The October 2003 geomagnetic storm (often called the Halloween storm) was one of the largest storms (as measured by Dst) yet recorded. The storm-induced synoptic-scale changes in the ionosphere's plasma content and density can be viewed through space weather maps created by objective analysis algorithms. For this study, these maps, which specify the electron density in altitude, latitude, and longitude, are created by the ionospheric data assimilation three dimensional (IDA3D), a three-dimensional variation algorithm of the ionospheric electron density. These maps, representing the average conditions in the ionosphere over a 15 min sampling time, show how dramatically the ionosphere changed during the Halloween storm. Following the southward turning of the interplanetary magnetic field, the dayside electron content is significantly reduced in the equatorial ionosphere between ±18° magnetic latitude and is enhanced poleward of this latitude. This is the expected behavior when the equatorial fountain is enhanced by a strong penetration electric field. In addition, the electron content is significantly increased in the dayside midlatitude ionosphere, which corresponds to a storm-enhanced density (SED) plume. Above 40° magnetic latitude, the dayside plasma content is significantly reduced in the regions adjacent to the SED structure, which enhances the electron content gradient. Electron density maps in the altitude–magnetic latitude plane show an increase in the topside electron densities within an SED plume.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The penetration electric field increases the E × B drift, which causes an upward ion drift at the magnetic equator. This increased drift pushes newly created plasma to higher altitudes where the ions and electrons diffuse downward and poleward along magnetic field lines [e.g., Anderson, 1973]. The equatorial anomaly has been extensively studied [e.g., Appleton, 1946; Hanson and Moffett, 1966; Walker, 1981; Andreeva et al., 2000; Whalen, 2003; Tsurutani et al., 2004] by a variety of instruments and measurement techniques. These studies have shown that the peaks of the anomaly move poleward as the penetration field increases [e.g., Whalen, 2003]. Recently, Tsurutani et al.  have shown that large total electron content (TEC) increases occurred at ±15°–20° magnetic latitude during a large magnetic storm in November 2001. The peaks moved poleward by ∼3° and showed an asymmetry with higher values in the northern, winter hemisphere. These changes in the equatorial anomaly occurred shortly after the arrival of the solar wind shock at the magnetopause. This behavior is suggestive of a penetration electric field [e.g., Kelley et al., 2003] and argues for a “superfountain” effect. In addition, a “shoulder” (a sharp drop) in the TEC was observed only in the Southern Hemisphere. It was argued that this shoulder is the signature of the plasmapause boundary layer [Carpenter and Lemaire, 2004]. The shoulder, located between −55° and −40° magnetic, moved equatorward during the storm's main phase and poleward during the recovery phase.
 This shoulder may also be an indication of a storm-enhanced density (SED) plume. SEDs [Foster, 1993] are regions of enhanced midlatitude plasma density typically occurring in the postnoon sector. In addition, most of the enhanced density occurs in the topside ionosphere. Foster et al. [2002, 2005b] demonstrated the relationship between SED plumes and plasmaspheric drainage plumes [Sandel et al., 2001]. In addition, Foster et al. [2002, 2004] have suggested that subauroral polarization streams (SAPS) [Foster and Burke, 2002], intense electric fields just equatorward of the auroral oval, are important in SED development. While the earliest SED studies relied mainly upon observations from the Millstone Hill incoherent scatter radar, more recent studies have examined vertical TEC maps over the continental United States. Recently, Foster et al. [2005a] have demonstrated that when SED plumes flow into the polar cap, they become tongues of ionization (TOIs) [Tsunoda, 1988].
 Both the equatorial anomaly and SED plumes are synoptic-scale (multiple degrees) structures. To study these features, it is often necessary to compile numerous observations into larger-scale space weather maps. As previously noted, several studies [e.g., Foster et al., 2002, 2005a; Tsurutani et al., 2004] have used TEC maps constructed from hundreds of TEC measurements. Typically, such slant TEC measurements are converted to vertical TEC by a thin shell approximation [Mannucci et al., 1998; Coster et al., 2003]. Unfortunately, it is difficult to include other forms of ionospheric measurements in these TEC maps. Numerous other ionospheric data sets are available, which provide auxiliary information about the ionosphere. GPS receivers located on low Earth orbiting satellites observe the over-satellite electron content (OSEC) [e.g., Heise et al., 2002; Stankov et al., 2003] and the electron content along an occultation path [e.g., Hajj and Romans, 1998; Angling and Cannon, 2004]. These measurements are an exciting new data set but are difficult to include in TEC maps. Usually, these satellites orbit above the F region so that OSEC measurements miss a significant proportion of the electron content. Similarly, GPS occultations pass through the F region at two different locations. Additional electron content measurements come from ionospheric tomography receivers [e.g., Bust et al., 2001] and sea surface altimeters [e.g., Fu et al., 1994; Jee et al., 2004]. In addition to electron content measurements, there are a limited number of point observations of the electron density. Ionospheric sounders, important instruments for observing the F region ionosphere, are only capable of measuring the height profile directly over the station. Satellite-borne in situ instruments provide latitude and longitude coverage but poor altitude and local time coverage. Incoherent scatter radars (ISRs), perhaps the single most powerful ionospheric instrument, are capable of measuring electron densities well above the F region peak. Several can scan in latitude and longitude. However, ISRs are geographically limited (mainly in the American sector), and the available observing times are limited for other reasons. To view the synoptic-scale electron density, all of these various observations need to be collected into space weather maps.
 These more inclusive space weather maps are generated by objective analysis (OA) algorithms. Commonly used in the meteorology community, objective analysis techniques provide a statistical minimization of the available observations and a background density field. Recently, several different OA algorithms have been developed [e.g., Garcia-Fernandez et al., 2003; Mitchell and Spencer, 2003; Bust et al., 2004], particularly as a component to data assimilation models [Scherliess et al., 2004; Wang et al., 2004]. This paper examines the synoptic-scale maps of the electron content and densities produced by the ionospheric data assimilation three dimensional (IDA3D) [Bust et al., 2004] to better understand the ionospheric response to solar wind changes during the 29–30 October 2003 magnetic storm. In particular, these maps introduce new views of the storm time changes in the equatorial anomaly and the development of SED plumes. The three-dimensional maps generated by IDA3D provide a new view of the plasma distribution within the SED plume. After providing an overview of what is commonly called the Halloween storm (section 2), IDA3D and the data sets available to IDA3D during the Halloween storm are described (section 3). Next, the space weather maps of the total electron content are presented as a time series in relationship to IMF Bz changes (section 4). Section 5 then presents altitude-latitude slices through SED and superfountain structures. The implications of these results are discussed in section 6, which is followed by a short summary.
2. Overview of the Halloween Storm
 October 2003 was a period of several intense geomagnetic storms, which had significant impacts upon satellite operations [Barbieri and Mahmot, 2004; Gopalswamy et al., 2005] and power grid systems [Kappenman, 2005], in addition to dramatically changing the state of the near-Earth space environment. Perhaps the largest of these storms began in earnest on 29 October (day 302) when the Kp index reached 9 at 0600 UT. Figure 1 shows Kp and Dst indices for the 29–30 October 2003 storm. It remained strong (7 ≤ Kp < 8), severe (Kp ≥ 8), or extreme (Kp = 9) for 24 hours. The intensity of the storm became strong at 1800 UT on 30 October (day 303) and remained a class G3 or higher storm for the next 21 hours, including 6 hours as a G5 storm. The geomagnetic storm classification and its relationship to the Kp index are described by the NOAA Space Environment Center (http://www.sec.noaa.gov/NOAAscales).
Skoug et al.  have discussed the extreme solar wind conditions near the end of October 2003. During 29 and 30 October, the solar wind reached extremely high speeds, seldom below 1000 km/s for these 2 days. Two shocks were observed at the ACE spacecraft on these 2 days, one at 0558 UT on day 302 and one on day 303 at 1619 UT. In between these shocks, a large coronal mass ejection was observed from 0800 UT on day 302 to 1600 UT on day 303. Figure 2 shows observed and derived solar wind quantities measured at the ACE spacecraft on these days. The top plot shows the high-resolution (16 s) measurement of the IMF Bz, while the second plot presents the hourly value of the solar wind velocity. The bottom two plots are derived quantities, the interplanetary electric field (IEF) = vsw × Bz and the time shift to 6.6 RE in front of the Earth. Geosynchronous orbit was chosen because the magnetopause was observed to be inside of geosynchronous orbit for much of 29 and 30 October [Dmitriev et al., 2005].
 Usually, the IEF is a good proxy for the strength of the convection electric field. However, Hairston et al.  showed that the polar cap potential drop saturated during the October–November 2003 period.
 The ionospheric response to these intense drivers is extensive, and many aspects of the response have been previously examined. In particular, several studies have examined the behavior of the equatorial anomaly during this storm. Mannucci et al.  presented evidence of a superfountain effect and a global increase in the electron content during the storm. In particular, OSEC measurements from the CHAMP spacecraft in the noon sector (1250–1350 local solar time (LST)) show a rapid (<3.5 hour) increase in the electron content (∼250 total electron content units, 1 TECU = 1016 el/m2) of the anomaly peaks. In addition, the fountain peaks moved poleward by ∼16° magnetic latitude and demonstrated a hemispheric asymmetry with larger OSEC measurements in the southern magnetic hemisphere. They also presented a correlation between the interplanetary electric field and the average TEC observed by ground receivers in the midlatitude, afternoon sector. The ionospheric response in the Southern Hemisphere has been examined by Yizengaw et al. . This study used data from GPS-ground receivers in the Australian sector, the TOPEX altimeter, International Monitor for Auroral Geomagnetic Effects (IMAGE) FUV spectrographic imager, and GPS receivers on board the FedSat satellite. These observations demonstrated a reduction in the Southern Hemisphere TEC on 29–30 October. This TEC reduction was attributed to the negative storm effect [e.g., Seaton, 1956; Crowley et al., 1989a, 1989b; Burns et al., 1995; Fuller-Rowell et al., 1996; Pavlov et al., 2004], which is likely given the increase in thermospheric mass density observed by the CHAMP spacecraft, especially in the Southern Hemisphere [Liu and Lühr, 2005]. Additionally, a tomographic reconstruction of FedSat data indicated a deep reduction of election density above 1500 km at 2220 UT on day 302 in the early morning sector equatorward of −35° magnetic latitude. This study also found enhancements in the dayside equatorial anomaly with higher TEC values in the Southern Hemisphere. A similar study [Lin et al., 2005b] examined the equatorial anomaly in the American and east Asian sectors using two chains of ground GPS receivers, data from the ROCSAT satellite, and Thermosphere, Ionosphere, Mesosphere Energetics and Dynamics (TIMED) Global Ultraviolet Imager (GUVI) measurements. An enhancement and slight poleward motion of the anomaly peaks on day 302 in the east Asian chain was observed. This was followed by TEC reduction in the Northern Hemisphere peak and an absence of the Southern Hemisphere peak on day 303. In contrast, the American sector saw TEC enhancements from roughly −45° to 50° magnetic latitude on day 302. These enhancements occurred around 1920 UT on 29 October. The asymmetry between the TEC in the Northern and Southern hemispheres increased, with the Northern Hemisphere reaching its maximum TEC earlier with a larger TEC and over a larger spatial extent. A similar pattern was reported on day 303, with the Northern Hemisphere TEC enhancement extending to 53° magnetic latitude compared with a Southern Hemisphere extension to −24° magnetic latitude at 1800 UT. The maximum TEC in the Northern Hemisphere occurred at 1930 UT near 24.5° magnetic latitude. The equatorward expansion occurred during periods of stronger vertical drifts measured by ROCSAT. This study was followed by a theoretical study [Lin et al., 2005a], which compared the role of the penetration electric field and neutral winds in generating the motion of the equatorial anomaly peaks. Lin et al. demonstrated that the hemispheric asymmetry in the equatorial anomaly peaks is caused by a combination of the neutral winds and the penetration electric field. The modeling results were in general agreement with the observations but did not fully match them. Finally, Anderson et al.  have compared observed TEC measurements along the western coast of South America with model predictions using vertical ion drifts derived from magnetometer measurements from South America. In general, the model equatorial anomaly peak locations agree well with observations. During this magnetic storm, the equatorial peaks moved poleward to ±20° magnetic latitude.
 In addition to variations at lower latitudes, the ionosphere was significantly altered at higher latitudes. Pallamraju and Chakrabarti  reported observations of the aurora at 48.3° magnetic latitude during the daytime and evening on 30 October. The polar ionosphere was examined by Mitchell et al. . Using a network of GPS receivers, this study investigated the occurrence of scintillation during the storm and presented some space weather maps of the TEC over the northern polar cap. The weather maps were generated by an OA algorithm, the Multi-Instrument Data Analysis System (MIDAS) [Mitchell and Spencer, 2003]. During the Halloween storm, severe scintillation was observed at high latitudes along horizontal density gradients. Additionally, evidence was presented for the occurrence of polar cap patches. Similarly, Foster and Rideout  examined ground-based GPS observations of TEC over North America. The TEC maps show a SED plume that formed over the continental United States during the Halloween storm. This plume appears to stream northwest from the peaks of the equatorial fountain toward the pole. The vertical TEC in this SED plume was over 250 TECU. Nearly simultaneously, a SAPS is observed at roughly 5 hours east of the SED plume, which “drives this broad SED plume” [Foster and Rideout, 2005, p. 2]. Additionally, significant scintillation events were observed in relationship to this SED structure [S. Basu et al., 2005; Su. Basu et al., 2005]. It was argued that the strong density gradients associated with the SED plume help to drive the scintillations [Vo and Foster, 2001].
3. IDA3D Description and Inputs
 The Ionospheric Data Assimilation Three Dimensional (IDA3D) is an objective algorithm, based upon three-dimensional variation (3DVAR) data assimilation. It solves the standard 3DVAR equations for the ionospheric electron density
where xa is the analysis electron density at a given time, xf is the forecast electron density for that time, yo is the set of electron density and electron content observations, Pf is the error covariance matrix for the forecast model, Po is the error covariance matrix for the observations, and the forward operator H and its associated matrix H (with a transpose HT) transforms the predicted electron density to the form and location of the observations. The forecast electron density and error covariance matrix are specified as
where xb is the specification from a background electron density model, τ is the estimated correlation time, tn is the present time step, tn−1 is the last time step, and Pb is the error covariance matrix for the background electron density model. The analysis, forecast, and background model electron densities are stored as vectors (xa, xf, and xb, respectively) for mathematical elegance and computational efficiency. Any known systematic bias to a data source is added into the forward operator H. Since the receiver bias for electron content measurements is almost always unknown, IDA3D uses the relative TEC measurements (in which one of the measured rays at a given time is subtracted from the other rays). A more complete discussion of IDA3D, including the specific treatment of each data type, is presented by Bust et al. .
 Since IDA3D is designed for flexibility, it requires several inputs. The most important inputs are the grid, the background electron density, the assumed model and data errors, and the assumed correlation lengths. For this study, the model grid extends from 75°S to 74°N geographic latitude with a latitude resolution of ∼250 km and a longitudinal resolution of ∼500 km. The vertical grid is 15 km in the bottomside ionosphere, 50 km in the topside below 1000 km, 200 km in the proton-dominated ionosphere (roughly 1000–5000 km), and 1500 km in the plasmasphere. The background electron density is specified on this grid by the Reilly ICED-Bent-Gallagher (RIBG) empirical model [Reilly and Singh, 2001, 2004]. The RIBG model is a smooth compilation of preexisting empirical models of the lower ionosphere (ICED [Tascione et al., 1988]), the topside ionosphere (Bent [Bent et al., 1976]) and the plasmasphere (Gallagher [Gallagher et al., 1988]) designed by Reilly. It is used over other empirical models because the model extends to 40,000 km. RIBG takes as input the location (latitude, longitude, and altitude), month, year, sunspot number, and Kp value. The model variances are set to 90% of the model density. Additionally, the vertical correlation length varies in altitude. In the E region, the correlation length is 20 km, while the F region correlation length is 30 km. The correlation length in the topside is 50–100 km, and the proton ionosphere correlation length is 300–500 km. The plasmaspheric length is 1500 km. The longitudinal correlation length is set to 15° in the midlatitudes and 12° at low latitudes, while the latitudinal correlation length is set at 3.5° equatorward of 15° latitude and 5° poleward. The data variance is assumed to be 10% for every data set except the ionosonde densities in which the variance is assumed to be 25%. It should be noted that the observational error is a combination of instrumentation error and error of representativeness (how well each measurement represents the value of the large-scale density in the voxell) [Petersen and Middleton, 1963]. Typically, the instrumentation error is small and the representativeness error dominates.
 Numerous data sets were available for this study. There were a total of 1107 ground, dual-frequency GPS receivers providing measurements on day 302 and 1094 on day 303. However, not every receiver is located within the model grid, and some receivers are removed because of redundancy. Because the typical receiver tracks around seven spacecraft, the total number of ground-based GPS TEC measurements exceeds 10,000 measurements. To increase computational efficiency, the ground GPS TEC data are decimated so that only one receiver within a 145 km radius is used. After removing these receivers, a typical IDA3D map uses between 250 and 300 receivers reporting 1400–2000 measurements. As mentioned previously, IDA3D uses the relative TEC rather than the absolute TEC. Typically, the highest-elevation angle ray for each instrument is subtracted off as a reference ray. Figure 3 provides a sample (before decimation) of the GPS observations for one time period (1930 UT on 30 October). The left plot shows the location of the GPS receivers that are used, while the right plot shows the location of the 350 km intercept points of each ray. The plots have been centered on the noon local time line. In addition to ground-based GPS slant TEC measurements, this study used data from two different satellite-based GPS receivers on day 302 and four different receivers on day 303. These satellite-based receivers can provide both TEC occultation and OSEC measurements. The relative TEC value is also used for treating these data. Table 1 gives the orbital information, available days, and type of measurement from these GPS receivers.
Table 1. Orbital Information of the Satellites With GPS Receivers
Orbital Altitude, km
Equatorial Crossing Local Time
Type of Measurement
∼0100 to ∼1300
OSEC and Occultation
∼1100 to ∼2300
∼0630- to ∼1830
∼0400 to ∼1600
 GPS TEC measurements do not provide the only electron content measurements. Two different arrays of computerized ionospheric Doppler receiver (CIDR) systems [Bust et al., 2001] made observations of the electron content. CIDRs observe the slant TEC from low Earth-orbiting (LEO) satellites in polar orbits. One array is located in Greenland, while the other is located in Alaska. Additional CIDRs are located at Austin, Texas, and Millstone Hill. Vertical TEC measurements from the TOPEX satellite [Fu et al., 1994] at ∼1000 km altitude are also included. TOPEX measurements are only taken over sea surfaces. Because these TEC measurements originate from lower altitudes, they help separate out the plasmaspheric content from GPS measurements.
 Because electron content measurements represent an integral of electron density, it is important to include electron density measurements. Recently, Garcia-Fernandez et al.  have demonstrated the importance of including ionosonde measurements in a GPS tomography algorithm. To improve IDA3D space weather maps, data from the global network of ionosondes have been included. Figure 4 is a map of the Digisonde and ionosonde stations that provided data for this study. The right plot is a map of the stations on day 302, while the left plot shows the stations for day 303. Because of differences in how Digisonde and ionosonde data are reported, it is necessary to treat different stations differently. Many Digisondes report to the Digital Ionogram DataBase (DIDB) (http://ulcar.uml.edu/DIDB/DIDBHome.html), which provides virtual height and the corresponding inverted true height measurements of the primary Digisonde echo. For these stations, the electron density is determined by treating the emitted frequency as the plasma frequency and using the true height corresponding to observed virtual height. The Automated Real Time Ionogram Scaler with True Height (ARTIST) routine [Reinisch and Huang, 1982, 1983; Galkin et al., 1996] is used to invert virtual height profiles to electron density profiles. These stations are indicated in Figure 4 by asterisks. The pluses indicate stations where only the virtual heights are reported. For these stations, the ARTIST inversions provide electron density values every 10 km. Only density measurements at or below F2 peak are used. Data from DIDB were visually inspected for quality, with easily scaled ionograms manually scaled and poor quality data rejected. Finally, some stations report only the peak densities and height to the Space Physics Interactive Data Resource (SPIDR) (http://spidr.ngdc.noaa.gov/spidr/). For these stations, only the peak values (nmF2, nmF1, and nmE) are used. A cross indicates the location of these stations. Additional electron density observations from spacecraft in situ electron density measurements are also used. Table 2 provides the orbital information for these spacecraft.
Table 2. Orbital Information for Spacecraft Providing in Situ Observations of the Electron Density on Days 302 and 303a
Orbital Altitude, km
Equatorial Crossing Local Time
Orbit Inclination, deg
The information for the Polar spacecraft is limited to measurements within the model grid (altitudes ≤20,000 km and latitudes equatorward of ±75°). NA means not applicable.
∼0100 to ∼1300
∼1715 to ∼0515
∼2030 to ∼0830
∼2115 to ∼0915
∼1200 to ∼2400
4. Low-Latitude and Midlatitude Response to IMF Change
Figures 5, 6, 7, and 8show maps of the storm time differences in the vertical TEC from the climate as a function of solar local time and magnetic latitude at different times during this storm. The storm time differences (TECIDA3D–TECRIBG) are given because they show the storm time changes more clearly than the IDA3D TEC maps alone. It should be noted that in data-poor regions (meaning regions more than one correlation length from an observation), IDA3D densities revert to the background model. As Figure 3 shows, much of the dayside Northern Hemisphere is within 48 min of local solar time of an electron content measurement. The TEC changes are compared with the high-resolution IMF Bz because changes in the IMF Bz have been suggested as a proxy for the strength of the penetration electric field [Kelley et al., 2003]. These maps were chosen to show the development of the superfountain and SED plumes. Figures 5 and 6 are related to the southward turning of the IMF on day 302 around 1845 UT, while Figures 7 and 8 are related to the southward turning around 1900 UT on day 303. For the time period presented in Figures 5 and 6, the Kp was constant at 8.67. The Kp was 9.00 for the time period shown in Figures 7 and 8. For constant Kp conditions within the same month, the RIBG model produces nearly identical electron density maps in a solar-fixed reference frame as a function of UT. Because the background climate is essentially static in the plots shown in Figures 7 and 8, changes in the TEC difference from the climate (IDA3D-RIBG) represent changes in the TEC.
Figure 5a shows the ionospheric changes at 1830 UT on day 302 before the southward turning of the IMF. In general, most of the midlatitude and low-latitude ionosphere experiences small changes in the plasma content (within 10 TECU) from the climate. These are minor changes considering that the mean error associated with the TEC measurements from ground-GPS receivers is roughly 8.25 TECU. However, there are several features which differ from the background climate. The most noticeable change is a daytime TEC reduction in the Northern Hemisphere poleward of 40° magnetic latitude. The plasma content is lower near the equator in the afternoon. A broad depletion region is also observed in the postdusk Northern Hemisphere around 15°. Figure 1b represents the map at 1900 UT after the southward turning. In general, the electron content has changed little since 1830 UT. By 1930 UT (Figure 5c), a region of enhanced TEC forms in the noon sector of the Northern Hemisphere. Associated with this enhancement is a deeper depletion in the daytime equatorial plasma content. By 2000 UT, a weaker enhancement region forms at a slightly later local solar time in the Southern Hemisphere. Over the next 4 hours (Figures 5e and 5f and all of Figure 6), a SED plume (seen as a triangular enhancement region) forms and evolves in the Northern Hemisphere. The SED plume can be seen as a channel of high TEC splitting the dayside depletion region poleward of 40°. Such a narrow region of enhanced TEC is not observed in the Southern Hemisphere, although an enhancement region does form near −45° in Figures 6a and 6b. Instead, a broad region of little or no enhancement forms in the daytime sector of the Southern Hemisphere. This likely reflects a region of little or no data coverage (see Figure 9). In the equatorial region, the plasma content decreases in the afternoon and dusk sectors as would be expected from a continued fountain effect. The dayside regions begin to move to later local times starting around 2200 UT (Figure 6b). This behavior suggests that the enhanced equatorial fountain is a temporary impulse driven by a prompt penetration electric field and that the enhanced plasma regions are raised to altitudes that are not dominated by chemical loss at night.
 The ionospheric response was different on day 303. Prior to the southward turning (1830 UT, Figure 6a), the TEC in the afternoon and evening sector was more depleted in the Northern Hemisphere than in the Southern Hemisphere. In contrast to the southward turning on day 302, the TEC enhancement takes longer to develop (compare Figures 7c and 7d with Figures 5c and 5d). From 2030 UT (Figure 6e) onward, the anomaly peaks in both hemispheres grow and persist. A SED plume appears to develop in the Southern Hemisphere at 2000 UT (Figure 6d). In general, the TEC outside of the SED plume and equatorial anomalies is more reduced on day 303 than on day 302, especially from 1500 to 2000 LST. This leads to steeper TEC gradients associated with the SED plumes on day 303. The horizontal gradient is also enhanced by the deep TEC reduction in the late afternoon (centered around 1600 LST). In this region, substantial plasma is lost (over 10 TECU) between 2000 to 2100 UT. This rapid loss of plasma content implies a dynamic rather than chemical process, especially since the TEC is closer to climate (less reduced) by 2300 UT. Additionally, the connection between the elevated midlatitude contents and the polar cap is more visible with the SED plume expanding to ±60°. Since the aurora oval was located closer to ±50° [Su. Basu et al., 2005; Pallamraju and Chakrabarti, 2005], this may in fact be a tongue of ionization.
5. Altitude Distribution of the Enhanced Plasma Region
 In order to better understand these enhancements, it is useful to view the altitude structure of the ionosphere. Figure 9 shows a series of altitude–geomagnetic latitude slices of the electron density changes at 2100 UT on day 302. The top plot in each pair of plots shows the difference between the log10 analysis density (IDA3D) and the climate density (RIBG), and the bottom plot indicates which grid points were affected by observations in this time step. Regions of enhanced or depleted plasma in the top plot without an accompanying observation in the bottom plot were affected by observations within the assumed Kalman decay time.
Figure 9a shows a significant decrease in the electron density poleward of 40°. This region corresponds to the midlatitude TEC reduction in Figure 9f. There is little change in the electron density elsewhere, but this is likely due to the lack of observations. A similar low-density region is seen extending to 1300 LST (Figure 9b). At 1400 LST (Figure 9c), a density enhancement is seen centered near 40° and 650 km. Only at low altitudes is a depletion seen (∼45° and 250 km). At 1500 LST (Figure 9d), the high-altitude density enhancement is seen around 40° with deep density depletion poleward. At later local times (Figures 9e and 9f), the poleward depletion is observed but not the enhancement. In addition, there are enough data to observe F region depletions in the midlatitude Southern Hemisphere.
Figure 10 shows a series of altitude-latitude slices at 2200 UT on day 303 for a series of solar local times from 12 to 17 hours. Figure 10 presents similar phenomena seen in Figure 9. However, a more complete picture is seen because more data are available. At 1200 LST (Figure 10a), the high-latitude depletion region is seen poleward of 45° at F region altitudes. The greater data coverage presents a clearer image of an enhanced equatorial fountain with two high-altitude (above 400 km) regions near ±35° and a deep depletion at low latitudes. The equatorial anomaly peaks are not as clearly seen at 1400 LST because of a lack of data. However, an enhancement is located at ∼35° and 500 km. This region extends poleward and upward and is the signature of the SED plume. At 1500 LST (Figure 10d), a large enhancement is seen centered around 30° without the poleward extension. In the later afternoon (Figures 10e and 10f), deep electron depletions occur in both hemispheres throughout much of the midlatitude ionosphere.
 To further demonstrate how the electron density changes in altitude, Figure 11 shows magnetic latitude–solar local time maps of the storm time variations in the electron density (as a percent change) for three different ionospheric altitudes at 2200 UT on day 303 in the afternoon and evening sectors. Figure 11 represents the bottomside (207 km), F region (312 km), and topside (825 km) ionosphere. Figure 11 shows that bottomside densities are reduced in the SED plumes, while the topside densities are significantly enhanced.
 The TEC difference maps and the electron density difference slices demonstrate several features of the ionospheric response to a southward turning of the IMF. Both SED plumes and enhanced equatorial anomaly peaks are observed. In addition, the high–electron content region poleward of 50° magnetic latitude is suggestive of a tongue of ionization (TOI). The apparent flow of the SED plume into the polar cap to become a TOI is in agreement with Foster et al. [2005a]. These space weather maps present several features of the ionospheric response, including the following.
 1. In an SED plume, the topside electron densities are significantly greater than the climate, while the bottomside electron densities are lower (see Figure 11). This is in agreement with both early ISR observations of higher topside densities in SED plumes [Foster, 1993] and recent modeling studies showing higher topside densities in the storm time equatorial fountain [Lin et al., 2005a].
 2. A density and content enhancement develops near 1300 LST between 30 and 60 min after a southward turning of the IMF Bz and begins to move into the evening sector after the IMF begins to turn northward.
 3. A triangular-shaped region of increased electron content forms poleward of 30° between 1300 and 1400 LST. This region is associated with an increase in the topside electron density. As the IMF turns northward, this structure begins to convect westward. This structure may be characterized as a SED plume.
 4. A region of low electron density and content exists on both day 302 and day 303 in the dayside Northern Hemisphere above 45°. This depletion region intensifies the horizontal density gradients associated with the SED plume and helps to generate the strong scintillation events observed during this storm [S. Basu et al., 2005; Su. Basu et al., 2005]: (1) The SED plume extends in a narrow band toward the polar cap, and (2) the dayside plasma is depleted in a band of ±18° with an increased TEC poleward of this band. Associated with this band are increased densities and content just poleward of this band. This behavior is indicative of the poleward motion of the equatorial anomaly peaks reported by other authors [Lin et al., 2005a, 2005b; Yizengaw et al., 2005; Anderson et al., 2006].
 These maps are in basic agreement with earlier studies of the equatorial fountain during the Halloween storm [Lin et al., 2005b; Mannucci et al., 2005; Yizengaw et al., 2005; Anderson et al., 2006]. The electron content in the equatorial anomaly peaks are greatly enhanced, and the observed peaks move to ±25° to ±35° poleward of the quiet time locations. Also observed is a region of enhanced plasma in the northern hemisphere at higher latitudes. This enhancement is a SED plume. These maps help illuminate the relationship between SED plumes [Foster and Rideout, 2005; S. Basu et al., 2005] and the equatorial fountain. It has been previously suggested that subauroral polarization streams (SAPS) are important for the development of SED plumes [Foster et al., 2002, 2004]. In this view, the strong SAPS electric field pulls plasma from the peak of an enhanced equatorial fountain and convects the plasma poleward to form the SED plume. Because the equatorial fountain [e.g., Anderson, 1973] and SAPS [e.g., Garner et al., 2004] are conjugate features, this hypothesis argues for simultaneous SED plumes in both hemispheres. However, SED plumes are not always observed in both hemispheres, in part because of a lack of data in the Southern Hemisphere. When SED plumes do form in both hemispheres, the magnitudes of the plumes are significantly different. On day 302 (Figures 5, 6, and 9), the lack of a southern SED plume may be related to a lack of observations. On day 303 (Figures 7, 8, 10, and 11), more data are available, and an SED structure is observed in the Southern Hemisphere. However, the Southern Hemisphere enhancement is smaller in both size and magnitude than the SED plume in the Northern Hemisphere. The lack of conjugacy argues that other factors, such as neutral winds and composition changes, play a role in SED development. Lin et al. [2005a] have recently examined the impact of the penetration electric field, storm time neutral winds, and storm time composition changes upon the equatorial fountain. This study demonstrates a strong hemispheric asymmetry in the location and the width of the equatorial anomaly peaks when the neutral winds are allowed to affect the equatorial fountain. If the northern anomaly peak is more poleward than the southern peak, then the SED formulation mechanism proposed by Foster et al. [2002, 2004, 2005b] becomes more efficient in the Northern Hemisphere. The regions of enhanced plasma seen in Figures 9c and 10c poleward of 45° magnetic latitude are consistent with the flow of plasma out of the anomaly peaks under the influence of a SAPS.
 This paper has used a series of space weather maps produced by the Ionospheric Data Assimilation Three Dimensional (IDA3D) to investigate the ionospheric response during the 29–30 October 2003 Halloween magnetic storms. Like earlier studies, these space weather maps demonstrate the dramatic poleward motion of the equatorial anomaly peaks. These maps also indicate deep daytime depletions in the northern, midlatitude ionosphere, and dusk sector ionosphere on 30 October. Because the depletion region is adjacent to a SED plume, the density gradient associated with the plume is large, which drives the scintillations observed by S. Basu et al.  and Su. Basu et al. . Unlike previous studies, this paper has presented space weather maps in the altitude–magnetic latitude plane. These maps show that much of the additional plasma in an SED plume is located in the topside ionosphere. The SED plumes that are observed occur particularly in the Northern Hemisphere, which argues for a nonconjugate formation mechanism. However, data coverage in the Southern Hemisphere is quite limited, and SED plumes may have occurred that were not observed.
 This work was supported by the Office of Naval Research under grants N00014-97-1-0236 and N00014-05-1-0198, the National Science Foundation under grant ATM-0228467, and the Air Force Office of Scientific Research under grant FA9550-05-1-0316 and by the Applied Research Laboratories, University of Texas at Austin under the ARL:UT postdoctoral fellowship. The authors are grateful to Ruth Skoug for providing the ACE SWEPAM data, Werner Singer and Jens Mielich for providing the Juliusruh/Ruegen Digisonde data, Terrence Bullett and the Air Force Research Laboratories for providing data from the Digital Ionospheric Sounding System, Bodo Reinsch and Ivan Galkin for providing data from the Digital Ionosonde Database, Cesar Valladares for GPS data from his chain of South American stations, Dave Cooke for providing the CHAMP in situ data, and S. Y. Su for providing the ROCSAT electron densities. Additionally, the authors wish to thank the International GPS Service, the Scripps Orbit and Permanent Array Center, and the Continuously Operating Reference Stations network for providing online GPS data; Eric Kihn and the Space Physics Interactive Data Resource at the National Geophysical Data Center for providing Web-accessible peak height measurements from Digisonde and geomagnetic indices; the Coordinated Data Analysis Web at Goddard Space Flight Center for providing the Polar data; the University of Texas at Dallas for providing the online DMSP database; NASA's Global Environmental and Earth Science Information System for providing the satellite-based GPS measurements, and the Open Madrigal Initiative for providing incoherent scatter radar data.