1. Modeling the Thermosphere and Ionosphere
 Interest in the storm time response of the midlatitude ionosphere has recently had a revival due to the contribution of a new, unique data source. Networks of ground-based, dual-frequency, GPS receivers are now able to map the response of the total electron content (TEC) to storms [Ho et al., 1996; Mannucci et al., 1998; Lu et al., 1998]. The maps are either produced by building up a picture of the individual measurements of vertical TEC derived from the slant path measurements [Tsurutani et al., 2004], or can be generated using data assimilation techniques by ingesting the observations into a Kalman filter. Figure 1 shows an example of the latter method [Spencer et al., 2004]. Figure 1 shows a map of the TEC over the contiguous United States (CONUS) at 2000 UT on two days. On the left is for 17 August 2003, a day of quiet geomagnetic activity, and on the right is the pattern following a geomagnetic disturbance on the following day. On the storm day a region of depleted ionosphere forms at midlatitude, possibly in response to changing neutral composition during the storm. The high-latitude magnetospheric sources heat the upper atmosphere, drive a storm time global circulation from pole to equator, and cause the molecular rich neutral gas to be upwelled into the upper thermosphere. This molecular nitrogen-rich neutral air is transported equatorward by the background and storm circulation where it can increase ionospheric loss rates. Figure 1 shows the consequence of these processes.
 The new observational technique also shows evidence of large increases in total electron content. Figure 2 (right), shows the response at 1945 UT on 31 March 2001, a time when the magnetosphere is strongly driving the upper atmosphere, compared with a quiet day on the left, four days earlier. Several features are apparent. The first is the much larger TEC at low latitudes; the second is the region of storm-enhanced density (SED) stretching from the east coast of the United States diagonally north and west across Canada to high latitudes. This SED has been shown by Vo and Foster , Foster et al. , and Foster and Vo  to be a consistent feature of storms when the United States is at the right local time and when the magnetosphere is strongly driving the upper atmosphere. The mechanism for the SED is uncertain but is possibly due to the transport of midlatitude dayside plasma on the equatorward edge of a subauroral polarization stream [Kelley et al., 2004; Foster and Burke, 2002].
 High-velocity plasma drifts can also give rise to depletions [Schunk et al., 1975], which further complicates the picture. In regions where the drift is high, the ions are frictionally heated via ion-neutral collisions, which raises the ion temperature. The elevated temperature increases the plasma loss rate due to the rapid conversion of O+ to NO+. The combination of transport and enhanced plasma loss can give rise to steep gradients in TEC.
 At low latitude the structure of the ionosphere is strongly controlled by electrodynamics. During quiet times, the electric fields are driven by a combination of the E and F region dynamo processes [Fesen et al., 2000; Millward et al., 2001; Fejer, 1981, 1991; Fejer et al., 1995]. The net result at the magnetic equator is eastward electric fields, or upward plasma drift, during the day, downward drift at night, and a prereversal enhancement (PRE) just after sunset. The upward plasma transport induced by the electrodynamics on the dayside, generates a strong equatorial ionization anomaly (EIA) and at times produces the conditions conducive to the generation of ionospheric irregularities. The latter are notoriously difficult to predict on a day-to-day basis.
 During geomagnetic storms, the dynamo electric fields are entirely altered because the normal quiet day thermospheric neutral winds are disrupted during a geomagnetic storm. Blanc and Richmond  were the first to describe the characteristics of the storm time disturbance dynamo, and their results are strongly supported by observations [Fejer and Scherliess, 1995, 1997; Fejer and Emmert, 2003]. The Blanc and Richmond theory relies on the build up of zonal winds at midlatitude under the action of the Coriolis force, in response to the increased equatorward winds. The meridional winds are forced by high-latitude heating. The dynamo action of the zonal winds causes a redistribution of charge, and the production of electric fields that tend to oppose the normal quiet time structure.
 The Blanc and Richmond theory predicts that the disturbance dynamo is slow to develop because of the gradual build up of the zonal winds, and also slow to abate. An additional mechanism was mentioned by Blanc and Richmond, and was explored by Fuller-Rowell et al.  in numerical simulations. The new mechanism appears to provide a means of generating a disturbance dynamo response about an hour or two after the onset of a geomagnetic storm. This second mechanism is driven by the meridional wind surges that respond within an hour to the high-latitude heating. Numerical simulations also indicate that the meridional winds “slosh” backward and forward from pole to pole in response the dynamic high-latitude forcing. These oscillations of the midlatitude F region neutral wind can induce a similar oscillation in the dynamo electric fields at equatorial latitude.
 The low-latitude ionosphere responds to the disturbance dynamo zonal electric fields by raising or lowering the height of the low-latitude plasma. The time sequence and magnitude of the disturbance dynamo will depend on the strength of the zonal winds and the strength and phase of the meridional wind surges. At times, the height of the ionosphere can be raised in the predawn sector leading to conditions that are ripe for the initiation of plasma bubbles, or irregularities, from the Rayleigh-Taylor instability mechanism. These irregularities cause scintillation in ground-to-satellite radio signals at a range of wavelength including UHF and GPS L band frequencies [Basu et al., 1996; Fejer and Kelley, 1980; Fejer et al., 1999; Groves et al., 1997].
 In addition to the dynamo fields, penetration electric fields are also a major source of disruption of the low-latitude ionosphere during geomagnetic storms. When the high-latitude magnetospheric convection increases, usually associated with a southward turning of the interplanetary magnetic field (IMF), the high-latitude electric fields are unshielded as the magnetospheric plasma begins to respond. As a result, the electric fields can penetrate directly to the equator [Kelley et al., 1979; Kelley, 1985; Spiro et al., 1988; Fejer et al., 1990].
 The empirical model of Fejer and Scherliess  attempts to separate the prompt penetration and disturbance dynamo electric fields using the time history of the AE geomagnetic index. The Rice magnetospheric convection model (RCM) is in good agreement with the empirical penetration electric field model [Fejer and Scherliess, 1997]. The magnitude of the penetration fields can be significantly larger than dynamo fields, but their duration tends to be shorter, typically lasting less than an hour. There has been speculation recently that penetration electric fields can remain unshielded for several hours [Huang et al., 2005] and that the magnetospheric response to “superstorms” is somehow different from that during more modest disturbances. Whether this is a fundamental change in the magnetospheric “system response” time is difficult to determine. Two other possibilities exist. The first is that the system response is the same but that an increasing solar wind electric field driver can continue to force the magnetosphere beyond the more typical shielding response time. The second is that penetration fields can be confused with rapid onset dynamo fields that can appear within two hours of an event and before wave surges have actually reached the equator [Fuller-Rowell et al., 2002]. Whether of short or long duration, penetration fields can be intense and can cause significant redistribution of plasma at low latitudes.
 Penetration electric fields are also generated in response to a decrease in high-latitude convection, often associated with a northward turning of the IMF, but with opposite sign. Even though penetration and dynamo fields can become confused in this overshielded case, the Fejer and Scherliess  model is able to separate the two effects. If the solar wind is highly variable, a series of prompt penetration electric fields will be initiated at low latitudes. The penetration field will interfere constructively or destructively with the dynamo component depending on the particular time history of the solar wind.
 Satellite observations of the dramatic changes that can occur at low latitude in response to geomagnetic storms were presented by Basu et al. . They showed data from the DMSP polar orbiting satellite at 850 km altitude during the Bastille Day storm in July 2000 indicating that a wide swath of plasma had disappeared over tens of degrees in latitude. A similar event occurred during the March 1989 storm [Greenspan et al., 1991] when vertical drift measurements exceeded 100 m/s. Modeling this event using the observed drifts raised the F region ionosphere to over 800 km [Batista et al., 1991].
 This huge restructuring of the ionosphere at low latitudes, as depicted in the DMSP observations, has never been successfully modeled. Numerical simulations are able to produce similar effects at 400 km altitude from the action of the various dynamo processes, but this in no way compares to the extreme changes that are observed. The most likely scenario is that the dynamo and prompt penetration fields are acting together, due to the particular time history of the solar wind drivers. Currently there is no physics-based thermosphere ionosphere plasmasphere model that combines the effects of both dynamo and penetration electric fields during storms.
 However, numerical simulations have been performed recently that compare the relative contribution of penetration and disturbance dynamoelectric fields on the restructuring of the midlatitude and low-latitude ionosphere [Maruyama et al., 2005]. The geophysical conditions for the storm event on 31 March 2001 are shown in Figure 3. Figure 3 shows the time history of the solar wind interplanetary magnetic field (IMF) Bz component, and the geomagnetic indices, AE, Kp, and Dst (SYM-H).
 Figure 4 compares the equatorial vertical plasma drift from both the penetration electric fields, as computed by the Rice Convection Model (RCM) [Sazykin et al., 2005], and the disturbance dynamo, as computed by the coupled thermosphere ionosphere plasmasphere electrodynamics model (CTIPe) [Millward et al., 2001; Fuller-Rowell et al., 2002]. A full description of the simulations and references to the two models can be found in the work by Maruyama et al. . Four simulations are shown for the storm on 31 March, for two different longitude sectors. On the left is longitude 127° in the Asian and Australian sector, and on the right is longitude 289° in the North American sector. The normal quiet time diurnal variation of the vertical drifts from the dynamo is shown as the dashed line, the storm time disturbance dynamo drifts from CTIPe are in black, and the RCM penetration electric field drifts in blue. The red curve shows the results from the CTIPe simulation where the RCM drifts are driving CTIPe in addition to the self-consistent disturbance dynamo drifts. In the latter case, the penetration drifts can alter the winds and conductivities so that the disturbance dynamo response is different. The response is therefore not the linear sum of the RCM and dynamo drifts alone.
 Note that the Asian/Australian sector (on the left) is on the dayside during the first period of Bz south and the response is dominated by the penetration fields. When this same sector rotates onto the night side, and the second pulse hits, the dynamo and penetration are of roughly equal importance. The North American sector (on the right) is on the night side during the first pulse and the combined response tends to follow disturbance dynamo fields. When the North American sector moves onto the dayside the disturbance dynamo tends to cancel the upward drift from penetration, resulting in a somewhat weaker response in this sector.
 The response of the F region ionosphere near the peaks in southward Bz, at 0700 and 1900 UT on 31 March are shown in the bottom plots of Figures 5 and 6, respectively. Figures 5 and 6 are from CTIPe simulations where both dynamo and penetration fields are included. The quiet reference days are shown for comparison in the top plots of Figures 5 and 6. Each plot shows the F region peak, in geographic coordinates, from 0° to 360° longitude, and from south to north pole. In this fixed Universal Time presentation, the local timescales at the bottom are different in Figures 5 and 6. At the earlier time, at 0700 UT in Figure 5, the Asian sector, at longitude 127°, is on the dayside near 1530 LT, and penetration electric fields transport the equatorial plasma upward (see Figure 4), moving the equatorial ionization anomaly (EIA) poleward by more than 10° in latitude compared to the quiet reference day. The separation of the EIA peaks approximately doubles on the dayside in response to the increased upward plasma drift. In the 2000 to 2100 LT sector the response to the prereversal enhancement (PRE) weakens, and on the night side between 0200 and 0300 LT over the North American sector, the upward drift mainly from the action of the disturbance dynamo produces a feature similar to that expected from a PRE. It is this feature that typically creates the conditions conducive to the formation of ionospheric irregularities in the post midnight or predawn sectors during storms. At this early phase of the storm depicted in Figure 5 the midlatitudes and high latitudes are only just beginning to be depleted, and the negative phase is still fairly weak. At the later storm time, 1900 UT, shown in Figure 6 (bottom plot), the predawn minimum in the 0200 to 0600 LT sector at the magnetic equator dominates the night side over Asian and Australian sectors, in contrast to the quiet time pattern. On the dayside over the North American sector between (1000 and 1600 LT), the EIA develops early because of the larger upward drift compared to the quiet day (see Figure 4) and because of preconditioning by the earlier event. The negative phase at midlatitudes and high latitudes is now much more exaggerated because of the transport of neutral composition changes.
 Note that Figures 4, 5, and 6 are from numerical simulation, and that detailed comparison with observations are needed to confirm the results and elucidate the relative importance of the various physical processes as attempted by Fedrizzi et al. .