JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, A03301, doi: /2009ja014788, 2010

Transcription

1 Click Here for Full Article JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115,, doi: /2009ja014788, 2010 Scintillation producing Fresnel scale irregularities associated with the regions of steepest TEC gradients adjacent to the equatorial ionization anomaly M. T. A. H. Muella, 1,2 E. A. Kherani, 1 E. R. de Paula, 1 A. P. Cerruti, 2 P. M. Kintner, 2 I. J. Kantor, 1 C. N. Mitchell, 3 I. S. Batista, 1 and M. A. Abdu 1 Received 16 August 2009; accepted 5 October 2009; published 4 March [1] Using ground based GPS and digital ionosonde instruments, we have built up at latitudes of the equatorial ionization anomaly(eia),inthebraziliansector,a time evolving picture of total electron content (TEC), L band amplitude scintillations, and F region heights, and we have investigated likely reasons for the occurrence or suppression of equatorial scintillations during the disturbed period of November During the prestorm quiet nights, scintillations are occurring postsunset, as expected; however, during the storm time period, their spatial temporal characteristics and intensity modify significantly owing to the dramatic changes in the ionospheric plasma density distribution and in the temporal evolution of TEC. The two dimensional maps showing both TEC and amplitude scintillations revealed strong evidence of turbulences at the Fresnel length (causing scintillations) concurrent with those regions of steepest TEC gradients adjacent to the crests of the EIA. The largest density gradients have been found to occur in an environment of increased background electron density, and their spatial distribution and location during the disturbed period may differ significantly from the magnetic quiet night pattern. However, in terms of magnitude the gradients at equatorial and low latitudes appear to not change during both magnetic quiet and disturbed conditions. The scenarios for the formation or suppression of scintillation producing Fresnel scale irregularities during the prestorm quiet nights and disturbed nights are discussed in view of different competing effects computed from numerical simulation techniques. Citation: Muella, M. T. A. H., E. A. Kherani, E. R. de Paula, A. P. Cerruti, P. M. Kintner, I. J. Kantor, C. N. Mitchell, I. S. Batista, and M. A. Abdu (2010), Scintillation producing Fresnel scale irregularities associated with the regions of steepest TEC gradients adjacent to the equatorial ionization anomaly, J. Geophys. Res., 115,, doi: /2009ja Introduction [2] Equatorial spread F plasma bubble (EPB) irregularities are depleted plasma density structures aligned with the magnetic flux tubes. The mechanism responsible for their generation is considered to be the Rayleigh Taylor plasma instability (RTI) [see Fejer et al., 1999; Abdu, 2005,and references therein]. As the plasma irregularities rise up above the magnetic equator, a subsequent cascading by secondary instability processes sets up the development of a hierarchy of irregularities with different scale sizes and characteristics [Haerendel, 1973; Muralikrishna, 2000; Abdu, 2005]. Depending on the apex height reached above 1 Divisão de Aeronomia, Instituto Nacional de Pesquisas Espaciais, São José dos Campos, Brazil. 2 School of Electrical and Computer Engineering, Cornell University, Ithaca, New York, USA. 3 Department of Electronic and Electrical Engineering, University of Bath, Bath, UK. Copyright 2010 by the American Geophysical Union /10/2009JA014788$09.00 the equator, the EPBs can extend/expand along the geomagnetic field lines to the latitudes of the equatorial ionization anomaly (EIA) crests, a region of increased background electron density and sharp density gradients. Experimentally, Basu et al. [2001] noticed through VHF/UHF scintillation effects that equatorial spread F (ESF) bubble events at latitudes of the EIA might be associated with the regions of strong density gradients. Motivated by the observations of Basu et al. [2001], Keskinen et al. [2003] showed from numerical simulation techniques that bubble like structures could be generated at latitudes of the EIA with extremely high ionospheric density gradients. Over South America the first attempt to relate the presence of irregularities (through GPS L band scintillation effects) with the large scale horizontal electron density gradients surrounding the anomaly region was made by Ledvina et al. [2004]. Ledvina et al. [2004] used total electron content (TEC) images and scintillation data obtained across Brazil to investigate such a relation. Muella et al. [2007a] and de Paula et al. [2008] showed from twodimensional ionospheric imaging, specific relationships between the time dependence and dynamics of small scale ionospheric structures (causing L band scintillations) and the 1of19

2 large scale horizontal plasma density structures (through TEC data). Muella et al. [2008a] reported from observations over conjugate sites in Brazil, evidence that Fresnel length irregularities causing GPS L1 frequency amplitude scintillations could be strongly correlated with the large TEC gradients at the boundaries of the enhanced anomaly crests. Also recently, Muella et al. [2007b, 2008b] reported that during a storm time period such a relationship becomes more complicated owing to the large spatial changes of the ionospheric plasma, which can modify drastically the intensity and the spatial distribution of the scintillations. The present paper explores the capability for imaging the ionosphere in the zone adjacent to the crests of the EIA to provide evidence for the relationship between GPS L band scintillations and the regions of steep TEC gradients, as well as their spatial and temporal variability during a case of very intense geomagnetic storm Horizontal Ionospheric Density Gradients [3] A characterization of horizontal ionospheric density gradients and their effects on transionospheric radio wave propagation has been demonstrated in many studies at different latitudes. However, only a limited number of works have broached the relationship among large scale and smallscale ionospheric properties, such as TEC and Fresnel length irregularities. As a primary effect, it is well known that the presence of density gradients in the ionosphere introduces bias into the GPS receivers carrier smoothing filter output [Jakowski et al., 2005]. As a consequence, the phase ambiguities resolution can be severely impacted when processing data received from satellite based navigation and positioning systems. In this way, the ionospheric corrections used to improve current positioning/navigation applications based on Global Navigation Satellite Systems (GNSS) may experience serious degradation of their performance and reliability [Jakowski et al., 2001]. Furthermore, at regions of the ionosphere with larger background electron density and increased ionospheric gradients, the formation of small scale irregularities seems to be more prone to occur [Keskinen et al., 2003; Ledvina et al., 2004; Muella et al., 2008a]. The presence of such irregularities imposes amplitude and phase scintillations into the satellite radio wave signals and also affects the positional and navigational accuracy [Kintner et al., 2007]. [4] The ionospheric plasma density gradients are known to be generally associated with the daily variations of the electron density. Therefore, an adequate knowledge about their spatial and temporal variability and their maximum expected conditions during geomagnetic disturbances can provide essential information for the technical and scientific communities. Hence, irregularity and propagation models [e.g., Costa and Basu, 2002] may consider these horizontal electron density gradients as an important parameter to predict ionospheric scintillation and to mitigate positioning errors on GNSS based applications. [5] It is normally found in the literature that the spatial plasma density gradients can be represented by means of TEC changes per latitude or longitude (TECU deg 1 )orby their changes in distance (TECU km 1 ), for which 1 TECU corresponds to electrons m 2. The ionospheric spatial density gradients can also be estimated from F region peak electron density (N m F 2 ) data by using a chain of radio soundings [Ram et al., 2006]. In general, the classification of the spatial electron density or TEC horizontal gradients is given as latitudinal (north south) and longitudinal (east west) [Strangeways, 2000; Jakowski et al., 2004]. The first techniques used to obtain the TEC gradients have investigated the midlatitude regions and used differential Doppler measurements of signals received from low Earth orbit and geostationary satellites (see review by Mendillo [2006] and references cited therein). At those latitudes (middle) the regions of steeper gradients have been associated with the borders of the ionospheric main trough [Vo and Foster, 2001]. The effects of electron density gradients on the most recent technologies of Earth space communications and GNSS applications, during quiet and disturbed periods, have been presented or reviewed in the works of Jakowski et al. [2004] and Radicella et al. [2004]. [6] The horizontal ionospheric gradients have also been demonstrated at equatorial and low latitude regions in order to define the spatial and temporal F region electron density and TEC gradients under different conditions. Differently from middle latitudes, the contour maps of Keroub [1976] clearly showed that the electron distribution at equatorial/ low latitude regions are irregular and that the slopes of equal density lines are not moderated, revealing the presence of tilts in the horizontal gradients. The diurnal pattern of electron density and TEC variations at equatorial and low latitudes are directly associated with the variability of the fountain effect, which may cause the crests of the equatorial anomaly to move backward and onward in latitude [Huang, 1984]. The crests of the anomaly when fully developed tend to be aligned approximately along the geomagnetic field lines, then resulting in the presence of strong spatial changes of the ionospheric plasma in the zone adjacent to the equatorial anomaly region [Franke et al., 2003; Rao et al., 2006]. On the basis of rocket ionospheric electron density data obtained from in situ experiments at the Brazilian equatorial ionosphere, Sobral et al. [1997] estimated the horizontal density gradients from predicted photoemission rates of the atomic oxygen red (l = 630 nm) and green (l = nm) lines. They calculated the horizontal gradients on the upleg and downleg trajectories of the rocket and demonstrated that the gradients change with altitude. Nava et al. [2007] estimated the horizontal gradients in the low latitude ionosphere over Brazil from mapping function errors of slant TEC to vertical TEC based on a coinciding pierce point technique [Radicella et al., 2004]. However, the application of the technique seemed difficult itself in several occasions owing to the low quantity of data from different satellite signals with coinciding puncture point. They showed strong electron density and TEC gradients during both geomagnetic quiet and disturbed conditions, which suggests that the quiet low latitude ionosphere might not be so different from the low latitude ionosphere during geomagnetic disturbances. Although not stated by the authors, the low quantity of coinciding piercing points found in their investigation might be associated with important features of the equatorial and low latitude ionosphere, that is, the presence of vertical velocities, tilts in the horizontal gradients, and irregularities in the plasma distribution. Recent investigations by Rao et al. [2006] discussed the close relationship between the occurrences of intense scintillations at L band frequency and the larger spatial electron density at the equa- 2of19

3 Figure 1. Geomagnetic indices Kp, AE, and SYM H and high resolution interplanetary and solar data B, B z, and V sw (see text for more details) for the space weather period of November torial ionization anomaly crest regions. Ray et al. [2006] showed for the Indian sector that the steep latitudinal gradient of TEC observed during the afternoon hours in the region between the anomaly crest and the equatorial trough could be taken as a precursor to the development of postsunset scintillations on transionospheric links. Analogously but analyzing ionosonde data, Ram et al. [2006] observed an enhancement in the ionization anomaly gradient prior to the onset of intense VHF scintillations. More recently, Abdu et al. [2009a] described that the asymmetries in the ambient ionization and density gradients over conjugate sites could be the possible cause of the corresponding asymmetries observed in the echo intensities of digital ionosonde data. [7] During disturbed periods, intense magnetospheric electric fields that penetrate promptly toward low latitudes can lead to an intensification of the normal eastward electric field generated from ionospheric, wind driven E and F region dynamo. Then, owing to the enhanced fountain effect, the ionospheric plasma is shoved out from the equator to latitudes/altitudes much higher than that observed during magnetic quiet conditions. This process results in the formation of enhanced ionization at higher latitudes and in an ionization trough at the dip equator, respectively, with TEC average values exceeding or decreasing drastically by a significant amount. The enhanced fountain effect due to prompt penetration of magnetospheric electric fields causes an expansion of the EIA and thus results in an enhancement of poleward horizontal convection effects at equatorial and low latitudes [Tsurutani et al., 2004]. As a consequence of the largest expansions of the fountain effect, the ionosphere at latitudes surrounding the geomagnetic equator and adjacent to the enhanced anomaly crests tends to exhibit remarkable spatial gradients of the electron distribution. Owing to the strong poleward gradients, the TEC structures respond accordingly and can bring high values of electron density distribution into the midlatitude ionosphere [Fedrizzi et al., 2005; Mendillo, 2006]. Additionally, thermosphereionosphere coupling effects (such as horizontal neutral winds) can modify the plasma surrounding the crests of the equatorial anomaly, affecting the location, width, and magnitude of the plasma density gradients. Besides the magnetic activity effects, the spatial electron density and TEC gradients can be expected to depend also on solar activity, season of the year, latitude, and local time Geomagnetic Superstorm of 20 November 2003 [8] The case studied in this work, which took place on 20 November 2003, is considered one of the major space weather events of solar cycle 23 [Gopalswamy et al., 2005]. Solar disk images from Solar and Heliospheric Observatory (SOHO) revealed that the sunspot active region (AR) 501 located at the near centered longitude of the Sun (N00E18) was the source of this very intense magnetic storm [Alex et al., 2006; Gopalswamy et al., 2005]. The active region 501 produced an M class solar flare at 0723 UT on 18 November [Alex et al., 2006]. The associated full halo coronal mass ejection (CME) evolved into a magnetic cloud (MC) and moved with a speed of 1100 km 1. The arrival of the shock into the Earth s environment was detected at the Lagrangian point L1 by the instruments on board SOHO around 0740 UT on 20 November. During the shock the Advanced Composition Explorer (ACE) spacecraft detected an MC speed of 730 km s 1. Figure 1 shows the solar wind velocity (V sw ), the MC s magnetic field ( B ), and the interplanetary magnetic field (IMF) B z component in GSM coordinates measured from ACE. The geomagnetic indices Kp (3 hourly planetary index), AE (auroral electrojet intensity), and SYM H (1 min high resolution intensity of the ring current) obtained from the World Data Center for Geomagnetism, Kyoto ( u.ac.jp), are also shown in Figure 1. Figure 1 illustrates that the shock resulted in the storm sudden commencement (SSC) at 0803 UT on 20 November. The compression of the Earth s magnetic field led the corresponding horizontal component 3of19

4 Figure 2. Location of the observation sites used in this study with single frequency (stars) or dual frequency (triangles) GPS receivers. to attain a maximum negative SYM H index value of about 490 nt around 1818 UT. After the SSC the 3 hourly Kp index increased from 6 + to 9 between 1500 and 2100 UT on 20 November, and the AE index enhanced drastically to values of above 2000 nt. The IMF B z turned abruptly to the south 2.5 h after the SSC, giving rise to the electrojet activity over broad regions and beginning a process of substorm activity. The IMF B z reached its minimum value of about 55 nt at 1500 UT. After its minimum excursion the SYM H index presented an extended recovery phase until November. According to Gopalswamy et al. [2005], the 20 November space weather event may be considered a superstorm owing to the large magnitude of B ( 56 nt) and also owing to the high inclination of the MC axial field nearly antiparallel to Earth s magnetic field. [9] The response of the ionosphere thermosphere system from equatorial to high latitudes during the severe storm of 20 November 2003 has been investigated by many authors using ground based and space based instruments. For example, Meier et al. [2005] presented the first results of global thermospheric temperature and composition changes during the 20 November magnetic storm by comparing parameter data extracted from the Global Ultraviolet Imager (GUVI) instrument on board the Thermosphere Ionosphere Mesosphere Energetics and Dynamics (TIMED) satellite, with run predictions from Advanced Space Environment (ASPEN) Thermosphere Ionosphere Mesosphere Electrodynamics General Circulation Model (TIMEGCM). Yizengaw et al. [2006] reported an unusual response of the dayside topside ionospheric density to this magnetic storm, when comparing TEC data obtained from groundbased GPS receivers with those detected by accelerometer measurements of the low Earth orbiting ( 400 km) Challenging Minisatellite Payload (CHAMP) satellite. The former presented larger dayside TEC enhancement, while CHAMP did not show any increase in TEC. Crowley et al. [2006] explained such TEC enhancements observed over Europe and Africa sectors in terms of changes in the thermospheric neutral winds, [O]/[N 2 ] ratio, and ionospheric electron density modeled by the TIMEGCM and obtained from the Ionospheric Data Assimilation Three Dimensional (IDA3D) algorithm [Bust et al., 2004]. For thesamestorm,yin et al. [2006] used two different data assimilation algorithms to demonstrate the electric field driven uplift in the ionospheric layer height over both the European and North American regions. Bruinsma et al. [2006] reported for the 20 November storm, anomalous large scale wavelike oscillations ( 1000 km) with comparable day night amplitudes that extended from auroral to low latitude regions in both hemispheres at speeds of m s 1. Such wavelike structures, in the form of gravity waves or traveling atmospheric disturbances (TADs), have also been observed by Becker Guedes et al. [2007] on ionospheric parameters scaled from ionograms recorded by ionospheric soundings installed at Brazilian equatorial and low latitude stations. Basu et al. [2007] studied for this severe storm the response of the ionosphere in the dusk sector, and they reported that the upward plasma drift at F region heights due to simultaneous electrodynamic effects (caused by prompt electric field penetration and F region dynamo) may be intensified in the longitudes of the South Atlantic Magnetic Anomaly (SAMA) region owing to increases in the east west conductivity gradient caused by energetic particle precipitation. 2. Database and Method of Analysis [10] In this study, L band amplitude scintillation data at MHz were collected by seven single frequency GPS receivers installed at different locations covering northto south and central eastern sides of the Brazilian territory. The distribution of the receivers is such that it includes data close to the magnetic equator, at the equatorward edge of the northern crest of the EIA, and beyond the southern anomaly crest (see Figure 2). One of the stations (São Luís) is located close to the magnetic equator. Another station (Manaus) is located north of the magnetic equator and under the equatorward edge of the equatorial anomaly. Five of the stations (Cuiabá, Cachoeira Paulista, São José dos Campos, Palmas, and São Martinho da Serra) are situated south of the magnetic equator. Cuiabá is located between the magnetic equator and the crest of the anomaly, São Martinho da Serra (S. M. da Serra) is situated well south, close to the poleward edge of the anomaly, and the other stations are close to the southern crest. GPS ionospheric scintillation monitors developed by the Electrical Engineering group from Cornell University (named CASCADE/SCINTMON; cornell.edu/spaceweather.html) were used to monitor the scintillations at the L1 frequency. The scintillation monitors in Brazil have been operated continuously and comprise the 4of19

5 Table 1. Geographic Coordinates of the Sites With Ground Based Scintillation Monitors and Their Overhead Geomagnetic Coordinates for an Ionospheric Piercing Point at 350 km Height Geographic Geomagnetic Station Latitude ( S) Longitude ( W) Dip Latitude Dip Angle (deg) Declination (deg) Manaus N São Luís a S Cuiabá S C. Paulista a S S. J. Campos S Palmas S S. M. da Serra S a Digital ionosonde instruments exist at these observatories. GPS array managed by the Instituto Nacional de Pesquisas Espaciais (INPE). These scintillation monitors compute accurate values of the S 4 scintillation index for each interval of 60 s from the normalized RMS deviation of signal power strength (for more details, see the work of Beach and Kintner [2001]). Table 1 lists the geographic and geomagnetic coordinates of the seven sites with single frequency GPS receivers used in this experiment. [11] Up to 70 dual frequency GPS receivers installed across the American continent (North, Central, and South Americas) were used to obtain TEC measurements from the ionospheric delay between the L1 ( MHz) and L2 ( MHz) frequencies. The receivers comprise different arrays, such as the International GPS Service for Geodynamics (IGS), the Continuously Operating Reference Station (CORS), the Rede Brasileira de Monitoramento Contínuo (RBMC), and others. The observation and navigation files for most of the sites can be obtained through file transfer protocol (FTP) from Scripps Orbit and Permanent Array Center (SOPAC) and California Spatial Reference Center (CSRC) Garner GPS archive ( Data from RBMC can be obtained from the Data Center for Geosciences of the Brazilian Institute of Geography and Statistics (IBGE) at Precise orbit files and other relevant GPS data files can be obtained from IGS/Jet Propulsion Laboratory and from National Oceanic and Atmospheric Administration (NOAA) home pages at and respectively. In the present study the scintillation and TEC data were obtained on the nights of November 2003 (hereinafter also referred to as prestorm quiet nights) and during the major space weather event that occurred throughout November The scintillation data (S 4 index) recorded by the CASCADE/SCINTMON receivers and the TEC data were processed for each of the satellite passes with an elevation angle 30. This angle mask ensures that only the local ionosphere over the observation site is being sampled and reduces the effects of low elevation angles, such as multipath and tropospheric effects. TEC maps covering the latitude range of in the American longitudinal sector with sufficient accuracy of a few TECU were generated using the Multi instrument Data Analysis System (MIDAS) algorithm of Mitchell and Spencer [2003]. The MIDAS algorithm includes computation and calibration of the differential phase values by constrained inversion to produce four dimensional electron density images. The images were generated using a (latitude longitude) grid size. Then the derived vertical TEC maps were used to evaluate qualitatively and quantitatively the horizontal ionospheric density gradients. However, it has to be mentioned that the GPS receivers are unevenly distributed across the American continent, which limits the accuracy to monitoring large scale horizontal ionization structures over some regions. Therefore, in the present work a small region over the Brazilian territory covered with a denser and evenly distributed chain of stations from RBMC has been used for quantitative estimations of density gradients with low maximum absolute percentage TEC errors. This region has been chosen following the recommendations reported in the studies of Materassi and Mitchell [2005] and Zapfe et al. [2006], and their simulations indicate that the absolute TEC errors will not exceed 16%. [12] With the aim to investigate specific relationships between the large scale plasma density structures and the irregularities at the Fresnel length, GPS amplitude scintillation data are coupled to the two dimensional TEC maps. The scintillations are represented in the maps by the ionospheric pierce points for each GPS satellite signal, and for the seven GPS scintillation monitors used in this experiment, over 30 piercing points can be depicted in the images. Figure 2 shows the distribution of the GPS receivers in South America for TEC measurements (triangles) and amplitude scintillation data (stars). In Figure 2 the white box indicates the region with high accuracy for imaging TEC gradients. [13] Information about the ionospheric F region peak (h m F 2 ) and base heights (h F) and critical frequency (f o F 2 ) parameters during the space weather period of November 2003 were scaled from 15 min ionograms recorded by two digital ionosondes operating at two different sites: one installed close to the magnetic equator at São Luís and another in the low latitude station of Cachoeira Paulista. The coordinates of these two stations with simultaneous ionosonde data available are also indicated in Table Observational Data Sets [14] Figure 3 presents the time variations (in UT) of three important ionospheric parameters scaled every 15 min from ionograms recorded at São Luís and Cachoeira Paulista during the period of November Both stations have the same local time with UT = LT+3 h. Figures 3a and 3b show the peak height (h m F 2 ) of the F region (scattered asterisks) observed over São Luís and Cachoeira Paulista (C. Paulista), respectively. The base height h F (scattered triangles) for the equatorial station of São Luís is also shown 5of19

6 Figure 3. The height variations of (a) h m F 2 and h F over São Luís, (b) h m F 2 over C. Paulista, and (c) f o F 2 over both stations, observed during the space weather period of November The intervals with hatched bars indicate nighttime, and the gray bars denote the times when any type of spread F was registered in the ionograms. in Figure 3a. Figure 3c shows the F region critical frequency (f o F 2 ) observed at São Luís (asterisks) and C. Paulista (triangles) for the same period. The hatched intervals in Figures 3a 3c indicate nighttime. The gray bars in Figures 3a and 3b indicate the time period for which any type of spread F was observed in the ionograms. [15] Figure 4 shows the spatial distribution of the S 4 amplitude scintillation index along the satellite tracks over Brazil (GPS signal ionospheric pierce point (IPP) projections for an elevation mask of 30 ) during the nighttime interval (Brazilian Standard Time, BST) from 1900 to 0600 LT ( UT) and throughout November The paths were colored according to the scintillation level (color scale on the right side of the plots). The plots in Figure 4 depict the scintillation activity distribution around the dip equator and at off equatorial latitudes during the prestorm nights (18 20 November) as well as during the stormrecovering phase (disturbed) nights (20 23 November). On 18 November no data were available at Cuiabá owing to power failure. Between 20 and 22 November no measurements were performed by the receiver installed at S. M. da Serra. On 21 and 22 November, problems with storage data in Palmas limited our observations at that latitude. In an ambient of very intense scintillations and/or extremely high ionospheric gradients, a partial or total interruption in the received signal may occur owing to a loss of lock in the phase channel of the GPS receivers. This factor explains why in Figure 4 sometimes at a given station the symbols representing the scintillations appear to be isolated or interrupted rather than to form contiguous tracks. [16] To investigate the relation of Fresnel scale irregularities causing GPS L1 amplitude scintillations with the overall evolution of equatorial TEC distribution, particularly at the southern anomaly crests, coupled maps of vertical TEC and S 4 index were generated for each 1 min interval during the prestorm period and during the disturbed period of November Figure 5 shows random intervals of daytime (from 1800 UT) and mainly nighttime coupled TEC and scintillation maps through the geomagnetic quiet days preceding the onset of the November 2003 superstorm. The hours in the images refer to universal time UT. TEC is shown in color according to the color bar in TECU on the right. The white triangles on the maps indicate the position of the sites with dual frequency GPS receivers whose data were used to generate the TEC images. For the amplitude scintillations the ionospheric pierce points (at 350 km) of each GPS signal (satellites with elevation angle 30 ) detected by the scintillation monitors (white squares) are shown, initially, as small black circles (S 4 < 0.2). When the scintillation index S 4 is above the noise and multipath effect levels (S 4 0.2) [Muella et al., 2009], the black dots change to white circles whose sizes are made proportional to the magnitude of the S 4 index. The largest circles represent S 4 equal to 1 or even higher (until approximately S 4 = 1.4). S 4 index exceeding 1 can occur when fluctuations in the refractive index of the ionosphere become much larger, and the computed S 4 is probably not due to a single scatter of the radio wavefield (see the work of Rino [1979] for more details about the multiscattering regime). [17] Figure 6 presents the nighttime variation, magnitude, and position in geomagnetic latitude of the maximum absolute values of the steepest latitudinal TEC gradients (in TECU deg 1 ) during the prestorm nights of November 2003 (Figure 6a), during the first night following the maximum negative value of SYM H, November (Figure 6b), and during the recovery phase nights of November (Figure 6c). The latitudinal TEC gradients were obtained from the vertical TEC maps for the region of lower TEC error denoted as a white rectangle in Figure 2. The TEC gradients were determined on grids with an extension of 2 in longitude 6of19

7 Figure 4. GPS L band amplitude scintillation (S 4 index) distribution in geographic latitude versus longitude during the nights ( LT Brazilian Standard Time, BST) of November The isolines represent the dip latitude. The satellite paths were colored according to the S4 intensity shown in the color scale on the right. 7of19

8 Figure 5. Geographic latitude versus longitude of coupled TEC and scintillation (S 4 ) images obtained during the prestorm quiet nights of and November Black dots represent the ionospheric pierce points of each GPS signal (elevation 30 ). When the signal scintillates (S 4 0.2), the ionospheric pierce point is depicted as a white circle. The size of the circles is made proportional to the S 4 index. and 1 in latitude. By fixing the grid that corresponds to the geographic longitude of 45 W (close to the longitude of São Luís and C. Paulista), the TEC gradients were derived over each distance of 1 in geographic latitude and expressed as TECU deg 1. Then the magnetic latitude that corresponds to the center of that grid of maximum derived TEC gradient was used to denote its location. The maps in Figure 7 are similar to those depicted in Figure 5 but for the disturbed nights of November. For a better demonstration of the relation between the scintillation distribution and the TEC structures at the anomaly region, the maps in Figure 7 during the day of the storm commencement (20 November) and the days following the SSC (21 23 November) are shown in random intervals after 1800 UT. 4. Discussion 4.1. Ionospheric F Region Height and Electron Density Variations [18] The F region base (h F) and peak (h m F 2 ) height variations throughout November 2003, over the equatorial station of São Luís, are shown in Figure 3a. Over the low latitude station of Cachoeira Paulista, only the F 2 peak height variation is shown in Figure 3b. The initial response of the ionosphere due to the storm can be seen on 20 November, when h m F 2 above the equatorial station São Luís showed a more dramatic uplift after 1200 UT (0900 LT BST) when compared to the prestorm days (18 19 November). The h m F 2 ascent between 1300 UT (1000 LT BST) and 1800 UT (1600 LT BST) on 20 November, when the SYM H index is decreasing rapidly, is probably associated with the penetration of eastward magnetospheric electric fields into the equatorial ionosphere. The intensification of the eastward equatorial electric field during daytime hours at the early stage of the storm (when IMF B z goes to southward) may be considered to be caused mainly by a sudden increase of the polar cap potential [Fejer et al., 1990]. Over the low latitude station of C. Paulista a rise in the F peak height was also observed, but the uplifts were more lasting and with smaller amplitude than those over São Luís. 8of19

9 Figure 5. (continued) [19] Following the onset of the storm, the large increase in the auroral electrojet (AE) index gives an indication of the amount of energy that was injected at high latitudes into the ionosphere and thermosphere. The global neutral winds respond to this energy input, and consequently, through the action of ionospheric dynamo process, drive disturbance dynamo electric fields into the equatorial ionosphere. It is noticed from h F and h m F 2 over São Luís (Figure 3a), during the postsunset period on 20 November, that the prereversal enhancement (PRE) of the vertical drifts is being suppressed. As shown in the model of Scherliess and Fejer [1997], during daytime and evening sectors the disturbance dynamo effect is to reduce the upward plasma drift and the evening PRE in the vertical drift/zonal electric field by the action of westward electric fields and westward disturbance neutral winds. Other possibilities for PRE suppression can be attributed to the penetration electric field of westward polarity arising from shielding layer electric field (overshielding processes) associated with B z turning northward (probably not the case in this study, as deduced from Figure 1) and/or auroral electrojet recovery [e.g., Kelley et al., 1979; Kikuchi et al., 2008]. As a consequence of the PRE suppression, the generation of equatorial irregularities was inhibited on this night (Figure 3, hatched interval), as noticed in Figure 3a from the absence of the gray bar during the first evening hours. Other indications that the PRE has been suppressed on this night can be seen in Figure 3c from the absence of f o F 2 enhancement (after sunset) in the lowlatitude station of C. Paulista. [20] The f o F 2 (proportional to the square root of F region peak electron density, N m F 2 ) response during the main phase of the storm shows a maximum of 16 MHz ( electrons m 3 ) over the low latitude station of C. Paulista around 1900 UT (1600 LT) on 20 November, whereas over the equatorial station of São Luís a precipitous drop is observed a few minutes earlier. This implies that an enhanced vertical ionospheric drift during daytime due to fountain effect intensification drained the plasma over São Luís and deposited it at latitudes higher than that of C. Paulista, resulting in a poleward movement of the southern anomaly crest. However, after its observed peak value a continuous drop of f o F 2 over C. Paulista suggests that ionization surrounding the region of the southern anomaly crest starts to subside. [21] On 21 November during the recovery phase of the storm, the peak (h m F 2 ) and base (h F) height parameters 9of19

10 Figure 6. Nighttime variations and geomagnetic latitude of the steepest latitudinal TEC gradients (in TECU deg 1 ) over the region of lower TEC error denoted as a white rectangle in Figure 1 for (a) the prestorm nights of November 2003, (b) the first disturbed night following the main phase of the storm (20 21 November 2003), and (c) the recovery phase nights of November over São Luís and C. Paulista present a rapid uplifting around 0300 UT (local midnight). As registered in the ionograms, this uplifting of the F region ( 60 m s 1 ) at São Luís (up to 650 km) coincided with the generation of short lived equatorial (frequency type) spread F (Figure 3a) at 0445 UT. Besides the large vertical drift E/B, as the layer moved upward to higher altitudes (where the collision frequency (n in ) is small), the presence of a steep bottomside electron density gradient probably created favorable conditions for the instabilities to grow. From the true height profiles of the ionograms recorded at São Luís during the spread F onset time, we estimated the electron density gradient scale length given by L N 1 = (1/N)(dN/dh) and the vertical drift velocity. Then we inferred the linear growth rate (g L ) of the instability by using the simplest equation given by g L = L N 1 (E/B + g/n in ) [Ossakow, 1981], where N is the background electron density, E is the eastward electric field, B is the magnetic flux density, and g is the gravitational acceleration. The values of n in were obtained from standard atmospheric models. During the spread F onset time at 0445 UT on 20 November, g L was s 1, which is much larger than its value of s 1 inferred, for example, during the onset of the spread F in the postsunset hours of the quiet night 18 November. At 0630 UT a strong negative phase is noticed in f o F 2, and after that the rapid F region descent ( 110 m s 1 ) coincides with the cessation of the irregularities at 0830 UT, when g L was estimated at s 1. Furthermore, the F region rapid uplift also coincides closely with the time of northward IMF B z (Figure 1). According to Kelley et al. [1979], a dusk todawn electric field perturbation (eastward at night) can penetrate to lower latitudes in the situation of sudden reversals of the IMF B z from steadily southward to steadily northward. However, in the present storm the B z turning to north occurred after its recovering from southward excursion. Even so, it is possible that combined effects of disturbance dynamo and prompt penetration electric fields might be occasioning the rapid uplift of the ionosphere around local midnight. [22] Throughout November the f o F 2 variations at São Luís and C. Paulista presented different behavior compared to those observed during the prestorm quiet days. For example, Figure 3c shows from 0900 UT on 21 November until 0300 UT on 22 November, higher values of f o F 2 over São Luís than over C. Paulista, when the opposite should be expected. Even on days 22 and 23 November, the f o F 2 variations at both stations assumed comparable values during most of the time. This phenomenon observed in f o F 2 variations is consistent with the assumption of wind driven migration of large density molecular nitrogen (N 2 ) from higher to lower latitudes, which results in a loss process enhancement at F 2 layer heights. Consequently, the reduction in the density ratio between the atomic oxygen ([O]) and the molecular nitrogen ([N 2 ]) at latitudes of C. Paulista could explain the low values of f o F 2 observed over this station during the storm recovery phase days. Otherwise, over the equatorial station of São Luís the positive f o F 2 disturbance on 21 November is probably related to effects of equatorward thermospheric winds and traveling atmospheric disturbances but not to neutral composition changes. Evidence of dramatic neutral composition changes in the ionosphere over the Southern Hemisphere has been demonstrated from GUVI data by Meier et al. [2005]. They showed a severe depletion of O/N 2 column density ratio (SO/N 2 ) extending to the Brazilian longitudinal sector on 21 November, which is in accordance with the low f o F 2 values observed at C. Paulista. Becker Guedes et al. [2007] also reported negative phases in f o F 2 over two observatories located in the Brazilian sector, 10 of 19

11 Figure 7. The same as Figure 5 but for the nights of November of 19

12 Figure 7. (continued) which agrees with the storm induced surges of thermospheric neutral composition perturbation associated with depletion in [O]/[N 2 ] ratio. Other physical processes that might explain the observed ionospheric F 2 layer response over Brazil to this superstorm, which are not pointed out here, deserve further investigation GPS L1 Amplitude Scintillation Observations [23] The amplitude scintillation S 4 index in the LT ( UT) interval throughout November 2003 is plotted in Figure 4 in geographic latitude against longitude. In the quiet night of 18 November the S 4 index around the magnetic equator rarely exceeds 0.3, while at latitudes of Palmas and S. Martinho da Serra (around the southern crest of the EIA), it had larger values (approaching 1.0). However, it seems that at the eastern side (around C. Paulista and S. J. Campos) the L band scintillation is somewhat less intense (maximum S 4 of 0.6). In the quiet night of 19 November it is also clear from Figure 4 that the scintillation is more intense at the southern EIA crest (S 4 higher than 0.9), moderate at latitudes between the dip equator and the southern crest (S 4 between 0.5 and 0.6), and weaker over the equator (S 4 lower than 0.4). On the other hand, the scenario of the scintillation distribution during the storm recovery phase is fairly different. For example, in the first night after the SSC (on 20 November), Figure 4 shows that the scintillations were absent over the Brazilian territory, which corroborates with the absence of postsunset spread F observations from digital ionosondes at São Luís and C. Paulista (Figure 3) and also with Defense Meteorological Satellite Program (DMSP) F14 in situ satellite data measurements reported previously by Basu et al. [2007]. Unfortunately, the loss of GPS receiver data in stations such as Palmas, São Martinho da Serra, and C. Paulista limited our observations over their latitudes. However, the absence of scintillations at Cuiabá suggests that irregularities were not generated over the magnetic equator farther to the west of these stations. Scintillations with very short patch durations were observed over Manaus, which have been found to coincide closely with the time of large upward drift that occurred around 0300 UT on 21 November. [24] On the night of 21 November the pattern of the scintillation distribution when compared to the nights of November is somewhat different at dip latitudes of São Luís, Manaus, and C. Paulista. This is a result of a different plasma distribution, as will be evidenced from TEC observations in section 4.3. On this night the maximum scintillation intensity (down to 0.7) is observed over Cuiabá 12 of 19

13 Figure 7. (continued) station. On the night of 22 November, Figure 4 shows that the amplitude scintillation is mostly weak (S 4 < 0.3) at the dip equator and somewhat more intense (S 4 < 0.5) at low latitudes, which is probably the result of the weak spread F development and low plasma density at off equatorial latitudes (the latter is also revealed in Figure 3c). All these features, however, are evidenced in more detail in section 4.3 from the coupled TEC and scintillation maps Coupled TEC and Scintillation Images Typical Quiet Night Scenario [25] The first vertical TEC image in Figure 5 at 2300 UT (2000 LT BST) shows clearly the development of the crests of the EIA due to their postsunset resurgence (occasioned by the PRE of the vertical drift). A close symmetry development of both anomaly crests with comparable crest to trough TEC is observed during the evening. We noticed from results not presented in Figure 5 that 40 min before 2300 UT, weak scintillations began to be observed above the equatorial station of São Luís. Then, at 2300 UT we can see that the scintillations (Figure 5, white circles) appear in the image close to the equatorward edge of the southern and northern anomaly crests; however, they seem to be more intense surrounding the southern crest. Since the irregularities in the ionosphere are field aligned structures that tend to drift eastward, those structures causing scintillations at stations such as C. Paulista and São José dos Campos were probably generated to the west of these stations. It is possible to see in the sequence of images labeled from 2330 UT on 18 November to 0100 UT (2200 LT) on 19 November that the population of strong events of scintillations (close to the southern crest) is initially observed close to the equatorward (inner) edge and the crest of the anomaly and then tends to move to the poleward (outer) edge of the anomaly peak. However, strong levels of scintillation can eventually still be observed at the equatorward edge of the southern crest of the EIA. It is easily noticed from the images that the population of piercing points (IPPs) and scintillation distribution are higher close to the southern crest, owing to the fact that more receivers are installed at that latitudinal sector. Although reduced in terms of population events, the images in Figure 5 reveal that comparable levels of scintillation are occurring close to the northern crest. Moreover, Figure 5 shows that as the time elapses the scintillations tend also to be observed for those GPS signals puncturing the ionosphere in the west sector of the sky above the stations. Figure 5 suggests that the scintillation events for later evening and postmidnight hours can be associated with the plasma bubbles created farther to 13 of 19

14 Figure 8. (a, b) The ambient ion density profile n o for Cases 1 and 2 and (c, d) the evolution of maximum value of de for different values of parallel wind W k in Cases 1 and 2. the west of the stations. Hence, the scintillation events observed, for example, at 0300 UT (local midnight, BST) are probably associated with fossil bubbles. It is worthwhile mentioning that for the night of November the scintillations ceased around 0500 UT (0200 LT) over the Brazilian region (results not shown). [26] In the continuation of images in Figure 5 it is noticed that at around 2200 UT (1900 LT) on 19 November the postsunset resurgence of the anomaly is becoming clear and that the onset of the scintillations is observed close to the equator. At 2300 UT an asymmetry in the development of the EIA becomes evident, with the southern crest presenting comparatively larger TEC values. Simultaneously, strong events of scintillations are seen easterly at the inner edge and crest of the southern anomaly, whereas weak scintillations are detected close to the northern crest. This suggests that the most intense scintillations are more probable to be observed at that crest with larger absolute TEC values. As the crest s TEC subsided to values of TECU, strong scintillations continued to be observed close to the southern crest and also at the northern crest until 0100 UT (2200 LT) on 20 November. Surprisingly, even after the crest densities have fallen to small values, the images show that the most intense scintillations can occur throughout the anomaly crests and at both east and west sectors of the sky above the stations, which may indicate the control by other factors on the scintillation activity. However, as the night progress, the Fresnellength irregularities producing scintillations tend to decay faster, and at around local midnight (0300 UT), only the weakest scintillation events are still observed. [27] In Figure 6 the TEC gradients are shown throughout November during UT. The geomagnetic latitude, magnitude, and variations of the observed absolute steepest TEC gradients (TECU deg 1 ) during the prestorm nights of November 2003 are shown in Figure 6a. The latitudinal TEC gradients are observed in a narrow strip around São Luís and C. Paulista, which have almost the same geographic longitude (45 W) but different latitudes. The largest TEC gradients for the quiet nights have been found to vary between 3 and 4.5 TECU deg 1. It is most likely that these gradients are associated with the equatorward or poleward boundaries of the equatorial anomaly crests. In Figure 6a we can see that the largest gradients may occur between 2200 UT (1900 LT) and 0400 UT (0100 LT) with a clear hourly and day to day variability. This corresponds to the time period when the strongest scintillations events are observed in the images of Figure 5. After 0400 UT the latitudinal TEC gradients begin to decrease until their lowest nighttime values around 0800 UT (0500 LT), and following this decrease is the fast decay of the scintillations. 14 of 19