1 Large Geomagnetic Storms: Introduction to Special Section 2 3 N. Gopalswamy 4 NASA Goddard Space Flight Center, 5 Greenbelt, Maryland, USA 6 7 Abstract: Solar cycle 23 witnessed the accumulation of rich data sets that reveal various 8 aspects of geomagnetic storms in unprecedented detail both at the Sun where the storm 9 causing disturbances originate and in geospace where the effects of the storms are 10 directly felt. During two recent coordinated data analysis workshops (CDAWs) the large 11 geomagnetic storms (Dst < -100 nT) of solar cycle 23 were studied in order to understand 12 their solar, interplanetary, and geospace connections. This special section grew out of 13 these CDAWs with additional contributions relevant to these storms. Here I provide a 14 brief summary of the results presented in the special section. 15 16 1. Introduction 17 The coordinated data analysis workshops (CDAWs) have been serving as a forum 18 to analyze large and disparate data sets by members of the science community. Two such 19 CDAWs were conducted in March 2005 (George Mason University) and 2007 (Florida 20 State University) focusing on the set of all large geomagnetic storms (Dst < -100 nT) of 21 solar cycle 23 until the end of 2005. There were 88 large storms in all [Zhang et al., 22 2007]. The solar cycle 23 started in May 1996 and continued into 2008, although 23 occasional observations active regions belonging to cycle 24 have been made since 1 24 December 14, 2007. After 2005, there have been only two additional large magnetic 25 storms in cycle 23: one on April 14, 2006 with Dst ~-111 nT and the other on December 26 15, 2006 with Dst ~-146 nT. Thus the CDAW storms represent an almost complete set 27 for the whole solar cycle. It was possible to assemble atmospheric, ionospheric, 28 magnetospheric, interplanetary, and solar data on the 88 storms. The uniform and 29 extended data on coronal mass ejections (CMEs) and the inner corona (including coronal 30 holes) available from the Solar and Heliospheric Observatory (SOHO) mission has 31 facilitated the study of solar connection of geomagnetic storms with unprecedented 32 clarity. It must be noted that solar cycle 23 is the first cycle in which CME data are 33 available over the whole cycle since the first detection of CMEs in the early 1970s. The 34 availability of simultaneous space and ground based data covering the Sun-Earth space 35 has made the solar cycle 23 storms as one of the best set of events that could serve as 36 bench mark to compare storms of future and past cycles. 37 Papers constituting this special section fall into three groups addressing solar – 38 interplanetary phenomena, magnetospheric phenomena, and ionospheric phenomena 39 related to cycle 23 storms with a single exception dealing with an important storm from 40 cycle 22 (Cliver et al., 2008). The superstorms of the Halloween 2003 and November 41 2004 periods are the subject of investigation in several papers. 42 2. Solar and Interplanetary phenomena 43 Asai et al. (2008) present a detailed examination of a peculiar active region (AR NOAA 44 10798) that emerged in the middle of a small coronal hole, and formed a sea anemone 45 like configuration. Two successive CMEs from this active region caused a large 46 geomagnetic storm (Dst = -216 nT) on 24 August 2005. The CMEs were very fast (1200 2 47 km/s for the first one and 2400 km/s for the second) as observed by SOHO. Based on the 48 height – time plots of the two CMEs, it was estimated that the CMEs interacted on the 49 way to Earth resulting in an interplanetary CME (ICME) with intense southward 50 magnetic field that was responsible for the large storm. It is suggested that the coronal 51 hole surrounding the active region might have channeled the CMEs with relatively 52 reduced friction between the solar wind and the CMEs. 53 Cliver et al. (2008) report on the solar source of the great geomagnetic storm (Dst = -354 54 nT) on 8-10 November 1991. The solar source is identified as the large-scale eruption of 55 a long (~25º) solar filament followed by a soft X-ray arcade that spanned ~90º of solar 56 longitude, distinguishing the geomagnetic storm as the largest yet associated with a 57 quiescent filament eruption. The storm was found to rank 15th on a list of Dst storms 58 from 1905 to 2004. The November 1991 event also underscores the difficulties in 59 predicting such storms. 60 Gopalswamy et al. (2008) report that many CMEs originating from close to the disk o 61 center (within ±15 in longitude) do not arrive at Earth, while the shocks driven by them 62 do. Such “driverless” shock events occurred only during the declining phase of solar 63 cycle 23. In each case there was at least one large coronal hole near the eruption 64 suggesting that the coronal holes might have deflected the CMEs away from the Sun- 65 Earth line. The presence of abundant low-latitude coronal holes during the declining 66 phase further explains why these events were found in the declining phase. As a control 67 study, they also examined CMEs that originated close to the disk center and arrived at 68 Earth as shocks with drivers. For these, the coronal holes were located such that they 69 either had no influence on the CME trajectories, or they deflected the CMEs towards the 3 70 Sun-Earth line. Disk-center CMEs interacting with coronal holes were not geoeffective, 71 while those minimally influenced by coronal holes were all geoeffective. This work 72 demonstrates that in addition to source and kinematic properties of CMEs, one also has to 73 consider the source environment in order to understand the geoeffectiveness of CMEs. 74 75 Jackson et al. (2008) present a low-resolution three-dimensional (3D) reconstruction of 76 the 27-28 May 2003 halo CME sequence observed by Solar Mass Ejection Imager 77 (SMEI) and the Solar and Heliospheric Observatory (SOHO) mission. These events are 78 known to have caused a major geomagnetic storm on 2003 May 28 (see 79 http://cdaw.gsfc.nasa.gov/CME_list/daily_plots/dsthtx/2003_05/dsthtx.20030528.html). 80 From the reconstruction they were able to infer the shape, extent, and mass of this CME 81 sequence as it reached the vicinity of Earth. The 3D reconstructed density, derived from 82 the remote-sensed Thomson scattered brightness agrees well with the in situ 83 measurements from the Advanced Composition Explorer (ACE) and Wind spacecraft. 84 Bisi et al. (2008) apply the same reconstruction technique to the early November 2004 85 events and compare the reconstructed structures with in situ measurements from the ACE 86 and Wind spacecraft, thus validating the reconstruction results. The early November 87 2004 events have caused two super intense (Dst ~ -373 nT and -289 nT) storms 88 (Gopalswamy et al., 2006). Information derived from the reconstruction technique serve 89 as input to the ENLIL 3D magnetohydrodynamic (MHD) numerical model of the solar 90 wind. 91 4 92 Zhang et al. (2008) report on the multiple dips in the Dst index profile during the storm 93 interval. They studied the properties of the interplanetary drivers of 90 intense 94 geomagnetic storms during 1996 to 2006 to trace the cause of the dips. Since the 95 decrease in Dst index is caused by an interval of southward component of the 96 interplanetary magnetic field, multiple dips mean multiple intervals of southward 97 magnetic field within the overall storm interval. The majority of the 90 storms (66%) 98 showed two or more dips. One frequent cause of two-dip storms is the occurrence of the 99 southward field in the sheath and in the ICME such that the first dip is caused by the 100 sheath field while the second dip by the ICME. Double or multiple dips are also caused 101 by the presence of multiple sub-regions of southward magnetic field within a complex 102 solar wind flow, resulting from two successive, closely spaced ICMEs. 103 3. Magnetospheric Phenomena 104 Liemohn et al. (2008) report on the simulation of the intense magnetic storms from solar 105 cycle 23 using the hot electron and ion drift integrator (HEIDI) model. The simulations 106 were run using a Kp-driven shielded Volland-Stern electric field, static dipole magnetic 107 field, and nightside plasma data from instruments on the Los Alamos geosynchronous 108 satellites. The storms were analyzed by grouping them according to their solar wind 109 driver: ICMEs and corotating interaction regions (CIRs). They find that the HEIDI model 110 was able to best reproduce the Dst time series for storms driven by ICME sheaths. Storms 111 driven by CIRs were the least reproducible class of storms, with simulated minimum 112 Dst* values typically only half to two-thirds of the observed minimum value. In general, 113 there was a strong correlation between the observed and modeled minimums of Dst*, and 114 essentially no correlation between the observed minimum Dst* and the modeled-to- 5 115 observed Dst* ratio. One of the implications of this study is that a Kp-driven HEIDI 116 simulation is consistently on the low side of predicting storm intensity, except for sheath- 117 driven events. 118 Jordanova et al. (2008) study the effect of electromagnetic ion cyclotron (EMIC) wave 119 scattering on radiation belt electrons during the large geomagnetic storm of 21 October 120 2001 (Dst=-187 nT) using their global physics-based model. They calculate the 121 excitation of EMIC waves (field-aligned and oblique) and evaluate particle interactions 122 with these waves according to the quasi-linear theory. They find that pitch angle 123 scattering by EMIC waves causes significant loss of radiation belt electrons at energies 124 >1 MeV due to precipitation into the atmosphere. On the other hand, the relativistic 125 electron flux dropout during the main phase of the storm at large L values (>5) is due 126 mostly to outward radial diffusion. Global simulations indicate significant relativistic 127 electron precipitation within regions of enhanced EMIC instability, whose location varies 128 with time but is predominantly in the afternoon-dusk sector. The minimum resonant 129 energy is found to increase at low L and relativistic electrons (<1 MeV) do not precipitate 130 at L<3 during the October 2001 storm. 131 Ilie et al. (2008) examine how the reference time selection affects the superposed epoch 132 analysis (SEA) for intense storms at solar maximum. Analyzing solar wind data from 133 ACE along with near-Earth data from the LANL MPA instruments, they find that for 134 different choices of the time stamp, different storm characteristics are reproduced in the 135 averaged data. In the ACE data they find that when using the storm sudden 136 commencement (SSC) as a time reference, the SSC-related jump in solar wind 137 parameters is very well reproduced, but near the storm peak, the vertical component of 6 138 the magnetic field (Bz) does not follow the criteria for intense storms (Bz<-10nT for 139 more than 3 hours). On the other hand, the Bz criterion is readily met when the zero 140 epoch time is chosen near the storm peak, but the jump in solar wind pressure is not as 141 sharp. 142 Keesee et al. (2008) present time resolved, remote ion temperature measurements of the 143 magnetosphere from 10 R to -60 R for the 2000 October 4-7 storm. They calculate the E E 144 ion temperatures from Maxwellian fits to IMAGE/MENA data. They find that the 145 calculated ion temperatures in the magnetotail are consistent with in situ measurements 146 from multiple geosynchronous spacecraft and GEOTAIL at x = -9 R . During the E 147 October 2000 storm, two separate instances of an Earthward propagating increase in ion 148 temperature are found. When the solar wind-magnetospheric coupling is strong, the 149 measured ion temperatures are consistent with predictions of a solar wind velocity 150 correlation equation; at other times, the measured ion temperature is 2-3 times larger than 151 the predicted value. 152 Manninen et al. (2008) investigate the steady magnetospheric convection period between 153 the two episodes of the November 2004 superstorm. During the interval in question ( 18- 154 04 UT on 8-9 November), the Dst index was stable but considerably low (-125 nT) and 155 the Bz was steady and slightly negative (~ -5 nT). The strongest magnetic disturbances 156 were observed in the midnight sector of the Earth, rather than in the expected morning 157 side geomagnetic activity and Pc5 geomagnetic pulsations. The results were obtained 158 using the Scandinavian multi-point observations of geomagnetic variations and 159 pulsations, visible auroras, and energetic particle precipitation. 160 4. Ionospheric Phenomena 7 161 Ding et al. (2008) report on the large-scale traveling ionospheric disturbances associated 162 with the major geomagnetic storms during 2002-2005. They use total electron content 163 (TEC) perturbation maps obtained from more than 600 GPS receivers in North America 164 (geographical latitudes of 25°N–55°N) and find 135 cases of such disturbances with 165 amplitudes of up to 3.5 TECU and a maximum front width of ׽ 4000 km. The mean 166 velocity (300 m/s) is slower than that observed at lower latitudes. The occurrence of the 167 disturbances peaks at 1200 LT and at 1900 LT. They also find that the UT dependence of 168 the occurrence of auroral geomagnetic disturbances plays a major role in the forming of 169 UT and LT dependence of the occurrence of the traveling ionospheric disturbances at 170 midlatitudes. Perevalova et al. (2008) report on the large-scale traveling ionospheric 171 disturbance registered in the auroral zone following the sudden storm commencement 172 (SSC) related to the 29 October 2003 event. The disturbance represented a large-scale 173 solitary type wave with an annular front shape whose center was located near the 174 geomagnetic pole. They also detected a “swirling” effect in the disturbance movement in 175 a direction opposite to the Earth's rotation. 176 Balan et al. (2008) report the occurrence of the F3 layer in the equatorial ionosphere at 177 American, Indian, and Australian longitudes during the November 2004 superstorms 178 (November 8 and 10). The observations show the occurrence, reoccurrence, and quick 179 ascent to the topside ionosphere of unusually strong F3 layer accompanied by large 180 reductions in peak electron density and total electron content. Observations and modeling 181 indicate that the unusual F3 layers arise mainly from unusually strong fluctuations in the 182 daytime vertical E × B drift. 8 183 Eriksson et al. (2008) report on an analysis of the great magnetic storm of 15 May 2005 184 associated with a well-known magnetic cloud (Yurchyshyn et al., 2006) using DMSP, 185 TIMED/GUVI, and IMAGE/WIC observations. In particular, they analyze the high- 186 latitude response of sunward E × B flow and Birkeland field-aligned currents. Using 187 DMSP observations, they were able to confirm a dawnward migration of a Northern 188 Hemisphere sunward E × B flow channel between a downward and upward field aligned 189 current pair. Using TIMED/GUVI observations, they also show that the dawnward 190 migration of the upward field aligned current coincides with a drifting transpolar auroral 191 arc. 192 Su. Basu et al. (2008) report on the impact of large ionospheric velocities on GPS-based 193 navigation systems within the midlatitude region in the North American sector during the 194 2004 November superstorm. The 2004 November storm was marked by the absence of 195 appreciable storm-enhanced density gradients compared to the 2003 Halloween storms. 196 This study demonstrates that it is possible to disable GPS-based navigation systems for 197 many hours even in the absence of appreciable TEC gradients, provided an intense flow 198 channel, generally known as the sub-auroral polarization stream (SAPS), is present in the 199 ionosphere during nighttime hours. 200 Mannuci et al. (2008) report the prompt daytime ionospheric responses for four intense 201 geomagnetic storms (during the 2003 Halloween period and 2004 November period). 202 They perform a superposed epoch analysis of the storms and use measurements from the 203 GPS receivers onboard the CHAMP satellite (400 km altitude) and from ground. The 204 TEC data indicate significant low- to middle-latitude daytime TEC increases for three of 205 the storms (-1400 local solar time) except for the 2003 November 20 storm, for which 206 the largest TEC increases appear several hours (-5–7) following the Bz event onset. 207 Estimates of vertical plasma uplift near the equator at Jicamarca longitudes (-281 E) 208 suggest that variability of the timing of the TEC response is associated with variability in 209 the prompt penetration of electric fields to low latitudes. They also found that for the 210 November 2003 magnetic storm the cross-correlation function between the SYM-H index 211 and the interplanetary electric field reached maximum correlation with a lag time of 4 h. 212 Such long delays of both the ionosphere and magnetosphere responses need to be better 213 understood. 214 Pokhotelov et al. (2008) apply a novel technique of extracting the storm time E × B 215 convection boundary from in situ measurements of plasma bulk motion obtained by LEO 216 DMSP satellites to the 20 November 2003 storm. They compare the results with the 217 global distributions of the ionospheric plasma deduced from characteristics of GPS 218 signals. The tomographic inversion of GPS data reveals that the convective flow 219 expanded low enough in latitude to encompass, in part, the formation of the midlatitude 220 TEC anomaly. Some features of the TEC dynamics observed during the 20 November 221 2003 storm, however, suggest that mechanisms other than the expanded ionospheric 222 convection (such as thermospheric neutral winds) are also involved in the formation of 223 the midlatitude anomaly. 224 Sahai et al. (2008a,b) report the effects of the November 2004 storms on the F-region in 225 the Latin American and East Asian sectors. Virtually no spread F (phase fluctuations) on 226 the nights of 09-10 and 10-11 November were observed in the Latin American sector. 227 The East Asian sector showed very pronounced effects during the second superstorm 228 which was preceded by two intense storms. There was no spread-F in the Vietnamese 10