in fof2 (positive storm effect) at OKI after the SSC at 0611 UT. The Japanese sector was closer to the dusk period. Schunk and Sojka [1996] have pointed out that as the ionosphere corotates with the Earth toward dusk, the zonal (eastward) component of the neutral wind increases and the increased eastward wind, in combination with sharp day-night conductivity gradient across the terminator, leads to a prereversal enhancement in the eastward electric field, which acts to raise the F layer as it rotates into darkness. Therefore at dusk time in the presence of decreasing conductivity gradient, it appears that the prompt penetration field associated with rapid decrease in the Dst index is superimposed on the existing normal prereversal conditions and the F layer is tossed up to unusual heights in the equatorial region. The positive storm phase observed at OKI is possibly associated with unusual F layer height rise in the Japanese equatorial region due to the direct penetration of the electric field (eastward), resulting in diffusion of plasma down the magnetic field lines due to gravity, creating an unusual enhancement in electron density over OKI. However, the observations at HCM show contrary effect (westward electric field) as evidenced by the suppression of the prereversal enhancement as seen in both h 0 F and hmf2 variations. [17] During the daytime on 29 October both PAL and SJC show wavelike oscillations in fof2 and hpf2 and HCM and OKI show wavelike oscillations in fof2, h 0 F, and hpf2/ hmf2 during the nighttime (29 30 October). These oscillations are possibly associated with neutral winds/gravity waves generated due to the Joule heating in the auroral region with the large enhancement of the auroral electrojet (AE) index between about 0600 UT on 29 October and about 0200 UT on 30 October. Huang and Cheng [1991] have explained the oscillatory ionospheric motion due to the compression and expansion of plasmasphere during geo- 7of15

8 Figure 6a. Plots of virtual height variations for six different frequencies (isofrequencies) observed at PAL (top panel) and SJC (bottom panel) for the UT days 29 to 30 October, Also the local time is shown at the top. magnetic disturbances. As pointed out by Crowley et al. [1989a, 1989b] and Fuller-Rowell et al. [1997], during geomagnetic disturbances, large-scale waves propagate efficiently from the high-latitude source and fast penetrating waves arise on the nightside, where they are hindered least by drag from the low ion density [18] As mentioned earlier, again during the period UT on 29 October, the Dst index was decreasing fast. The Brazilian sector was in the dusk period, whereas the East Asian sector was in the dawn period. Figures 4 and 5 show that possibly due to the prompt penetration of electric field (eastward), there was unusual simultaneous uplifting of the F layer at HCM and OKI (predawn with very low conductivity) and during the period 2030 UT to 2200 UT the F layer traces were outside the usual height range of the ionograms at these two locations. However, during this period of prompt penetration, the response in the Brazilian sector (dusk period with prereversal conditions) appears to be more complex with wave-like variations (Figure 6a) between 2000 UT on 29 October and 0600 UT on 30 October. As mentioned earlier, the enhancement of the auroral electrojet (AE) index at about 0600 UT on 29 October and continuing for several hours had possibly set up the ionospheric disturbance dynamo before this prompt penetration. The prompt penetration of the electric field in the presence of disturbance dynamo during the dusk period has been recently investigated by Sa. Basu et al. [2001, 2005]. The present case study indicates that the response of the equatorial ionosphere when the penetration of electric field occurred in the presence of ionospheric disturbance dynamo is very different in the dusk (Brazilian) and dawn (East Asian) sectors. [19] Figure 6b shows an amplified version of the isofrequency plots (Figure 6a) between the period 1800 UT on 29 October and 0600 UT on 30 October. During the period 1800 to 2000 UT on 29 October, the isofrequency plots at SJC clearly show downward propagation indicating the presence of the traveling ionospheric disturbance (TID), associated with the Joule heating in the auroral zone as evidenced by the large enhancements in the auroral electrojet (AE) index starting at about 0600 UT on 29 November and continuing for several hours (Figure 1). However, the observations at PAL do not show such feature. TID will show up much more at SJC than at PAL because of the larger inclination of the field lines. As one approaches the equator, the TID height effects go to zero. Thus at PAL there could be the same TID, but the height effects would be 8 of 15

9 Figure 6b. An amplified version of the isofrequency plots (Figure 6a) between the period 1800 UT on 29 October to 0600 UT on 30 October. difficult to see. This possibly implies that the TID effect on ionosphere reaches the anomaly crest region and diminishes toward equatorial regions [Huang and Cheng, 1993]. However, after 2000 UT it becomes difficult to say whether there is downward or upward propagation. Of course, vertical upward motion can be driven by eastward electric fields. Figures 2 (PAL) and 3 (SJC) very clearly show that the postsunset prereversal at both the stations is considerably enhanced compared with the quiet day conditions, indicating the additional influence of eastward electric field. This indicates that there was a change at about 2000 UT in what was happening before. Prior to 2000 UT the fluctuations at SJC are gravity waves. The evidence for this is both downward phase propagation and the increased amplitude of the waves with height (usually near the critical frequency gravity waves produce large perturbations due to changing the electron densities, which in turn causes large variations in virtual height (this is a different effect from the increased amplitude of gravity waves with height due to the reduced neutral densities). After about 2000 UT the height variations tend to all be of the same magnitude. Thus there may be a change from gravity wave effects to electric field-driven variations. [20] The multiple trace features that show up at SJC between about 2300 UT (29/10) to about 0100 UT (30/10) might be range spread that is associated with the spreading at PAL. Between 0000 UT and 0030 UT we also see the presence of satellite trace on ionograms at SJC, possibly associated with reflections from overhead plus a reflection from a moving kink. Both PAL and SJC (Figures 2, 3, and 6b) show the presence of strong spread F on the night of October. Possibly, the strong spread F observed in the Brazilian sector on this night was also influenced by the magnetic disturbance, resulting in an unusual enhanced prereversal mentioned earlier. As reported by Basu et al. [2005], the present investigation also shows the onset of spread F in the postsunset period during the ionospheric disturbance dynamo conditions existing prior to the prompt penetration of magnetospheric electric field. Possibly, during strong magnetic disturbances the model predictions [Fejer and Scherliess, 1997] are contrary. [21] The Dst index was again fast decreasing at about UT on 30 October. This period again corresponds to dusk time in the Brazilian sector and dawn time in the East Asian sector. At about 2000 UT (1700 LT; see Figures 2, 3, and 6a (marked with line B )), close to the prereversal time, the F region was violently tossed up simultaneously at PAL and SJC, with h 0 F attaining a very unusual height of about 925 km (see inset in Figure 2) at PAL (h 0 F uplifting at SJC was less prominent compared with PAL), possibly due to the prompt penetration of electric field (eastward). During this period the fof2 values decreased considerably at PAL, in relation to the quiet time observations. During the major magnetic storm of 13 March 1989, Batista et al. [1991] have reported similar dramatic rise of the F layer height simultaneously over Fortaleza (4 S, 9of15

10 Figure 7. A sequence of the OI 630 nm all-sky images observed at Sata (SAT) on the night of 31 October to 1 November W; 8.5 dip) and Cachoeira Paulista (22.5 S, 45 W; 26 dip) for about 2 hours. The unusual lifting of the F layer was followed by the onset of spread-f at PAL (1845 LT) and SJC (1900 LT). In the East Asian sector (predawn with very low conductivity), the observations at HCM (Figure 4) show that possibly due to the prompt penetration of electric field (eastward), there was unusual uplifting of the F layer during the period 2015 UT to 2215 UT and the F layer traces were outside the usual height range of the ionograms. However, the F layer uplifting at OKI in this sector was less evident during this period, but the period of uplifting in the equatorial region was followed by unusual increase in fof2 at OKI for about 3 hours. These observations at OKI are very similar to the observations at SJC after the SSC at 0611 UT on 29 October. [22] During the daytime on 31 October, in the recovery phase, the fof2 variations at SJC show negative storm phase. This is possibly associated with the extension of changes in thermospheric composition (O/N 2 density ratio) during geomagnetic disturbances from middle to low latitudes. In the East Asian sector on the night of 31 October to 1 November, at about 1300 UT (2200 LT), the variations in fof2 at OKI show the start of an unusual increase in fof2 (Figure 5), which attains maximum at about 1600 UT 10 of 15

11 (0100 LT). Both h 0 F and hpf2 also show slight increase at OKI at about 1300 UT. Figure 7 shows a sequence of the OI 630 nm emission all-sky images obtained at Sata (SAT) (north of OKI) on this night with relatively clear sky. The enhancement of airglow between 1400 and 1700 UT on 31 October is clearly seen in Figure 7. Also, we do see intensification of airglow from south (a strong northeast movement of plasma during the fof2 increase observed at OKI (Figure 5)), suggesting a penetration of eastward electric field and subsequent poleward shift of the equatorial ionospheric anomaly and appearance of equatorial plasma bubbles (quasi north-south aligned) in the images (with spread F at OKI starting at 1400 UT (2300 LT) and continuing up to 2130 UT (0630 LT); this is the only night among the three nights studied which showed the presence of equatorial spread F in the Japanese sector). It should be pointed out that there is some substorm-like activity starting from 1300 UT. This activity can be seen in the auroral zone data in the 210MM magnetic field data at gm0310/ gif. Also, during this period a sharp increase of the geomagnetic horizontal field component (H) was observed at VAS, a low-latitude station (Figure 1), indicating substorm activity. The Dst index continued to increase on 31 October but temporarily decreases between 1300 and 1500 UT probably because of this substorm activity as seen in Figure 1. Thus this substorm activity possibly caused the prompt penetration of electric field at around 1300 UT. During the prompt penetration of electric field, the ionospheric observations at HCM do not show any appreciable changes in the F region height (hf and hmf2) but fof2 variations show continuation of positive storm phase observed during the daytime on 31 October up to about 1600 UT and then the fof2 values start decreasing. [23] No spread F was observed at HCM on any of the three nights. As mentioned earlier, spread F was observed at OKI on the night of 31 October to 1 November. It should be mentioned that the ionospheric sounding observations from Yamagawa (YAM), very close to SAT, also showed the presence of spread F between 2315 and 0345 LT on this night. The white color in Figure 7 shows the highest intensity. The upward velocity of bubble development is m/s [Tsunoda et al., 1981], which is about km/h. Because the apex height of magnetic field line at SAT is about 1500 km, it takes 1 3 hours for the bubble to reach SAT latitudes from the magnetic equator. The spread F observations in the Japanese sector on the night of 31 October to 1 November are associated with the magnetic disturbance. Since no spread F was observed at HCM on this night, the spread F at OKI and plasma bubbles observed at SAT could be generated at longitudes west of OKI but east of HCM on this night. It is not possible to say why spread F was not observed at HCM on this night, although the longitudinal width between OKI and HCM is fairly narrow (separated only by 2 hours in local time). [24] During this prompt penetration of electric field, the Brazilian sector was in daytime. At PAL the observed variations in h 0 F, hpf2, and fof2 during the daytime on 31 October (Figure 2) are fairly similar to those observed on 30 October (daytime). Therefore it is difficult to visualize any influence of the effect of prompt penetration. At SJC, during the period 1130 to 1515 UT due to strong radio-wave absorption we had no F region ionogram traces and between 1600 and 1700 UT there was an instrumental problem. However, a perusal of Figure 3 indicates that around 1530 UT, when ionogram traces were back, the fof2 variations show unusual increase from large negative storm phase and the hpf2 variations show a decrease, possibly associated with westward electric field. [25] It should be pointed out that the two longitudinal sectors investigated in the present investigation clearly show the global nature of the storm-time effects. However, the responses to the storm-time effects are also associated with the local times in the two sectors. The present investigation shows that there are both similarities and differences in the storm-time response in the two sectors. Both positive and negative storm phases have been observed at all the four stations. On occasions the storm-time response shows latitudinal differences in the same sector as well. It is hoped that future investigations with relatively larger grid of ground-based ionospheric sounding observations during storm-time will provide more useful and important results related to space weather effects in equatorial and lowlatitude regions Comparison of the Multisite Observations With the TIMEGCM Model Results [26] The variations in hmf2 and fof2 (hourly values) for the four locations (PAL, SJC, HCM, and OKI) obtained during the period 29 to 31 October from the TIMEGCM model results are also shown in Figures 2 5 (blue line). The model runs were appropriate for the geomagnetic conditions and F10.7 solar flux for the day. The extent of auroral oval and particle fluxes were based on Hemispheric Power obtained from the DMSP and NOAA satellites (values every 15 min). The cross-cap potential was obtained from the IMF B y component [Heelis et al., 1982]. The Weimer [1996] model driven by solar wind inputs was used to obtain the cross-cap potential difference. [27] It should be mentioned that the observed median values for the quiet days shown in Figures 2 to 5 (black lines) and TIMEGCM median values for the quiet days (not shown; every 1 hour; 9, 10, 11, 12, 23 October and 5, 7, and 8 November 2003) agree reasonably well for all the four locations. A comparison of the observed fof2 with the TIMEGCM model results in the East Asian sector (Figures 4 (HCM) and 5 (OKI)) shows that, in general, the diurnal patterns on the disturbed days 29 and 30 October agree reasonably well. Of course there are some differences between the observed and model results. It should be mentioned that the observed short period oscillations/enhancements are not reproduced by the model. However, on the recovery day of 31 October, from about 0400 UT to about 2100 UT, the observed fof2 values at OKI show positive storm phase, whereas the model results show negative storm phase. Again similar results are seen at HCM but for the period between 0900 UT and 2200 UT. The unusual enhancements in fof2 observed at both OKI and HCM on the night of 31 October to 1 November are not reproduced by the model. In the Brazilian sector (Figures 2 (PAL) and 3 (SJC)), a similar comparison shows that in general, the observed and model diurnal patterns of fof2 values compare reasonably well up to about 2000 UT on 29 October. Again some differences between the observed and modeled variations are there and the observed short 11 of 15

12 period oscillations/enhancements are not reproduced by the model. However, after this both the observed and model values of fof2 at PAL show negative storm phase during the nights of and October. During the daytime on 30 October only the observed fof2 values show positive storm phase and whereas on the daytime on 31 October only the model fof2 values show negative storm phase. The model fof2 values at SJC show negative storm phase starting from about 2000 UT on 29 October till 2400 on 31 October. The observed fof2 values at SJC show positive storm phase between about 2000 UT on 29 October and about 0300 UT on 30 October. Possibly, there is negative storm phase at SJC between 0300 UT and 0900 UT on 30 October (no fof2 data because of spread F). The observed fof2 values follow the median values during the daytime on 30 October. On the night of October, for most part of the night there are no observed fof2 values due to spread F. However, it appears that the observed fof2 values show positive storm phase in the beginning of the night and negative storm phase at the end on the night. The negative phase at the end of the night on October accompanies the model results for most of the daytime on 31 October (up to about 1600 UT) and the observed fof2 values reach to the median level around 1800 UT on 31 October. [28] Figures 2 5 show the hmf2 variations obtained from the model with the observed hpf2/hmf2 variations for the four locations. In general, it appears that there is a reasonable agreement between the model-generated hmf2 and the observed hpf2 in the Brazilian sector during the period 29 to 31 October. In the Brazilian sector, both the model and observed F region peak heights show strong short-period uplifting during nighttime. It should be pointed out that in the Brazilian sector, on the disturbed nights of October and October, both the locations show the presence of strong spread F and it is not possible to estimate hpf2; however, the strong variations in hmf2 from the model are reflected in the observed h 0 F variations. In the East Asian sector, the model results of hmf2 are much higher compared with the observed F region peak heights on the recovery day (31 October). As mentioned earlier, both the sectors show fast uplifting of the F layer during dawn/dusk periods in association with prompt penetration of electric field when rapid changes in the Dst index occurs. The fast uplifting of the F layer in either of the two sectors are not reproduced in the model results. [29] There are major discrepancies between the model results and the observed ionospheric parameters during strong geomagnetic disturbances investigated. This possibly indicates that some of the input parameters in the model used may need a reevaluation. Most of the global 3-D models such as the TIMEGCM have received little validation against large data sets. On the average they perform reasonably well against empirical models such as MSIS and IRI. However, there is interest in validating the models against real data to evaluate their performance. In this study, we report on a validation using ionosonde data from four stations. As mentioned in the paper, the model performed very well for quiet times at all four sites, but this was not the focus of the present paper. Rather, this paper focuses on an extreme space weather event during 3 days in October [30] During storm times, there may be significant changes in the factors affecting the ionosphere, namely composition, winds, and electric fields. The storm time electric fields consist of storm time dynamo fields driven by the winds, together with penetration E fields. The TIMEGCM selfconsistently computes changes in the composition, winds, and the dynamo electric fields. However, there are no penetration E fields included in this version of the model. Thus to the extent that penetration E fields drive the ionospheric hpf2/hmf2 and fof2, those changes will not be captured by the model. [31] The modeled storm time response is a strong function of the high-latitude inputs specifications. We describe how the high-latitude inputs used in the present study were fairly coarse, consisting of analytical models of the convection [Heelis et al., 1982] and auroral precipitation [Roble and Ridley, 1987] driven by simple parameters such as By, cross-cap potential difference, and hemispheric power index. We have found that these input specifications are adequate for numerical experiments, and for simple studies of typical storm effects; however, we are aware that they are limited when it comes to simulating real storms. Studies such as the one reported here are driving changes in the way the model inputs are specified, and work is now in progress to improve the particle and electric field specifications in the model. For the present time, we are content to learn how the model responds with the current input specification. [32] We find that the major discrepancy between the model and the data for the October storm reported here is the variation of hpf2/hmf2 and fof2 on the recovery day (31 October 2003). Specifically, the model predicts a negative storm phase at all four sites, but only the SJC site observed a negative phase during the recovery. Similar effects in other storm studies [e.g., Meier et al., 2005] have been reported. The model appears to be forced too strongly during the storm, producing global-scale changes in the thermospheric composition that appear to be too extreme and which last for too long. In particular, the O/N2 ratio is reduced globally due to the upwelling and transport of molecular nitrogen. The corresponding increase in hmf2 is caused by exospheric temperatures that are too large, so the thermosphere (and ionosphere) expand to greater altitudes. We anticipate that this problem will be ameliorated by improvements in the high-latitude input specifications that are planned for the near future. [33] On the subject of electric fields, these would be expected to modify the height of the ionosphere and indirectly the fof2. Although the modeling study by Maruyama et al. [1998] indicates a fairly modest effect at low latitude to the prompt penetrating electric field, as pointed out by Fuller-Rowell et al. [2002], this source has the potential to raise the F region fairly rapidly and impact the low-latitude ionospheric structure. We have mentioned that penetration electric fields were not used in the model; however, dynamo E fields should be simulated reasonably well assuming the model predicts wind patterns that are reasonable. It is difficult to validate this question without neutral wind or corresponding electric field measurements. We note, however, that the model appears to predict the large height changes at SJC and PAL throughout 30 and 31 October, suggesting most of these effects are disturbance dynamo-produced and any 12 of 15

13 penetration E field contribution is fairly small. On the other hand, hmf2 is consistently too high in the Asian sector, which is partly due to the temperature effect mentioned above but may also be due to incorrect disturbance dynamo effects and lack of penetration E field in the model. In future, it would be useful to obtain estimates of the penetration E fields in the two sectors for further comparisons with the model and the data presented here. However, such a study is outside the scope of present work. [34] The high-latitude inputs to the TIMEGCM are still fairly crude. Studies such as the one reported here are revealing how the models and their inputs need to be improved. We are aware that the analytical representation of the auroral flux and energy distribution [Roble and Ridley, 1987] used here is limited and undoubtedly does not capture the variability observed during the storm by the GUVI instrument on the TIMED satellite. This lack of fidelity in the high-latitude inputs is probably responsible for much of the discrepancy between the model and the ionosonde data. One of the coauthors (GC) is a coinvestigator for the GUVI instrument and is well aware of the fabulous images being produced by that instrument, showing the auroral structure that would ideally be captured in global models such as the TIMEGCM. However, these models are currently limited to a coarse horizontal grid resolution that is incapable of representing the detailed auroral structure. It appears that for this storm the statistical auroral oval obtained from NOAA and DMSP satellites was much different from the near-instantaneous auroral images obtained from the GUVI instrument. We must await greater computational power before models with sufficiently high resolution can be developed. There is an ongoing debate within the community about how much of the high-latitude heating and momentum forcing is being missed because of our inability to capture detailed structure in the high-latitude electric fields and particle precipitation. 4. Conclusions [35] The ionospheric sounding observations carried from two locations PAL (near equatorial) and SJC (low latitude) in the Brazilian sector and two locations HCM (near equatorial) and OKI (low latitude) in the East Asian sector, obtained during the month of October 2003 on geomagnetically disturbed days, which included two major geomagnetic storms and one intense geomagnetic storm, are presented and discussed in this work. The two sectors are separated by about 12 hours in local time. The principal results from the investigation are presented below: [36] 1. The two major storms have rapid decrease in the Dst index at 1900 UT (the first major storm has another earlier intense Dst variation). It is usually the time of the strong Dst decrease that is the important feature related to the storm-time effect. In Brazil the storms began just before nighttime (dusk time) and the stations show stronger prereversal uplifts. This caused a nighttime repeat of the equatorial anomaly effects with enhanced ionization at SJC. For the stations HCM and OKI the major storms began during the predawn period. The storm effects result in unusual uplifting of the F layer. [37] 2. It should be pointed out that the two longitudinal sectors investigated in the present study clearly show the global nature of the storm-time effects. However, the response to the storm-time effects is also associated with the local time in the two sectors. The present investigations show that there are both similarities and differences in the storm-time response in the two sectors. [38] 3. The present investigations also show the onset of spread F in the postsunset period during the existing ionospheric disturbance dynamo conditions, prior to the prompt penetration of magnetospheric electric field. [39] 4. Although in the East Asian sector, HCM, and OKI are located fairly close in longitude, with only 2 hour difference in local lime, on occasions the storm-time responses have been very different. [40] 5. Some differences in the latitudinal response of the F region were also observed in the two sectors. [41] 6. Both positive and negative storm phases have been observed at all the four stations. [42] 7. It is felt that more simultaneous studies using ionospheric sounding observations from several longitudinal zones will be important for space weather studies. [43] 8. This study has also provided the opportunity to validate a global 3-D model under extreme storm conditions. A comparison of the ionospheric parameters obtained from the TIMEGCM model runs and the observed ionospheric parameters at the four locations shows a reasonable agreement during the quiet periods. In addition, the model reproduces many of the observed features of the storm-time response, including height variations in the Brazilian sector. However, during the storm periods, there are also some considerable differences between the model and the observed fof2 and height variations. In particular, the model exaggerates the global negative storm phase and extends it for too long into the observed recovery phase. Finally, the peak heights predicted for the Asian sector are too high and probably correspond to exospheric temperatures that are too large. These discrepancies in the model predictions are likely caused by the rather crude representation of the high-latitude inputs during the storm. There have been very few validation studies for actual storm data, and the investigation reported here is an important addition, showing where the models can be improved. Following future model improvements, we anticipate the new model being run for the same storm conditions to validate the improvements. [44] Acknowledgments. Thanks are due to J. Kozyra for kindly providing helpful information related to the October 2003 events and to Ronaldo Marins de Carvalho, Observatório Nacional, Rio de Janeiro, Brazil, for kindly providing the magnetometer observations carried out at Vassouras, Brazil. Thanks are also due to the Brazilian funding agencies CNPq and FAPESP for the partial financial support through grants / 2003/7, /2003-8, and / (CNPq) and 1998/ and 2002/ (FAPESP). One of us (GC) was supported by NASA grants NAG , NAG and NNG04GN04G. [45] Arthur Richmond thanks Ivan Kutiev and another reviewer for their assistance in evaluating this paper. References Abdu, M. A. (1997), Major phenomena of the equatorial ionosphere-thermosphere system under disturbed conditions, J. Atmos. Sol. Terr. Phys., 59, Abdu, M. A., I. S. Batista, I. J. Kantor, and J. H. A. Sobral (1982), Gravity wave induced ionization layers in the night F-region over Cachoeira Paulista (22 S, 45 S), J. Atmos. Terr. Phys., 44, of 15