Case study of the 15 July 2000 magnetic storm effects on the ionosphere-driver of the positive ionospheric storm in the winter hemisphere

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. A11, 1391, doi: /2002ja009782, 2003 Case study of the 15 July 2000 magnetic storm effects on the ionosphere-driver of the positive ionospheric storm in the winter hemisphere Hyosub Kil, 1 Larry J. Paxton, 1 Xiaoqing Pi, 2 Marc R. Hairston, 3 and Yongliang Zhang 1 Received 16 November 2002; revised 15 July 2003; accepted 13 August 2003; published 1 November [1] The ionospheric response to the magnetic storm of 15 July 2000 is investigated using the global total electron content (TEC) maps provided by global positioning system and the measurements of ion density, composition, and drift velocity from the Defense Meteorological Satellite Program (DMSP) F13 and F15 spacecraft. The global TEC maps showed clear seasonal effects that can be characterized by a dominance of a negative ionospheric storm (decrease in plasma density) in the summer (northern) hemisphere and the pronounced positive ionospheric storm (increase in plasma density) in the winter (southern) hemisphere. The northern negative storm phase rapidly expanded to the equator at midnight and even penetrated to the opposite hemisphere during the storm main phase. In the southern hemisphere, the negative storm phase began in the morning sector but was confined to narrow latitude and local time sectors owing to strong poleward winds and ion drag on the dayside. The negative storm phases in the opposite hemispheres kept out of phase, lasted for a day, and corotated with Earth. These characteristics show good qualitative similarity with the predictions of global model simulations. The positive storm phase prevailed in the southern low-middle latitudes and was most pronounced during nighttime. In that region, the quiet time DMSP measurements at 1800 LT, 2100 LT, 0600 LT, and 0900 LT showed low ion density, low O + proportion, and large downward ion drift velocity compared with those in the northern hemisphere. During storm time the O + proportion and ion concentration increased to the levels seen in the northern hemisphere while the downward ion drift velocity was much decreased. The excellent temporal and spatial correspondence of the increase in ion concentration with the decrease in downward ion drift velocity indicates that the maintenance of the F layer at high altitudes by the storm-induced equatorward neutral winds was the main driver of the positive ionospheric storm. The quiet time hemispheric asymmetry was most significant at nighttime, and therefore the positive storm effect appeared most pronounced at nighttime in the winter hemisphere. INDEX TERMS: 2435 Ionosphere: Ionospheric disturbances; 2437 Ionosphere: Ionospheric dynamics; 2427 Ionosphere: Ionosphere/atmosphere interactions (0335); 2419 Ionosphere: Ion chemistry and composition (0335); KEYWORDS: ionospheric storm, ionosphere-thermosphere coupling, thermospheric disturbance Citation: Kil, H., L. J. Paxton, X. Pi, M. R. Hairston, and Y. Zhang, Case study of the 15 July 2000 magnetic storm effects on the ionosphere-driver of the positive ionospheric storm in the winter hemisphere, J. Geophys. Res., 108(A11), 1391, doi: /2002ja009782, Introduction [2] During geomagnetic storms the magnetospheric energy input into the polar upper atmosphere can significantly 1 Applied Physics Laboratory, Johns Hopkins University, Laurel, Maryland, USA. 2 Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USA. 3 W. B. Hanson Center of Space Sciences, University of Texas at Dallas, Richardson, Texas, USA. Copyright 2003 by the American Geophysical Union /03/2002JA modify the chemistry and dynamics of the ionospherethermosphere (I-T) system [Prölss, 1995; Buonsanto, 1999]. Intense auroral particle precipitation increases the conductivity of auroral oval, while the increased convection electric field mapped from the magnetosphere drives fast ion motions. The collisions of high-speed ions with neutral gases force the thermosphere into motion and drive energy deposition into the upper atmosphere via Joule heating. Heating of the upper atmosphere leads to an upward expansion of heavier gases, which produces a molecular composition bulge characterized by an enhanced mean molecular mass [Prölss, 1993a; Fuller-Rowell et al., 1994]. Upwelling of the atmosphere promotes O + loss in SIA 1-1

2 SIA 1-2 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM the F region by reaction with molecular gases and subsequent dissociative recombination of the molecular ions: O þ þ N 2! NO þ þ N; O þ þ O 2! O þ 2 þ O; NO þ þ e! N þ O; O þ 2 þ e! O þ O: [3] Reactions (1) and (2) occur much slower than the dissociative recombination of NO + and O 2 + with electrons. Therefore the loss rate of O + is critically dependent on the density of molecular gases N 2 and O 2. Since the production rate of O + is proportional to atomic oxygen density and since the O 2 measurements were not available at most of the time, the atomic oxygen to molecular nitrogen concentration ratio (O/N 2 ratio) has been used as a suitable indicator of I-T storms [e.g., Prölss, 1995; Strickland et al., 1995, 1998, 2001]. The change in electron density in the ionosphere is termed a positive ionospheric storm (or positive storm effect) when the electron density increases as a response to a storm, while it is termed a negative ionospheric storm (or negative storm effect) when the electron density decreases [Mendillo, 1978]. With the onset of a geomagnetic storm the molecular composition bulge is created in the auroral oval, it then expands to lower latitudes by the horizontal neutral winds that arise from the pressure gradient force in the auroral oval and by ion drag in the polar cap. As a result, the negative storm effects are driven in the middle and high latitudes by the thermospheric composition change. [4] The coincidence of ionospheric perturbations with magnetic storms was reported as early as the 1920s [e.g., Anderson, 1928; Hafstad and Tuve, 1929]. The early studies of auroral phenomena suggested a heating of neutral atmosphere during magnetic storms [Appleton and Ingram, 1935; Kirby et al., 1937; Berkner and Seaton, 1940], but its connection to the ionospheric perturbation was not clearly understood owing to a lack of information on the neutral upper atmosphere at that time. Seaton [1956] first suggested that the decrease in the F-peak electron density might be caused by a change in the neutral composition. The possibility of neutral composition change effects was first identified from the observation of depressed O I 1304 Å airglow intensity in the polar region by Meier [1970] and Meier and Prinz [1971]. Modeling efforts by Hays et al. [1973] showed an upward transport of nitrogen-rich air by the Joule heating in the auroral oval, and Mayr and Volland [1973] showed that nitrogen-rich air could be redistributed by horizontal winds causing neutral composition changes beyond the auroral oval. The observational evidence of an atmospheric disturbance (or thermospheric storm) associated with a magnetic storm was first obtained from the satellite drag data that showed a change in the decay rate of satellites associated with geomagnetic activity [Jacchia, 1959]. Following the studies of magnetic storm effects on ð1þ ð2þ ð3þ ð4þ the atmosphere using drag data [Jacchia, 1972; Roemer, 1972], the in situ measurements of atmospheric neutral composition [Prölss, 1980, 1981, 1987, 1993a; Miller et al., 1990; Burns et al., 1995a; Werner et al., 1999] and satellite-borne far ultraviolet (FUV) measurements of O column density or O/N 2 column density ratio [Barth and Schaffner, 1970; Chubb and Hicks, 1970; Meier, 1970; Hicks and Chubb, 1970; Meier and Prinz, 1971; Strickland and Thomas, 1976; Craven et al., 1994; Gladstone, 1994; Meier et al., 1995; Immel et al., 1997, 2000; Nicholas et al., 1997; Strickland et al., 1998, 2001; Drob et al., 1999; Daniell and Strickland, 2001; Zhang et al., 2003] provided observational evidence of thermospheric composition changes during storm time. The advance of theoretical thermospheric physics and development of thermospheric general circulation models in the 1980s further enriched our understanding of the process of magnetospheric energy deposit into the polar upper atmosphere and the I-T response to this energy input [Mayr and Volland, 1973; Dickinson et al., 1981, 1984; Fuller-Rowell and Rees, 1981, 1983; Roble et al., 1982, 1987, 1988; Killeen and Roble, 1984, 1986; Rishbeth et al., 1987; Crowley et al., 1989; Fesen et al., 1989; Burns et al., 1991; Fuller-Rowell et al., 1991, 1994, 1996; Richmond et al., 1992; Codrescu et al., 1997; Förster et al., 1999; Lu et al., 2001]. [5] Now it is a well-established fact from observations and model simulations that the negative ionospheric storms in the high and middle latitude regions are induced by thermospheric neutral composition change. In the lowlatitude region the ionospheric perturbations are produced by different processes and show a different morphology. The heating of high-latitude region launches equatorward wind surges that drag the low-middle latitude plasma to higher altitudes along the magnetic field lines. An uplift of the plasma in that region induces an increase in plasma density owing to the decrease in molecular gases (or decrease in O + loss rate) at higher altitudes. Observational evidence has been provided by measurements of F-peak height (h m F 2 ) and F-peak electron density (N m F 2 ) from worldwide ionosonde stations [Prölss, 1980, 1982, 1993a, 1993b; Richards and Wilkinson, 1998; Bauske and Prölss, 1998; Prölss et al., 1998; Szuszczewicz et al., 1998; Werner et al., 1999; Prölss and Očko, 2000] and the observation of traveling atmospheric/ionospheric disturbances [Hajkowicz and Hunsucker, 1987; Hajkowicz, 1990; Pi et al., 1993, 2000; Ho et al., 1996, 1998; Saito et al., 1998]. Prölss [1993b] used the AE index as a reference for substorm onset and observed an increase in h m F 2 and N m F 2 with a few hours delay (1 2 hours delay in the middle latitude and 3 5 hours delay in the low latitudes) after substorm onset. The increase in N m F 2 following an increase in h m F 2 and its time delay at lower latitudes was provided as evidence of traveling atmospheric disturbances (TADs) that drive positive ionospheric storms in the low and middle latitudes. [6] On the other hand, the model simulations showed that the divergence and upwelling of the polar upper atmosphere during a storm could set up a Hadley-type circulation cell that produces its convergence and downwelling in the low and middle latitudes [Rishbeth et al., 1987; Rodger et al., 1989; Crowley et al., 1989; Burns et al., 1995a, 1995b; Fuller-Rowell et al., 1994, 1996; Immel et al., 2001]. The

3 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM SIA 1-3 downwelling of the atmosphere causes a decrease in molecular gases (or decrease in O + loss rate) in the F region and induces a positive storm effect. The model simulations of National Center for Atmospheric Research-thermosphere/ ionosphere general circulation model (NCAR-TIGCM) by Burns et al. [1995a] predicted a large enhancement (200%) of O/N 2 ratio in the middle latitude, evening-tomidnight sector of the winter hemisphere. They deduced the O/N 2 ratio using the DE 2 data and observed an increase in the O/N 2 ratio in the region where the model simulation predicted an increase in the O/N 2 ratio. The recent study by Immel et al. [2001] indicates both processes, uplift of the ionosphere and thermospheric composition change, could occur simultaneously. Their observations show that the perturbed h m F 2 measured from ionosonde network and the enhanced O I nm emission obtained from DE 1 data propagate in phase from the high latitudes to lower latitudes, which was also identified from their model simulations using the thermosphere/ionosphere/mesosphere electrodynamics general circulation model (TIMEGCM). Thermospheric neutral composition change is an alternative explanation of the positive ionospheric storm that requires further observational evidence for its validation. [7] In this paper, we investigate I-T response to the magnetic storm of 15 July 2000 by using global TEC maps provided by global positioning system (GPS) and the measurements of ion density, composition, and drift velocities from the Defense Meteorological Satellite Program (DMSP) F13 and 15 spacecraft. It was a big storm (minimum Dst index of 370 nt), occurred near solstice, and is expected to provide a good opportunity to compare observations and model simulations. Our investigation is focused on identifying the driver of positive ionospheric storms at low-middle latitudes in the winter hemisphere. While the observations and model simulations show general agreement on the most significant positive storm effects in that region, there is a disagreement on the primary source mechanism. In this paper, we attempt to give a confident answer to this question by providing observational evidence. In section 2 we describe the magnetic storm of 15 July In section 3 the context of ionospheric storms is presented. In section 4 we investigate the time evolution of ionospheric storms by using TEC maps, and in section 5 we investigate the changes in the topside ionospheric conditions during the storm using DMSP data. In section 6 we discuss our observation in comparison with global model simulations. Finally, the summary of our findings and conclusions is given in section Description of Magnetic Storm on 15 July 2000 [8] At midday on 14 July 2000, a major coronal mass ejection was observed by various solar imaging instruments. Since this is a day celebrated by the French as Bastille Day, this solar event was called the Bastille Day Event. Its impact on the Earth s magnetosphere appeared about 32 hours later. Following a steep increase in the interplanetary magnetic field (IMF) southward component, an extreme geomagnetic storm was recorded on 15 July Figure 1 shows the Kp index (top panel), Dst index Figure 1. The Kp index (top panel), Dst index (second panel), IMF B z (third panel), and AE index (bottom panel) during the magnetic storm of 15 July Following the SSC at 1500 UT on the 15th the storm phase is divided into an initial phase (I) during UT on the 15th, a main phase (M) during UT on the 15th, and a recovery phase (R) after 2300 UT on the 15th. (second panel), IMF B z component measured from WIND (third panel), and auroral electrojet (AE) index (bottom panel) during the storm. The Dst index is calculated from 37 stations below 40 magnetic latitude, and the AE index is calculated from 72 stations between magnetic latitude (G. Lu, private communication, 2003). Following the sudden storm commencement (SSC) at 1500 UT on the 15th, the storm main phase started at 1930 UT when B z southward component increased rapidly, and the recovery phase started around 2300 UT. We divide the storm phase into the initial phase (I) ( UT on the 15th), the main phase (M) ( UT on the 15th), and the recovery phase (R) (after 2300 UT on the 15th). Before the SSC there were several substorms on the 14th and 15th. On the 15th the AE index started to increase at 0700 UT, and the substorm activity between UT is attributed to the IMF orientation that has a small but steady southward component. 3. Context of Ionospheric Storms on 15 July 2000 [9] To examine the storm-time ionospheric response, global TEC maps were generated at the Jet Propulsion Laboratory using GPS data collected from about 100 ground stations of the global network of the International GPS Service for Geodynamics. The TEC mapping process

4 SIA 1-4 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM Figure 2. Comparison of global TEC difference map with in situ ion density measurements from DMSP F13 and F15. The top panel shows the TEC map at UT on the 16th, and the ground tracks of F13 and F15 during UT and UT, respectively, on the 16th. The numbers on the map indicate ionosonde stations used for Figure 3. The second panel shows ion density measurements along the F13 track on the 16th (red) and on the 14th (blue). The third panel shows the percentage TEC changes along the F13 pass on the 16th. The fourth and fifth panels have the same format for F15. The observations on the 14th are shown as quiet time reference. was conducted through a Kalman filter, which solves for vertical TEC (on a local-time and geomagnetic latitude frame) and instrumental biases with all ionospheric delay measurements along lines-of-sight of GPS receiver-to-satellite in a time interval [Iijima et al., 1999]. The biasremoved line-of-sight TEC obtained using the mapping technique has an uncertainty of about 1 2 TECU (1 TECU = m 2 ). When mapped to the vertical, additional errors can be introduced due to the adopted simplified mapping function which does not perfectly represent reality in regions of large TEC gradients. An averaged pattern was obtained from TEC maps of the 15 days prior to the storm day, and the relative difference maps with respect to the average were generated for every

5 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM SIA 1-5 hour interval for the event days. The top panel in Figure 2 shows the global TEC difference map at UT and ground tracks of F13 ( UT) and F15 ( UT) on the 16th. The DMSP F13 and F15 satellites have Sun-synchronous orbits at an altitude of 840 km with an inclination of 98. Their measurements covered local solar times of 1800 LT and 0600 LT for F13 and 2100 LT and 0900 LT for F15. In the figure the satellite proceeds from the right to the left. The numbers 1 6 on the map indicate ionosonde stations used in Figure 3. The second panel shows the ion concentration measurements on the 16th (red) and on the 14th (blue) from F13. The third panel shows the percentage TEC difference along the F13 pass. The fourth and fifth panels have the same format for F15. The observations on the 14th in the second and fourth panels are shown as a reference of quiet time values. The comparison of the two measurements show excellent correspondence even in the oceans where there are no GPS stations. The TEC map shows a clear seasonal effect and also the local time dependence of ionospheric storms. The negative storm effect is dominant in the northern (summer) hemisphere and extended to the equator at nighttime. In the southern (winter) hemisphere, the positive storm effect is most pronounced at low-middle latitudes at nighttime. The southern negative storm effect is confined to middle and high latitude regions. [10] Figure 3 shows N m F 2 measurements during 13 17th from the ionosonde stations indicated in Figure 2. The N m F 2 measurements are ordered from station 1 on the top to station 6 on the bottom. The monthly average is shown by solid lines without diamond symbols. At stations 1 and 2 a decrease of 50 90% in N m F 2 is observed from the afternoon of the 15th to the end of the 16th. Those two stations were located in the middle of the negative storm phase in the TEC map shown in Figure 2. At station 3, spike-like increases of N m F 2 are observed from the afternoon of the 15th to the morning of 16th. After 0900 UT of the 16th the N m F 2 decreased by 50 70% until the morning of the 17th. This station was located near the boundary of negative and positive storm phases at UT on the 16th in Figure 2. The large fluctuations of N m F 2 might be produced by surge-like meridional winds or by the perturbation electric field before the negative storm phase dominates this area. The measurements in the southern stations show a different pattern. On the TEC map in Figure 2, station 4 was located at the edge of the positive storm phase. Although the data gap at on the 16th makes it difficult to compare N m F 2 with TEC, the increase of N m F 2 in the morning of 16th indicates that the positive storm phase was dominant in this area. The fourth and fifth panels in Figure 2 (around 0900 LT in the figure) show that station 5 was located in a very complicated region. The TEC and ion density are seen to be normal at this location, but the increases in TEC and ion density are observed at its north and south. The decrease in N m F 2 on the 16th seemed to be caused by the intervention of the negative storm phase into the region where the positive storm phase prevailed. Station 6 was located at the center of the positive storm phase where the TEC increased by more than 100% at UT on the 16th. N m F 2 increased by more than 200% during nighttime and 50% during daytime of the 15 16th. By and large, the global-scale ionospheric changes seen on the TEC map agree very well with DMSP and ionosonde measurements. We note that the 1-hour average TEC maps is not enough to catch the local changes in a short time period, especially in the regions where the opposite storm phases compete. [11] The hemispheric difference of the ionospheric response is caused primarily by the seasonal difference of the magnetospheric energy deposition into the polar upper atmosphere and of the background wind field [e.g., Fuller-Rowell et al., 1996]. We can infer, qualitatively, the atmospheric heating effect from the ion temperature measurements of DMSP. Figure 4 shows the variation of ion temperature as a function of magnetic latitude and UT (UT is counted from 0000 UT of the 15th.) in the northern hemisphere (top panels) and in the southern hemisphere (bottom panels). The local times of the observations are given on the top of the plots. The storm phase is monitored by Dst and AE indices on the right columns. In the northern hemisphere, a significant increase in ion temperature is observed in all local time sectors (1800 LT, 0600 LT, 2100 LT, and 0900 LT) during the storm main and recovery phases but is more significant at 0900 LT. In the southern hemisphere, the temperature increase is not that significant compared to its increase in the northern hemisphere. The increase is again more significant at 0900 LT. The differential hemispheric heating is induced by the difference of conductivities in the two hemispheres [e.g., Berthelier, 1976; Banks et al., 1981; Foster et al., 1983; Weimer et al., 1990]. In the summer hemisphere, most of the polar region is on the sunlit side (high conductivity), and therefore Joule heating can be significant. The more significant increase in ion temperature at 0900 LT in both hemispheres is considered to be associated with the westward electrojet in the morning sector [e.g., Lu et al., 1995; Thayer et al., 1995]. An unusual increase in the ion temperature is observed at 2100 LT between 60 and 30 magnetic latitudes during the recovery phase (after 3400 UT in the figure on the bottom). It is an interesting feature, but we do not have good explanation of this phenomenon. The magnetospheric energy dissipation in the I-T system is a complicated process, and its discussion is outside of the scope of this paper. From the ion temperature measurements we infer the presence of a hemispheric and local time difference in the energy input that leads to a differential ionospheric response to the geomagnetic storm. 4. Temporal and Spatial Evolution of Ionospheric Storms Seen From TEC Maps [12] The I-T storm starts from the high-latitude region, but its impact reaches all the way to the equator. Owing to global-scale changes in the I-T system its study can be done more effectively using global observations. In spite of the advances in theoretical thermospheric physics in the last 30 years, its validation has relied on limited data sources. FUV observations from space have the advantage of providing global-scale images of changes in thermospheric composition, but this information is available only on the sunlit side from a few FUV instruments operating during limited time periods [e.g, Strickland et al., 1998, 2001].

6 SIA 1-6 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM Figure 3. N m F 2 measurements during July at the ionosonde stations shown in Figure 2. The plots are ordered from the top to the bottom following the station ID 1 to 6. The solid lines without diamond symbols are the monthly average of N m F 2.

7 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM SIA 1-7 Owing to the paucity of neutral measurements, thermospheric changes have been inferred by using the perturbations of ionospheric parameters. The frequently used ionospheric variables are N m F 2 and TEC. Recently, global TEC maps have been used to study I-T storms [Ho et al., 1996, 1998; Lu et al., 1998; Buonsanto et al., 1999; Aponte et al., 2000]. In this section, we investigate the global-scale ionospheric changes during the storm period using high temporal and spatial resolution TEC maps. The TEC maps that we present here, compared with the previous ones, include newly available data from sites in Africa, South America, and Asia. [13] Figure 5 shows the time series of global TEC maps from 1600 UT on the 15th to 2200 UT on the 16th in the magnetic local time (MLT) and magnetic latitude coordinates. The white circles in each plot are drawn every 30 from the magnetic equator to the magnetic pole. There is a data gap at UT on the 15th owing to a problem associated with data retrieval. We divide time into three sectors based on the storm evolution initial phase (before 2000 UT of 15th), main phase ( UT of 15th), and recovery phase (after 2300 UT of 15th) Initial Phase (TEC Maps During UT on the 15th) [14] The negative and positive ionospheric storms existed before the SSC at 1500 UT on the 15th. Before 1500 UT the negative (positive) storm phase was developed along the auroral oval in the northern (southern) hemisphere. The AE index increased on the 14th and in the morning of the 15th before SSC (see Figure 1), and the ionospheric perturbations before the SSC are attributed to those substorm activities. During the initial phase, the negative storm phase in the auroral oval intensified and broadened in the northern hemisphere but had not yet reached low latitudes. In the southern hemisphere, the positive storm phase extended from the auroral oval to the polar cap and strengthened in the low and middle latitudes. The features of negative storm phase appeared only at localized areas in the nighttime middle latitudes Main Phase (TEC Maps During UT on the 15th) [15] During the main phase of the storm the northern negative storm phase expanded to the equator and even penetrated to the opposite hemisphere at midnight. The coincidence of main phase onset with the movement of northern magnetic pole into nightside seemed to prompt its rapid equatorward expansion during UT [e.g., Fuller-Rowell et al., 1994]. The northern negative storm phase expanded to lower latitudes in the morning sector than in the evening sector. Significant heating in the morning sector as shown in Figure 4 and the development of strong EIA in the evening sector might cause the preferential expansion in the morning sector. In the southern hemisphere, the signature of negative storm phase that is developed in its own hemisphere appeared in the morning sector at UT. The onset time of the negative storm phase in the morning sector agrees with the time of ion temperature increase in the 0900 LT sector in Figure 4. After the initiation of the negative storm phase the positive storm effect disappeared in the MLT sector in the high and middle latitude regions but was intensified in the evening and night sectors Recovery Phase (TEC Maps During UT on the 16th) [16] The northern negative storm phase further expanded to lower latitudes during the recovery phase. Its equatorward expansion at midnight and co-rotation with the Earth broadened the MLT extent of the negative storm effect. The TEC map at UT on the 16th shows an expansion of the negative storm phase to the equator along the MLT sector as well as during nighttime in the northern hemisphere. This rapid expansion during daytime is quite an unusual feature owing to the strong poleward wind and ion drag [Fuller-Rowell et al., 1994]. The strong positive ionospheric storm in the evening sector and the strong poleward wind in the noon sector might lead the negative storm phase to expand preferentially between those barriers. In the southern hemisphere, the negative storm phase expanded from the middle latitudes to high latitudes (compare the TEC maps at UT on the 15th and UT on the 16th). It might be due to the strong positive storm effect (or ion drag) at low-middle latitudes and the poleward wind in the noon sector. As the positive storm effect diminished (after 1700 UT on the 16th), the southern negative storm phase could expand to the equator and penetrate to the opposite hemisphere. The corotation of the negative storm phases in both hemispheres is evident during the storm recovery phase. At UT the negative storm phase in the northern hemisphere extended to the equator at night, while the negative storm phase in the southern hemisphere was elongated along the noon sector. Their phases rotated 180 degrees at UT. The negative ionospheric storms in the opposite hemispheres maintained opposite phases during the whole recovery phase. [17] As the negative storm phase expanded to lower latitudes, the positive storm effect diminished in the northern hemisphere. In the southern hemisphere, the positive storm phase competes with an expansion of negative storm phase at low-middle latitudes [e.g., Strickland et al., 2001]. The strong nighttime positive storm phase in the south persisted more than a half day (compare the TEC map at UT on the 16th with the TEC map at UT on the 16th). On the other hand, the daytime positive storm phase appeared and disappeared depending on the location of negative storm phase. As a test case, we focused our attention on the MLT sector around 30 magnetic latitude. At UT on the 15th a positive storm phase is present, and then it gradually disappears at UT of the 16th with an expansion of the negative storm phase. As the negative storm phase rotates in the counterclockwise direction, the positive storm phase gradually reappears in that MLT sector at UT on the 16th. The long time persistence of the nighttime positive storm phase and the repetition of the appearance and disappearance of the daytime positive storm phase imply that those features do not corotate with Earth. The different behavior of the negative and positive ionospheric storm phases may indicate that they are produced by a different mechanism. We note that this morphology does not apply for all

8 SIA 1-8 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM

9 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM SIA 1-9 Figure 5. Temporal evolution of the global TEC difference maps during July. The TEC maps are presented in polar magnetic latitude and magnetic local time coordinates. In the text the storm phase is divided into the initial phase ( UT of 15th), main phase ( UT of 15th), and recovery phase (after 2300 UT of 15th). positive storm phases. The positive storm phase in the Middle and North America during the storm initial and main phases and the peak positive storm phase in the southern hemisphere during the storm recovery phase appeared to corotate with Earth. We could not clarify whether the noncorotating and corotating positive storm phases were produced by the same mechanism or by a different mechanism. We note that the local plasma enhancement could persist for a long time giving the corotation effect if its source mechanism acted continuously. 5. Changes in Low-Middle Latitude Topside Ionospheric Conditions During Storm Time [18] The measurements of N m F 2, FUV emissions, and model simulations reflect the phenomena below 500 km Figure 4. (opposite) In situ ion temperature measurements from DMSP F13 and F15. The top two panels are observations in the northern hemisphere, and the bottom two panels are observations in the southern hemisphere. The local times of the satellite orbits are given on the top of the plots. The Dst and AE indices are shown on the right columns as an indication of the storm phase.

10 SIA 1-10 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM Figure 5. (continued) altitude. Extensive studies of the response of topside ionosphere to geomagnetic storm has yet to be performed. The DMSP observations have been used for a climatological study of topside ionosphere in the equatorial and middle latitude regions. Before we present our observations from DMSP, we briefly describe the topside ionospheric physics that is relevant to our data interpretation. [19] The plasma distribution in the equatorial and midlatitude ionosphere is subject to transport along the magnetic field lines by a number of processes involving thermospheric neutral winds, diffusion along the magnetic field line, and E B drifts [e.g., Heelis et al., 1978; Vickrey et al., 1979; Heelis and Hanson, 1980; Kutiev et al., 1980; Greenspan et al., 1994; Sulzer and Gonzalez, 1996; West and Heelis, 1996; MacPherson et al., 1998; Venkatraman and Heelis, 1999a, 1999b, 2000]. During daytime, the photoionization of atomic oxygen in the F region creates a downward pressure gradient (or upward pressure gradient force) in the topside ionosphere. As a result, the newly created O + diffuses upward along the magnetic field lines. The plasma also undergoes upward E B drift motion during the daytime. The upward diffusion and E B drift of plasma during daytime make the oxygen ion the dominant species at the altitude of DMSP satellites (840 km). At night, the rapid recombination of the ion species in the lower ionosphere induces downward diffusion of the topside plasma along the field line. The plasma also undergoes downward E B drift at nighttime. As a result the H +

11 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM SIA 1-11 Figure 5. (continued) population increases at nighttime in the topside ionosphere. As well as the diffusion and E B drift the neutral wind motions that vary with season and local time can significantly modify the ionospheric dynamics and composition. From the DMSP F10 data West and Heelis [1996] observed substantial longitude variations of O + /H + composition in dip latitudes at different local time and season. The longitudinal variations of ion composition at the height of the satellite were attributed to the modulation of F-layer height by the neutral winds. Venkatraman and Heelis [2000] used DMSP F10 measurements and found that the fieldaligned plasma flows are maximized in regions where the effects of the F-region neutral meridional and zonal winds maximize. Both studies emphasize neutral winds as a main driver of interhemispheric plasma transport in the topside ionosphere. [20] The TEC map in Figure 2 has shown the most pronounced positive storm effect at night in the winter hemisphere. Among the four local time orbits of DMSP F13 and F15, the orbits of F15 at 2100 LT pass through near the center of the positive ionospheric storm. Figure 6 shows the DMSP F15 ground tracks at 2100 LT (top panels) and their measurements of ion concentration (second panels), O + proportion (third panels), and vertical ion drift velocity (fourth panels). The left columns are observations at UT, and the right columns are observations at UT during three consecutive days, the 14th (blue), 15th (green), and 16th (red). The cross mark indicates the

12 SIA 1-12 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM Figure 5. (continued) magnetic equator. Owing to the similarity of satellite orbits at similar UTs during those days, we could study the temporal variations of those parameters while minimizing their spatial variations. In the figures, the observations of the 14 15th are used as references for the state of the ionosphere before the development of the ionospheric storm. We note that the vertical drift velocity at middle latitudes has a contribution of field aligned motion of a plasma owing to an increase in dip angle. The distinguishing features of observations on the 14 15th are the hemispheric asymmetry of those observations at low-middle latitudes. In the southern hemisphere, ion density and O + proportion are much reduced, and large downward ion drift velocity is observed before the storm onset. As was Figure 6. (opposite) DMSP F15 ground tracks (top panels) and their measurements of ion concentrations (second panels), O + proportion (third panels) and vertical ion drift velocity (fourth panels). The observations on the left and right columns were made at UT and at UT on the 16th, respectively. The satellite tracks and their measurements of different days are distinguished by colors (14th is blue, 15th is green, and 16th is red). The cross marks indicate the location of the magnetic equator. The observations on the 14 15th are shown as references of quiet time behavior.

13 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM SIA 1-13

14 SIA 1-14 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM described above, those hemispheric asymmetries during nighttime resulted from the downward diffusion and E B drift of plasma and lowering ionosphere in altitude by the summer-to-winter neutral winds. The observations of the 16th are totally different from those of previous days. At UT (left column) the vertical drift velocity at winter low-middle latitudes is comparable to that of summer hemisphere, and simultaneously the O + proportion increased up to 100% and the ion concentration increased more than an order of magnitude. We note that the increase of ion concentration is also observed in the northern hemisphere. Since the background ion concentration in the north was much higher than that in the south, the positive storm effect appears much weaker in the north. Similar changes are observed at UT (right column). Compared with the observations at UT in the American sector, the EIA in the northern hemisphere disappeared at UT. This was due to the expansion of northern negative storm phase into the Pacific Ocean (see the TEC map in Figure 2 or the TEC map at UT on the 16th in Figure 5). Those two sample observations indicate that the positive ionospheric storms are produced in the low-middle latitude in both hemispheres, but the positive storm effect appears more pronounced in the southern hemisphere owing to the hemispheric asymmetry of the plasma density before the storm and owing to the dominance of negative storm effect in the northern hemisphere. [21] We further investigated the temporal evolution of ionospheric conditions at the local time sectors of DMSP orbits. Figure 7 shows the temporal (it is also longitudinal) and latitudinal variations of ion concentration (first column), O + proportion (second column), and vertical ion drift velocity (third column) for the four local time sectors, 2100 LT (a), 1800 LT (b), 0600 LT (c), and 0900 LT (d). The Dst and AE indices (fourth and fifth columns) are shown as a storm phase reference. In the following, the UT is counted from 0000 UT of 14th. The observations at 2100 LT (Figure 7a) show excellent correspondence of the decrease in downward drift velocity with an increase in ion density and O + proportion at low-middle latitudes in the southern hemisphere. The significant changes of ionospheric parameters occurred at around 4600 UT and continued until 6000 UT. The recovery of hemispheric symmetry of those parameters during storm time indicates that the process that produced the hemispheric asymmetry is removed or lessened in importance. [22] Similar changes in the ionospheric parameters are observed in other local time sectors. The observations at 1800 LT (Figure 7b) show variations of ion composition and vertical drift velocity with longitude. They are due to the variation of zonal wind contribution to the interhemispheric plasma transport depending on the magnetic declination [West and Heelis, 1996; Venkatraman and Heelis, 2000]. Oxygen ions are the dominant constituent at the satellite altitude except in the regions where the large downward drift velocity was observed. Compared to the quiet time values, significant changes in ionospheric parameters have arisen around 4400 UT and lasted until 6000 UT. The correspondence of the decrease in downward drift velocity with an increase in O + proportion and ion concentration is again evident at 1800 LT. [23] The covariance of ionospheric parameters is also observed in the morning sector. However, their changes during storm time are not that obvious owing to the presence of longitudinal variation of the parameters during quiet time. At 0600 LT (Figure 7c) an increase in O + proportion in the south is obvious during storm time ( UT). Careful comparison of the ion concentration and vertical drift velocity during that time period with those at UT and at UT also reveals their changes in a correlated way during storm time. The minimum ion concentration and O + proportion occurred at 0600 LT owing to the downward drift and diffusion of ionospheric plasma during nighttime. [24] At 0900 LT (Figure 7d) an increase in ion concentration was observed at UT. The increase in O + proportion occurred around 4400 UT. Compared to the downward drift velocity at other local times, its magnitude is a minimum at 0900 LT owing to the upward plasma diffusion and E B drift during daytime. Although the change is small, we still see an increase in vertical drift velocity at UT. It is interesting to compare the O + proportion and vertical drift velocity at 0900 LT with those at 1800 LT. The locations of O + proportion increase (and also downward velocity increase) are well ordered by the magnetic declination in the two local time sectors. This is due to the opposite zonal wind velocity (eastward at 1800 LT and westward at 0900 LT) that contribute to the interhemispheric plasma transport in opposite ways in regions of opposite magnetic declination [West and Heelis, 1996, Venkatraman and Heelis, 1998a, 1998b, 2000]. [25] Over the four local time sectors an increase in the ion concentration in concert with a decrease in the downward ion drift velocity was observed in the southern low-middle latitudes during the storm period. Compared with the observations at quiet time the maximum change in the vertical drift velocity and correspondingly the maximum increase in the ion concentration was observed at 2100 LT. This morphology agrees with the observation of the most pronounced TEC increase at nighttime in the southern hemisphere. As the thermospheric neutral winds are the main driver of hemispheric asymmetry of topside ionosphere during quiet time, it is believed that the storminduced neutral winds are the primary source of the storm time changes. The intensification of negative storm effect in the southern hemisphere during the late storm recovery phase (after 6000 UT) is contrasted to its dilution in the northern hemisphere during the same time period. The TEC maps in Figure 5 have also shown an intensification of the TEC decrease and its expansion at nighttime after 1600 UT on the 16th in the southern hemisphere. The ion temperature measurements from the DMSP 2100 LT orbits showed its significant increase in the southern middle latitudes during that time period (see the bottom panels in Figure 4). While we do not know the cause of this temperature increase, the intensification of negative storm effect in the southern hemisphere during the late storm recovery phase is seen to be associated with this heating. 6. Discussion [26] The case study of the ionospheric response to the magnetic storm of 15 July 2000 has revealed many interest-

15 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM SIA 1-15 Figure 7. Temporal and spatial variations of ion concentration (left), O + proportion (middle), and vertical ion drift velocity (right) at 2100 LT (a), 1800 LT (b), 0600 LT (c), and 0900 LT (d). The Dst and AE indices are shown on the right side. The black dotted lines indicate the magnetic equator. ing features that reflect the characteristics of thermospheric storms. In this section, we discuss the seasonal and local time dependence of I-T storms, and the drivers of positive ionospheric storm in comparison with global model simulations Seasonal Effect [27] The seasonal dependence of the ionospheric storm phase is a well-known fact from observations [Prölss, 1993a; Ho et al., 1998; Szuszczewicz et al., 1998] and numerical simulations [Fuller-Rowell et al., 1994, 1996]. In agreement with previous studies, our observations show a clear seasonal effect that can be characterized by a negative storm effect in the summer hemisphere and positive storm effect in the winter hemisphere. The two main factors that produce the seasonal effect are the differential magnetospheric energy deposition and the direction of background wind field. Most of the magnetospheric energy is deposited in the summer hemisphere

16 SIA 1-16 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM Figure 7. owing to the high conductivity on the sunlit side. The molecular composition bulge induced by heating can rapidly expand to lower latitudes at nighttime in the summer hemisphere in concert with the summer-to-winter background wind circulation and the storm-induced equatorward wind surges. The nighttime equatorward expansion of the molecular composition bulge and its corotation with Earth further broadens the negative storm effect in the summer hemisphere. On the other hand, the Joule heating effect is less significant in the winter hemisphere owing to low conductivity. As well, the background summer-towinter wind fields (poleward winds in the south) act (continued) oppositely to the equatorward expansion of the molecular composition bulge at nighttime. As a result, the expansion of negative storm phase is limited to high-middle latitudes while the positive storm effect is pronounced in the lowmiddle latitudes. These characteristics of I-T storms agree with model simulations [Fuller-Rowell et al., 1994, 1996]. One new finding that we would like to add to the seasonal effect is an onset condition of the negative storm phase in the winter hemisphere. The onset of negative storm phase in the southern hemisphere was delayed a few hours after the storm main phase onset. A significant increase in ion temperature was observed at 0900 LT (Figure 4) when the

17 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM SIA 1-17 Figure 7. (continued) southern magnetic pole moved to the dayside. At that time the negative storm effect was first observed in the morning-noon sector (TEC map at UT on the 15th in Figure 5). It is seen to be due to the change of conductivity in the auroral oval depending on the location of the magnetic pole. When the southern magnetic pole is in the nightside, the greater part of the southern auroral oval is in darkness, and therefore the Joule heating effect is minimal. As the magnetic pole moves into the dayside, some portion of the auroral oval would be placed on the sunlit side where the Joule heating effect becomes significant. If the molecular composition bulge is always created in the dayside in the winter hemisphere, then the negative storm effect would be always minimal in the winter hemisphere Local Time Effect [28] Fuller-Rowell et al. [1994, 1996] suggested that there is a rapid equatorward expansion of summer negative storm phase at nighttime owing to the background summer-to-winter wind circulation and reduced ion drag. Fuller-Rowell et al. [1994] showed that this effect could

18 SIA 1-18 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM Figure 7. (continued) be maximized (minimized) when a storm begins while the magnetic pole is in the nightside (dayside) (longitude/ut effect). The TEC maps at UT on the 15th in Figure 5 show a rapid equatorward expansion of the northern negative storm phase at midnight. In this time interval, the Dst index reached its minimum ( 370 nt) and the northern magnetic pole was moving into nightside. This observation may support the longitude/ut effect. In the southern hemisphere, the negative storm phase began on the dayside. It preferentially expanded to higher latitude rather than to lower latitude owing to strong poleward wind and ion drag on the dayside (compare the TEC map at UT on the 15th with TEC map at UT on the 16th in Figure 5). The onset of negative storm phase in the dayside minimized its effect in the winter hemisphere. This is an another example that supports the longitude/ut effect Driver of Positive Ionospheric Storm at Low and Middle Latitudes in the Winter Hemisphere [29] In section 4 we described the positive storm phases that did not corotate with the Earth. Whether the storm

19 KIL ET AL.: DRIVER OF THE POSITIVE IONOSPHERIC STORM SIA 1-19 phase corotates or not has implications as to whether its source mechanism acts locally (fixed in longitude) or globally (not fixed in longitude). In that sense, the negative ionospheric storm is a local phenomenon produced by local neutral composition change. It takes a long time for the changes of neutral composition to recover to normal and, as a consequence, the negative storm phases appear to corotate with the Earth. On the other hand, the positive storm phase that did not corotate with Earth is a global phenomenon dependent on local time. The repetition of its appearance and disappearance in the noon sector in the southern hemisphere indicates that its source mechanism acted quickly within a few hours. The positive storm phase at night persisted for a half day until the storm-induced wind surges are diminished. Those characteristics of positive ionospheric storms suggest that the positive ionospheric storm is produced by a global mechanism that is different from the source mechanism of the negative ionospheric storm. [30] The DMSP observations (Figure 7) have shown a large hemispheric asymmetry in the plasma density in the low-middle latitudes during quiet time. The severity of asymmetry varied with local time, but the asymmetry was observed at all local time sectors of DMSP orbits. Our observations and previous studies by West and Heelis [1996] and Venkatraman and Heelis [2000] suggest that the summer-to-winter thermospheric neutral winds are the main driver of the asymmetry. During storm time the hemispheric asymmetry was totally removed or lessened in all local time sectors. As a consequence, the positive storm effect appeared most pronounced in the low-middle latitudes at nighttime where the ionosphere experienced the most severe hemispheric asymmetry. Its effect appeared weaker in the terminator and mid morning sectors where the asymmetry was relatively small. The excellent correspondence of an increase in ion concentration with the decrease in the downward drift velocity during storm time strongly supports the idea that the storm-induced wind surges are the primary driver of positive ionospheric storm in the winter hemisphere. [31] The DMSP observations show that the storminduced neutral winds maintain their strength for a long time during the recovery phase. The AE index decreased rapidly at the beginning of the 16th, but a significant change in the vertical drift velocity was observed until around 1200 UT on the 16th. The long duration of the perturbation winds during recovery phase seemed to contradict the model simulation of Fuller-Rowell et al. [1994]. They found that the neutral winds return to close to their quiescent state fairly rapidly during recovery phase, and therefore, are not responsible for the positive storm phases that persist for a long time during recovery phase. Although their model simulation could detect rapid equatorward wind surges after storm onset and suggested the possibility of plasma density increase by the meridional wind surges, they emphasized neutral composition change as the primary source mechanism of the positive ionospheric storm. While we could not rule out the contribution of neutral composition change to the positive ionospheric storms, our ionospheric observations strongly support that the storminduced neutral winds can drive a positive ionospheric storm regardless of the presence of neutral composition change. 7. Summary and Conclusions [32] We summarize our findings as follows. [33] 1. The ionospheric response to the magnetic storm on 15 July 2000 showed a clear seasonal effect. The negative storm effect was dominant in the northern (summer) hemisphere while the positive storm effect was pronounced in the southern (winter) hemisphere. [34] 2. During the storm initial phase the ionospheric storms showed opposite phases in the high-latitude region of the opposite hemispheres. While the ionization by particle precipitation would be more or less equivalent in the opposite hemispheres, the difference of Joule heating effect caused the negative storm effect in the northern hemisphere and a positive storm effect in the southern hemisphere. [35] 3. The northern negative storm phase expanded rapidly to the equator and penetrated to the opposite hemisphere at midnight during storm main phase. The coincidence of storm main phase onset with the movement of northern magnetic pole into nightside seemed to maximize its expansion at nighttime. [36] 4. The southern negative ionospheric storm initiated in the morning sector when the southern magnetic pole moved into the dayside. Its expansion was hindered by strong poleward winds and ion drag on the dayside, which minimized the negative storm effect in the winter hemisphere. [37] 5. The equatorward expansion of the negative storm phase was effectively interrupted by the ion drag force in the areas where the positive ionospheric storms were pronounced. [38] 6. The negative ionospheric storms persisted more than a day and corotated with the Earth. The ionospheric storm phases in the opposite hemisphere remained out of phase during the whole recovery phase. [39] 7. There appears a corotating and noncorotating component of the positive storm phases. In the southern hemisphere, the strength of the daytime positive storm phase depends on the location of the corotating negative storm phase. The nighttime positive storm phase persisted more than a half day. Those positive storm features that do not corotate are driven by the storm-induced neutral winds. [40] 8. There was a similar amount of corotating positive storm phase. This corotating component of the positive storm phase might be produced by a different process such as neutral composition change. However, the effect of the composition change on the ionospheric plasma density may be too small to account for the observed effect [e.g., Prölss et al., 1998]. We note that the plasma density enhancement in the local region can persist for a long time and appears to corotate with Earth if the storm-induced neutral winds continue to blow plasma up the field lines long enough. [41] 9. Hemispheric asymmetries in ion concentration, composition, and vertical drift velocity existed at low and middle latitudes during quiet time. This asymmetry was, in order of magnitude of the effect, present at night, the