Long-lasting disturbances in the equatorial ionospheric electric field simulated with a coupled magnetosphere-ionosphere-thermosphere model

Authors


Abstract

[1] The Magnetosphere-Thermosphere-Ionosphere-Electrodynamics General Circulation model of Peymirat et al. [1998] is used to investigate ionospheric-wind-dynamo influences on low-latitude ionospheric electric fields during and after a magnetic storm. Simulations are performed with time-varying polar cap electric potentials and an expanding and contracting polar cap boundary. Three influences on equatorial electric fields can be of comparable importance: (1) global winds driven by solar heating; (2) direct penetration of polar cap electric fields to the equator that are partially shielded by the effects of Region-2 field-aligned currents; and (3) disturbance winds driven by high-latitude heating and ion-drag acceleration. The first two influences tend to have similar magnetic local time (MLT) variations in a steady state, while the disturbance-wind influence tends to have the opposite MLT variations. The nighttime disturbance winds at upper midlatitudes that affect the global ionospheric wind dynamo are predominantly westward after the simulated magnetic storm. The nighttime winds drive an equatorward dynamo current that tends to charge the low-latitude ionosphere positively around midnight, which can lead to reductions or reversals of the normal equatorial night-side east-west electric fields. The simulations partly support the theories of the so-called “disturbance dynamo” [Blanc and Richmond, 1980] and “fossil wind” [Spiro et al., 1988], both of which predict long-lasting disturbances in the equatorial eastward electric field associated with magnetic storms. However, the simulations do not support the element of fossil wind theory that links the disturbance-wind influence on equatorial electric fields to polar cap contraction following the storm. The simulations show a stronger wind-produced enhancement of steady state shielding than predicted by the model of Forbes and Harel [1989], due to the fact that the disturbance winds extend well equatorward of the Region-2 currents.

1. Introduction

[2] Ionospheric electric-field disturbances at middle and low latitudes are generated both by time-varying magnetosphere-ionosphere interactions at high latitudes accompanied by direct penetration to lower latitudes [e.g., it Nishida et al., 1966; Vasyliunas, 1970, 1972; Jaggi and Wolf, 1973; Fejer et al., 1979, 1990; Gonzales et al., 1979; Kelley et al., 1979; Peymirat and Fontaine, 1994; Fejer and Scherliess, 1995; Kikuchi et al., 2000] and by secondary effects of disturbance thermospheric winds [e.g., Blanc and Richmond, 1980; Spiro et al., 1988; Forbes and Harel, 1989; Fejer et al., 1990; Fejer and Scherliess, 1995; Fuller-Rowell et al., 2002]. These disturbance electric fields can have a strong influence on the low-latitude ionosphere [e.g., Tanaka, 1981; Kelley and Maruyama, 1992; Deminova, 1995; Lakshmi et al., 1997; Abdu, 1997; Palmroth et al., 2000].

[3] The direct penetration of electric fields from the polar cap to low latitudes is modulated by interaction of the hot plasma in the magnetosphere with the ionosphere. This interaction drives the Region-2 currents that flow into and out of the auroral region equatorward of the polar cap. The ionospheric electric field associated with these Region-2 currents has a tendency to counteract the east-west component of the electric field coming from the polar cap, and thus to reduce the penetration of the polar cap electric field toward middle and low latitudes [e.g., Schield et al., 1969; Vasyliunas, 1972]. This effect is called “shielding” [e.g., Wolf and Jaggi, 1973]. Modeling studies show that the shielding effect is established with a characteristic timescale on the order of 3–300 min, depending on magnetospheric plasma properties and ionospheric conductivity [Vasyliunas, 1972; Jaggi and Wolf, 1973; Wolf and Jaggi, 1973; Southwood and Wolf, 1978; Wolf et al., 1982; Senior and Blanc, 1984; Spiro et al., 1988]. A rapid increase of the polar cap potential is not effectively shielded from lower latitudes on timescales shorter than this, and thus strong, rapidly varying disturbance electric fields can reach the equator. An interesting situation arises when strong magnetospheric convection has been active for an extended period of time and then undergoes a rapid decrease, as can happen, for example, when the interplanetary magnetic field (IMF) suddenly turns northward. The magnetospheric hot plasma that feeds the Region-2 currents, which had been strong enough largely to counteract the directly penetrating electric field from the polar cap, is then temporarily unbalanced, creating a shielding electric field that is too strong: the equatorial electric field is not simply shielded from the polar cap, but is “overshielded,” such that the net equatorial electric field associated with direct penetration can temporarily reverse direction [e.g., Kelley et al., 1979; Spiro et al., 1988; Fejer et al., 1990]. According to model studies, this temporary reversal should last only on the order of 10–60 min [e.g., Spiro et al., 1988; Peymirat et al., 2000].

[4] Equatorial ionospheric electric fields are also affected by thermospheric winds in the dynamo region, which lies primarily between 100 km and 200 km at day and in the lower F region at night. The winds are driven by day-night differences in solar heating, by upward propagating atmospheric tides, by collisional interaction with rapidly convecting ions in the presence of strong high-latitude electric fields, and by Joule heating associated with the strong high-latitude electric currents. The latter two sources are highly variable, and depend on the level of geomagnetic activity. They produce thermospheric disturbance winds, which affect the global ionospheric dynamo [Blanc and Richmond, 1980; Spiro et al., 1988; Fejer et al., 1990; Fuller-Rowell et al., 2002]. Because of the inertia of the neutral air, a few hours are required to set up disturbance winds. Once set up, they can persist for several hours. Thus low-latitude electric-field disturbances associated with the disturbance winds tend to be more persistent than those associated with changes in direct penetration from the polar cap and with changes in the Region-2 currents. Different aspects of the dynamo influences of disturbance winds have been given different names. Banks [1972] and Coroniti and Kennel [1973] pointed out that high-latitude winds accelerated by coupling with the convecting ions can have a dynamo effect that may tend to smooth out fluctuations of magnetospheric-ionospheric convection by acting as what Banks [1972] called a type of “flywheel.” Blanc and Richmond [1980] examined the ionospheric “disturbance dynamo” associated with midlatitude winds that are driven by high-latitude Joule heating. Spiro et al. [1988] examined the dynamo effects of “fossil winds,” which, like winds of the “flywheel,” are accelerated by strong ion convection in the auroral regions, but which can be found lying equatorward of the Region-2 currents during the recovery phase of a magnetic storm, after the polar cap and auroral region have contracted poleward. Because the disturbance winds lie equatorward of the Region-2 currents, any electric fields they generate are not shielded from lower latitudes, and thus may affect the equatorial electric fields.

[5] The present study pursues the earlier fossil-wind studies by Spiro et al. [1988] and Fejer et al. [1990], which used the Rice Convection Model (RCM) [Harel et al., 1981] to quantify the effect of a sudden decrease of the polar cap potential drop that corresponds to a northward turning of the IMF. This decrease results in an immediate overshielding effect, which then decays away. Spiro et al. [1988] and Fejer et al. [1990] compared their simulations with observations of the SUNDIAL 1984 and 1986 campaigns and found general agreement of the amplitudes. However, the observed disturbances persisted for hours, while the simulated disturbances lasted only about 10–60 min. To reconcile the model with the observations, Spiro et al. [1988] examined how fossil winds might affect global electric fields. Only when they specified an equatorward displacement of the neutral wind distribution associated with ion convection in the auroral region during the period of southward IMF, were they able to get long-lasting equatorial electric-field perturbations that agreed with the observations. This equatorward displacement of the winds relative to the Region-2 currents was intended to simulate the contraction of the polar cap, which in reality was held fixed in their simulations. Fejer et al. [1990] pursued this study with the RCM and discussed the fossil wind idea in a more quantitative way. They reached similar conclusions as Spiro et al. [1988], but also proposed another mechanism which should lead to mid- and low-latitude electric field perturbations similar to those of the fossil wind mechanism: the reconfiguration of the magnetospheric magnetic field that occurs when the magnetic activity quiets very abruptly, resulting in a poleward motion of the polar cap necessary to explain the mid- and low-latitude perturbations. However, they did not simulate this proposed mechanism.

[6] In the present study, we use the Magnetosphere-Thermosphere-Ionosphere-Electrodynamics General Circulation Model (MTIE-GCM) built by Peymirat et al. [1998]. The MTIE-GCM couples the model of inner-magnetospheric plasma convection of Peymirat and Fontaine [1994] with the Thermosphere-Ionosphere-Electrodynamics General Circulation Model discussed by Richmond et al. [1992]. Although the MTIE-GCM uses a simple dipole geomagnetic field and thus cannot simulate the influences of magnetic-field reconfiguration proposed by Fejer et al. [1990], it has an advantage over the RCM used in the earlier fossil-wind studies in that the winds in the MTIE-GCM are calculated using the full three-dimensional dynamical equations with realistic forcing. Furthermore, in the present study we make an attempt to simulate realistically the expansion and contraction of the polar cap, instead of artificially moving the wind pattern equatorward with respect to a fixed polar cap boundary.

2. Model Input Parameters

[7] The MTIE-GCM requires as inputs the solar flux, the polar cusp precipitation, the polar rain, the polar cap electric-potential pattern, upward propagating tides at 97 km, and the density and temperature of the magnetospheric plasma in the tail source. We run the model for solar-maximum equinox conditions corresponding to F10.7 = 243 × 10−22 W m−2 Hz−1. The large solar-maximum ionospheric ion densities result in considerably stronger magnetosphere-ionosphere-thermosphere coupling effects than would be expected to occur at solar minimum. Particle precipitation in the polar cap and upward propagating atmospheric tides are disregarded. Six different simulations are performed, the differences corresponding to different polar cap potential models and the inclusion or the neglect of the neutral wind dynamo effect on the ionospheric electric field. We shall often refer to the simulations that include or exclude ionospheric wind dynamo effects as “wind” or “no wind,” respectively, although the wind itself is always calculated. The assumed variations of the polar cap potential and of the magnetospheric plasma source are discussed below.

2.1. Polar Cap Potential

[8] To test the fossil-wind theory suggested by Spiro et al. [1988] with winds that are realistically simulated, one needs to expand and contract the polar cap. Several studies focused on the expansion and contraction of the polar cap [Siscoe and Huang, 1985; Moses et al., 1988, 1989, 1994; Lockwood et al., 1990; de la Beaujardière et al., 1987, 1991; Craven and Frank, 1987; Frank and Craven, 1988; Ridley and Clauer, 1996]. They emphasized that the polar cap boundary is very dynamic during non-steady states, and can contract or expand in complex ways. Following Siscoe and Huang [1985], these studies are based on the assumption that the polar cap boundary can be represented by one or more adiaroic segments that move equatorward or poleward with the plasma flow velocity such that the tangential electric field in the reference frame moving with the plasma vanishes, plus one or more segments with specified tangential electric fields, across which plasma does flow. Moses et al. [1988, 1989, 1994] built the Expanding/Contracting Polar Cap (ECPC) model, which has two non-adiaroic segments, or gaps, one on the day side and the other on the night side. When the potential drop across the day-side gap exceeds that across the night-side gap, the polar cap expands, while it contracts when the night-side potential drop exceeds that on the day side. The locations of the gaps, their widths in magnetic local time (MLT), the amplitude of the potential drops, and as a consequence the polar cap expansion and contraction rates vary with the magnetic activity. One can expect potential drops between 15 and 350 kV, leading to expansion and contraction rates between 0.01 per minute and 0.72 per minute. de la Beaujardière et al. [1987, 1991], Craven and Frank [1987], Frank and Craven [1988], and Ridley and Clauer [1996] observed velocities of this magnitude, between 0.016° and 0.65° per minute.

[9] In this study we simulate the polar cap boundary as a line of given magnetic latitude θ0, with a night-side gap extending from 21.9 MLT to 2.1 MLT and a day-side gap extending from 9.9 MLT to 14.1 MLT. The potential along the polar cap boundary, Φb(φ) is given by

equation image
equation image
equation image
equation image

where φ is MLT, and where Φn and Φd are respectively the values of the potential drops on the night and day sides.

[10] Inside the polar cap the potential is computed similarly to the formalism of Siscoe and Huang [1985], and is given by

equation image

where θ is the magnetic latitude. The coefficients Am are calculated from the Fourier transform of Φb(φ) along the polar cap boundary and are given by

equation image

In practice, the summation is limited to m = 60 for computational purposes.

[11] To calculate the boundary motion dθ0/dt, where t is time, we note that the lengths of the dawn- and dusk-side adiaroic lines are each 7.8 hours of MLT, or 0.65π radians. Letting RE represent the radius of the Earth and ∣Z00)∣ the magnitude of the vertical dipole geomagnetic field at the Earth's surface and at magnetic latitude θ0, we find that the motion of the polar cap boundary is determined by

equation image

We place the polar cap boundary θ0 initially at 71.97°, which corresponds to a grid point of the MTIE-GCM, and let the polar cap evolve according to the above equation.

[12] We analyze three cases with different time variations of Φn and Φd. Cases 1 and 2 represent simulated disturbances that we will call storms, even if the MTIE-GCM cannot fully simulate magnetospheric storm effects because of its fixed magnetospheric field. Case 3 represents quiet steady state conditions. For each case we perform model runs that include and exclude ionospheric wind dynamo effects.

[13] Figure 1 shows the evolutions of Φn and Φd for our first two cases. Case 1 is shown by the solid line. For the first 2 hours, both Φn and Φd are kept constant at 25 kV. This allows the inner magnetosphere to attain a steady state plasma distribution. The day-side potential drop is then increased to 90 kV over 10 min, starting at 2 universal time (UT), initiating the expansion of the polar cap. This expansion stops when the night-side potential drop is increased up to the day-side potential drop, 80 min later. At 3.5 UT, the polar cap boundary has reached 66.72° latitude, having moved equatorward by 5.25 in 1.5 hours from its initial position. The potential drops on the night and day sides are then kept constant for 2 hours to set up a pattern of strong neutral winds. Starting at 5.5 UT the day-side potential drop decreases back to 25 kV over 10 min, and 80 min later the night-side potential drop also decreases, such that the initial position of the polar cap boundary is reached at 7 UT. The potential drops are kept constant for the following three hours.

Figure 1.

Time evolution of the potential drops across the day-side (top) and night-side (bottom) gaps. The solid line is for Case 1 and the dashed line for Case 2 (see text).

[14] The potential drops for Case 2 are illustrated by the dashed lines in Figure 1, and are identical to Case 1 until 5.5 UT. After 5.5 UT both the day-side and the night-side potential drops are reduced simultaneously in two steps (5.5–5.6 UT and 6.93–7.0 UT), such that the polar cap boundary stays at its expanded position of 66.72° magnetic latitude. The two-step reduction of the potential drops is designed to start and end at the same overall times as the reductions in Case 1, in order to keep a somewhat similar driving force on the neutral wind for both runs. This allows us to test the importance of the contraction of the polar cap on the fossil-wind mechanism.

[15] Case 3 keeps both the day-side and night-side potential drops constant at 25 kV, such that the polar cap boundary stays at its initial position of 71.97°.

[16] These particular evolutions of the potential drops in the day and night sides were chosen to allow the polar cap boundary to move by 5° with reasonable contraction and expansion rates. The extensions in MLT of the gaps were chosen to be more or less compatible with observations. Note, however, that for quiet magnetic activity, the polar cap boundary is usually located at magnetic latitudes higher than the 71.97 we take at the beginning of the simulations. For instance, for quiet steady state conditions, Hairston and Heelis [1990], Peymirat and Fontaine [1997], and Foster et al. [1986] showed that the polar cap is located between 73° and 77° invariant latitude for potential drops between 30 and 38 kV. However, because our magnetosphere is a simple dipole, it cannot reasonably approximate realistic magnetic field mapping for lines at such high ionospheric magnetic latitudes. Furthermore, the polar cap boundary is not exactly a circle, as illustrated by Moses et al. [1988, 1989, 1994], and it can be displaced dawnward or duskward of the magnetic pole, depending on the By component of the IMF [Boyle et al., 1997; Hairston and Heelis, 1990]. These facts are neglected in the present study.

[17] Figure 2 shows the electric potential between the pole and 60° magnetic latitude for Case 1 when the ionospheric wind dynamo is included. (The ionospheric wind dynamo affects the potential only equatorward of the polar cap boundary in the MTIE-GCM.) The bold dashed line shows the polar cap boundary. Each frame corresponds to one of the times A–E of Figure 1, that is, just before the storm, near the end of the expansion phase, at the end of the expanded phase, near the end of the storm during the contracting phase, and 80 min into the recovery phase. At time A, the potential drops across the day-side and night-side gaps are both 25 kV, and the polar cap boundary is located at 71.97° magnetic latitude. At time B, the day-side potential drop is larger than the night-side potential drop. More magnetic flux enters the polar cap than leaves, so that the polar cap expands. At time C, the day and night sides potential drops are equal to 90 kV, and the polar cap boundary is stationary at 66.72° magnetic latitude. At time D, the night-side potential drop is larger than that on the day side, and the polar cap contracts. At time E, the two potential drops have recovered their initial values of 25 kV, and the polar cap has contracted to its initial position of 71.97° magnetic latitude.

Figure 2.

Ionospheric electric potentials for Case 1, including wind dynamo effects, poleward of 60° magnetic latitude at the times A–E shown in Figure 1. The contour interval is 2 kV, where solid and dashed contours show positive and negative potentials, respectively. The maximum and minimum values are given on the left axis in kilovolts. The bold dashed line shows the polar cap boundary.

2.2. Magnetospheric Plasma Source

[18] At the initial time of the simulations, the magnetospheric plasma source is located at 10.44 Re, which is the equatorial conjugate of the position of the polar cap boundary at 71.97° magnetic latitude. As the polar cap boundary moves during the simulations, the temperature and the density of the source plasma are adjusted so as to maintain compatibility with the plasma just inside the boundary.

[19] Statistical observations of the ion and electron populations show that the average values of the density range from 0.05 cm−3 to 1 cm−3 and the ion temperature from 1 keV to 8.7 keV in the plasma sheet [Huang and Frank, 1986; Huang et al., 1989, 1992; Baumjohann and Paschmann, 1989; Baumjohann et al., 1989; Goertz and Baumjohann, 1991; Huang and Frank, 1994; Escoubet et al., 1997]. They also show that those quantities are strongly dependent on the magnetic activity and the position in the plasma sheet, dependence that we do not consider in our simulations. At the initial time of the simulations, the density is set to 0.4 cm−3, the ion temperature of the source is set to 5 keV, and the electron temperature is set to 1 keV.

[20] We calculated the electron and ion precipitation at respective rates of 40% and 10% those for strong pitch angle diffusion. The electron rate is an average between that used by Spiro et al. [1988] and that of Schumaker et al. [1989], which showed that the electrons tend to precipitate with roughly 50% the strong pitch angle diffusion rate for energies greater than 1 keV.

3. Results and Discussion

3.1. Global Patterns of Wind and Electric Potential

[21] Figure 3 shows the influence of the wind on the electric potentials. Unlike Figure 2, the lowest magnetic latitude plotted is 10°. The top three rows are for Case 1, in which the polar cap contracts at the end of the storm, while the bottom row is for Case 2, in which the polar cap does not contract at the end of the storm. The times A, C, and E that are shown for Case 1 represent pre-storm, pre-recovery, and post-recovery. For Case 2 the results at times A and C are identical to those for Case 1, and so only time E is shown. All of these times are at least 80 min after the end of a change in either the day-side or night-side polar cap potential drop, so that any residual effects associated with transient under- or overshielding should be small. On the left are the electric potentials for the cases that include ionospheric wind dynamo effects. In the middle are shown the wind patterns at 150 km, which is roughly representative of the height-averaged winds, weighted by the Pedersen conductivity, that largely drive the ionospheric dynamo. On the right are the Region-2 field-aligned current densities, that is, the current densities for latitudes that are linked to the magnetospheric portion of the MTIE-GCM. (Field-aligned current densities are not shown for the polar cap.)

Figure 3.

Ionospheric quantities poleward of 10° magnetic latitude for times A (pre-storm), C (pre-recovery), and E (post-recovery) for Case 1 with winds, and at time E for Case 2 with winds. (left column) Electric potentials in kilovolts with a contour interval of 2 kV. (middle column) Neutral winds at an altitude of 150 km. (right column) Downward Region 2 currents in amperes per square meter with a contour interval of 10−7 A m−2. Solid and dashed contours are for positive and negative values, respectively, and the maximum and minimum values are indicated on the left axis.

[22] For Case 1, the polar cap potentials are essentially identical at times A and E, but the potentials at middle and low latitudes show significant differences. This is primarily due to the different wind patterns at times A and E: the winds above 50° are considerably stronger at time E, and the winds below 50° are also significantly changed. In particular, we note that the night-side winds tend to be more westward at time E than at time A. The westward shift of the wind causes an equatorward shift of the wind-driven Pedersen current, which causes an increased positive charging of the low latitudes, as seen in the electric potential. At 10°, the greatest increase in the nighttime potential at time E with respect to time A is around midnight.

[23] The winds shown in Figure 3 differ from the winds used in the modeling studies of Spiro et al. [1988], Forbes and Harel [1989], and Fejer et al. [1990]. Those three studies assumed that the disturbance winds are primarily driven by collisional coupling with the rapidly convecting ions at high latitudes, without taking into account Joule heating effects or momentum transport away from the acceleration region of the winds. Spiro et al. [1988] and Fejer et al. [1990] assumed that the wind velocity at all heights after the magnetospheric disturbance was simply one-half of the plasma convection velocity present during the main phase of the disturbance. They did not account for the tendency of the wind pattern to rotate with the Earth. Although they did not show their wind distribution, it should have the same pattern as the plasma convection, and should therefore be generally westward in the afternoon and evening and eastward in the morning at auroral latitudes. This is roughly compatible with our 150 km winds at time C, during the time the storm is still active. However, it is less compatible with our winds at time E, nearly 3 hours later, which tend to be westward during most of the night and eastward during most of the day. Because the winds of Spiro et al. [1988] and Fejer et al. [1990] are rotated toward earlier times with respect to ours, we would expect their winds to produce the maximum low-latitude disturbance potential closer to dusk than to midnight.

[24] Forbes and Harel [1989] used two models of height-varying disturbance winds, one of which took Earth rotation into account. In the Pedersen conductivity region their nighttime disturbance wind that considered Earth rotation was westward between 18 and 2.6 MLT equatorward of the polar cap, more compatible in this respect with our winds in Figure 3 at time E. However, Forbes and Harel [1989] did not allow the ion-convection-driven wind to move equatorward with respect to the acceleration region, whereas the winds in Figure 3 at times C and E extend considerably equatorward of the auroral acceleration region. Therefore, the influence of disturbance winds on the low-latitude potential found by Forbes and Harel [1989] was less than what our simulations show, in part because ionospheric wind-dynamo effects within the auroral region are partly shielded from lower latitudes in the same manner that polar cap fields are shielded from lower latitudes, and in part because the region of disturbance winds used by Forbes and Harel [1989] was smaller and farther from midlatitudes than what our simulations show.

3.2. Eastward Electric Field at the Magnetic Equator

[25] The top portion of Figure 4 shows, for Case 1, the eastward electric field as a function of MLT at simulation times A, C, and E, on geomagnetic field lines that intersect the 90 km altitude level at a magnetic latitude of 11.645°. The apexes of these field lines reach 364 km altitude over the magnetic equator, and thus Figure 4 is representative of the equatorial F-region eastward electric field (apart from a small reduction in amplitude due to the slight spreading of geomagnetic field lines as one maps the electric field upward and equatorward along the field lines). Solid lines are for the simulation that includes ionospheric wind dynamo effects, while dashed lines are for the simulation that neglects these wind effects. The bottom portion of Figure 4 shows the difference electric fields for these two simulations with respect to the respective control runs of Case 3 (no storm) with and without wind effects. Before the storm (time A) the difference electric fields are zero.

Figure 4.

Eastward electric fields at 11.645° magnetic latitude and at 90 km altitude, for times A (pre-storm), C (pre-recovery) and E (post-recovery) of Figure 1. (top) Total eastward electric field for Case 1 with (solid) and without (dashed) winds (see text). (bottom) Differences between the eastward electric fields of Case 1 and those of the quiet Case 3 for runs with (solid) and without (dashed) winds.

[26] At time A the simulated equatorial eastward electric field, including wind effects, has the general features of observed fields: it is eastward at day and westward during most of the night, except for a large eastward enhancement in the early evening, the so-called “pre-reversal enhancement.” The responsible winds are predominantly the global winds driven by solar heating of the thermosphere. Fesen et al. [2000] showed that the addition of atmospheric semidiurnal tidal forcing at the lower boundary of the TIE-GCM, which we have neglected, can modify the electric field pattern without changing these basic features. It is a noteworthy coincidence that the equatorial electric field shows similar features even without the ionospheric wind dynamo (dashed lines), although the electric field amplitude is typically only about one third as large. The equatorial electric field in the no-wind simulation is caused by penetration of the imposed polar cap electric field to lower latitudes, due to incomplete steady state shielding.

[27] At time C, just before the end of the main phase of the storm, the total equatorial eastward electric field (top of Figure 4) for the run including wind effects (solid line) retains the same general features as at time A, but becomes somewhat stronger. The electric field for the run excluding wind effects (dashed line) also retains the same general features as at time A, but becomes much stronger, comparable with the run including wind effects. The difference between the solid and dashed lines at time C is considerably smaller than at time A, indicating that the influence of winds on the equatorial electric field has become smaller. That is, the dynamo effects of the disturbance winds are tending to cancel those of the normal quiet-time winds. This approximate cancellation is probably fortuitous, as we would expect the strength of the disturbance winds to depend on the strength and time history of the storm, which have been arbitrarily specified for our study. However, we expect that the tendency for the influence of the disturbance winds to oppose influence of the quiet-time winds on equatorial electric fields may be a general characteristic of disturbance winds. Since the MLT variations of the steady state penetration electric field are similar in form to those of the quiet-time ionospheric wind dynamo, i.e., the dashed curves at times A and C in Figure 4 are similar in form to the difference between the solid and dashed curves at time A, though different in amplitude, the disturbance-wind influence on the equatorial electric field also tends to oppose the steady state penetration field. This is seen in the bottom central panel of Figure 4, during the storm main phase at time C. The dashed line shows the storm-produced change in the incompletely shielded penetration electric field, which is considerably larger than the total quiet-time penetration field shown by the dashed line in the upper left panel of Figure 4. Note that steady state shielding should be in effect at time C, since the polar cap electric potential has been steady for over 2 hours. In comparison, the storm-produced change in the total electric field that includes wind effects, shown by the solid line, is considerably weaker. Again we are finding a fortuitous near-coincidence in the amplitudes of two opposing effects: the amplitude of the incompletely shielded disturbance penetration electric field and the amplitude of the electric field generated by the disturbance winds. For a different assumed time history of the storm such an approximate cancellation of the two disturbance effects would not necessarily hold. Nevertheless, we find that the inclusion of disturbance wind effects has a tendency to offset the electric field disturbance associated with steady state direct penetration.

[28] Note that the reduction of the disturbance in the equatorial eastward electric field when wind effects are included at time C is in the same sense as the reduction associated with steady state shielding by the Region-2 currents. Spiro et al. [1988], Forbes and Harel [1989], and Peymirat et al. [1998, 2002] discussed how the winds can increase the effective shielding, by driving ionospheric currents that tend to oppose those driven by the Region-2 field-aligned currents, and therefore have a similar effect as that produced by a reduction in Pedersen conductivity. These studies emphasized the winds which are accelerated within the auroral region, coincident with the region of strong auroral electric fields and ion convection. However, we noted earlier that the winds at time C in Figure 3 differ from those at time A not only in the auroral region, where the field-aligned current densities are largest (above 55° magnetic latitude), but also at latitudes equatorward of the auroral region, where they can have larger effects on the equatorial field.

[29] At time E, after the storm, the equatorial electric field for the Case 1 run including wind effects has become more eastward with respect to time A at MLTs of 0–6, and less eastward at MLTs of 15–23. These changes can readily be seen in the difference plot on the bottom of Figure 4 for time E, and reflect the increased low-latitude electric potential around midnight at that time seen in Figure 3. The eastward shift of the early-morning field is in the same sense as observed for post-disturbance equatorial electric fields in this time sector [e.g., Kelley et al., 1979; Spiro et al., 1988; Fejer et al., 1990]. There is a much smaller post-storm electric field disturbance for the run excluding wind effects (dashed line). This disturbance is associated with the increased storm-time injection of plasma into the inner magnetosphere, which persists long after the storm as a decaying ring current. The increased plasma density changes the distribution of field-aligned currents and tends to increase the strength of steady state shielding. The directly penetrating electric field therefore tends to be reduced. However, this effect on the equatorial electric field is much smaller than the effect of the post-storm disturbance winds.

3.3. Importance of Polar Cap Contraction

[30] If the polar cap is kept at its storm-time expanded size while the cross-polar cap potential is reduced, as in Case 2, the post-storm low-latitude electric fields are considerably different from those of Case 1. The bottom row of Figure 3 shows the patterns of the potential, wind, and field-aligned current at time E for Case 2, to be compared with those from Case 1 directly above. The field-aligned current patterns look similar relative to the position of the polar cap boundary, but the currents for Case 2 lie about 5° equatorward of those for Case 1, because of the different boundary positions. The wind patterns are quite similar. The low-latitude electric potentials show significant differences between Cases 1 and 2.

[31] The top portion of Figure 5 shows the eastward equatorial electric field at time E for both Cases 1 and 2, for the simulations that include wind effects. Case 2 (dot-dashed curve), as compared with Case 1 (solid curve), has a considerably stronger westward field between 2 and 5 MLT, and a more eastward field between 18 and 22 MLT. Compared with the pre-storm fields shown by the solid curve at time A in Figure 4, the equatorial electric field has changed more for Case 1 than for Case 2. That is, the disturbance is larger when the polar cap contracts, in apparent agreement with the fossil wind theory of Spiro et al. [1988] and Fejer et al. [1990]. However, it turns out that the different equatorial electric field disturbance amplitudes between Cases 1 and 2 has little to do with differences in the wind influences for the two cases. The difference between the eastward equatorial electric fields for the Case 2 and Case 1 simulations that neglect wind effects (not shown) is found to be very similar to the difference between the two curves on the top of Figure 5. The bottom portion of Figure 5 presents the wind influences on the electric field in a different way than used in Figure 4: the solid curve shows the difference between the eastward equatorial electric fields simulated with and without wind for Case 1, while the dot-dashed curve shows the corresponding difference for Case 2. Clearly, the post-storm wind influence on the equatorial electric field is nearly the same whether or not the polar cap is contracted.

Figure 5.

Eastward electric fields at 11.645° magnetic latitude and at 90 km altitude, for time E (post-recovery) of Figure 1. (top) Total eastward electric field for Cases 1 (solid) and 2 (dot-dashed), with winds (see text). (bottom) Differences between the total eastward electric fields for runs with and without winds, for Cases 1 (solid) and 2 (dot-dashed).

[32] This result disagrees with one of the key elements of the fossil-wind theory presented by Spiro et al. [1988] and Fejer et al. [1990]: that contraction of the polar cap after the storm is important in order to have the disturbed winds lie equatorward of the shielding region. Even for moderately sized storms like the ones simulated here, the winds are disturbed well equatorward of the Region-2 currents, and these winds influence the equatorial electric field in a similar manner whether or not the polar cap contracts after the storm. Part of what disturbs the winds equatorward of the Region-2 currents is the effect described by Blanc and Richmond [1980]: high-latitude Joule heating drives winds equatorward into midlatitudes, which tends to cause westward winds to develop through action of the Coriolis force. In addition, the high-latitude equatorward winds on the night side, which are driven both by ion convection and by Joule heating, are large enough to help transport momentum out of the polar region. The net result is disturbed winds at upper midlatitudes, especially on the night side and predominantly in the westward direction, that create ionospheric wind dynamo influences which extend to the equator.

[33] The reason why Case 2 produces a smaller post-storm disturbance in the eastward equatorial electric field than Case 1 appears to be caused by a weakened shielding effect when the polar cap is expanded. When shielding is reduced, the amplitude of the directly penetrating electric field that reaches the equator, as represented by the dashed curves in the top frames of Figure 4, will increase. Since this penetration electric field is generally in the opposite sense to the equatorial electric field disturbance produced by the disturbed wind, it tends to reduce the net electric field disturbance at the equator, more so when the polar cap is expanded than when it is contracted. The reduced shielding when the polar cap is expanded can possibly be explained partly by a reduced volume of magnetospheric plasma that contributes to shielding, and partly simply by the fact that the polar cap is closer to the equator, so that there is less geometrical attenuation of the penetration electric fields.

3.4. Temporal Variations at Fixed Locations

[34] The primary observations relating to fossil wind effects have come from a ground-based station (the Jicamarca radar in Peru, at the magnetic equator) and so it is useful to examine how the equatorial electric fields from our simulations would appear to a ground-based observer. We choose two hypothetical sites located in longitude such that they reach 5 MLT and 19 MLT at time E, corresponding approximately to the maximum and minimum difference electric fields for the normal storm case at that time.

[35] Figure 6 shows the storm-induced changes of the electric field that would be observed at these two sites, differenced from the background run. The eastward field at 90 km and 11.645° magnetic latitude represents that in the F region over the magnetic equator. The upper plot is for a site where the storm begins at 12.67 MLT, and the site then rotates with the Earth into the evening sector. The lower plot is for a site where the storm begins at 22.67 MLT, and the site then rotates into the morning sector. Whenever there is an abrupt change in either the day-side or night-side polar cap potential drop, abrupt changes of the electric field are seen. These changes rapidly reach a peak and then tend to decay partially away, illustrating either the effects of electric field penetration and the subsequent development of shielding, or else the effects of overshielding and its subsequent decay, depending on whether the polar cap potential has increased or decreased. In addition, the electric field can grow or diminish even after shielding has approached a steady state, owing to the rotation of the observing location into or out of a region of stronger electric field. There are also some irregular changes of the electric field during the periods when the polar cap is either expanding, approximately between times A and B, or contracting, approximately between times C and D. These are artifacts of the numerical procedure, representing times when the expanding or contracting polar cap boundary passes a latitude of fixed MTIE-GCM grid points. That is, these irregular changes result from the changing interpolation between the fixed grid points.

Figure 6.

Differences between the eastward electric fields of Cases 1 and 3 (storm minus quiet) with (solid) and without (dashed) winds, at 11.645° magnetic latitude and 90 km altitude for two fixed locations. (top) Time variations of the difference fields for the location which, at the post-recovery time E, has rotated with the Earth to 19 MLT, where the maximum westward disturbance field is found (see Figure 4). (bottom) Time variations of the difference fields for the location which, at the post-recovery time E, has rotated with the Earth to 5 MLT, where the maximum eastward disturbance field is found.

[36] For both locations shown in Figure 6, the wind-dynamo effect tends to reduce the amplitude of the difference electric field during the storm between times B and C, but to increase it following the storm (after time D + 10 min), as can be seen by comparing the solid and dashed curves. The effect of the wind, given by the difference between the solid and dashed curves, is generally as large as the effect of the penetration electric field, that is, the difference electric field between Cases 1 and 3 without winds, that is illustrated by the dashed line. As noted earlier, the wind effect tends to oppose the steady state penetration electric field, and it produces a prolonged disturbance following the storm that is westward around 19 MLT and eastward around 05 MLT at time E.

4. Conclusions

[37] In our simulations, we find comparable influences on equatorial electric fields due to three effects: (1) global winds driven by solar heating; (2) direct penetration of polar cap electric fields to the equator that are partially shielded by the effects of Region-2 field-aligned currents; and (3) disturbance winds driven by high-latitude heating and ion drag acceleration. Not only are the amplitudes of these three effects often similar, but their variations with respect to MLT tend to be either similar or opposite. That is, in a steady state (1) and (2) tend to have similar MLT variations, while (3) tends to have MLT variations of opposite sign to (1) and (2). Thus the disturbance winds tend to decrease the equatorial electric field disturbance during a storm, but to increase it after the storm. Furthermore, other studies [e.g., Spiro et al., 1988; Fejer et al., 1990] have shown that the direct-penetration field (2) can essentially reverse sign under conditions of overshielding, when the polar cap potential rapidly decreases. The combinations of these various effects can give rise to enhancements, reductions, reversals, or distortions of the usual pattern of equatorial electric fields, depending on the timing and strength of the geomagnetic disturbance.

[38] The nighttime disturbance winds at upper midlatitudes are predominantly westward after the geomagnetic storm. They appear to be produced by a combination of factors, including (1) equatorward flow out of the auroral region driven by increased high-latitude Joule heating, that leads to westward wind acceleration due to the Coriolis force, and (2) direct ion drag acceleration of winds in the expanded auroral region during the geomagnetic storm, together with momentum transport out of the auroral region due to the strong high-latitude equatorward winds. The nighttime westward disturbance winds at upper midlatitudes drive an equatorward dynamo current that tends to charge the low-latitude ionosphere positively during the night, especially around midnight. This changes the eastward gradient of the low-latitude potential in the evening and early morning and therefore changes the east-west electric field at these times. The electric fields generated by the winds tend to oppose the normal equatorial night-side east-west electric fields.

[39] Our simulations partly support both the disturbance-dynamo theory of Blanc and Richmond [1980] and the fossil wind theory presented by Spiro et al. [1988] and Fejer et al. [1990]. Both theories predict that long-lasting disturbances in the equatorial eastward electric field similar to those that have been observed, generally opposite in direction to both the quiet-day field and the steady state penetration electric field, can be produced by alterations of the upper-midlatitude thermospheric winds during a geomagnetic storm. Our simulated upper-midlatitude disturbance winds are influenced both by Joule heating and by ion drag acceleration in the auroral region, as proposed separately by the two theories. However, our simulated winds vary strongly with MLT, unlike those of Blanc and Richmond [1980], and they extend considerably farther equatorward from the auroral region, and tend to be shifted later in MLT, than those postulated by Spiro et al. [1988] and Fejer et al. [1990].

[40] Our simulations do not support the element of the fossil wind theory that links the disturbance-wind influence on equatorial electric fields to polar cap contraction following the storm. Instead, we find that the influence of the disturbance winds on equatorial electric fields depends very little on polar cap contraction. Polar cap contraction can indeed lead to larger post-storm changes in the equatorial electric field, but those larger changes are attributed mainly to a reduced direct-penetration electric field, a field that tends to offset the disturbance electric field produced by the winds.

[41] Our simulations support the prediction by Spiro et al. [1988] and Forbes and Harel [1989] that winds which are accelerated by high-latitude ion convection tend to enhance steady state shielding, by reducing the net electric field that penetrates to middle and low latitudes. However, we find that storm-time disturbance winds can cause a considerably stronger reduction in the penetration electric field at the equator than predicted by the model of Forbes and Harel [1989], due to the fact that the disturbance winds extend well equatorward of the Region-2 currents.

Acknowledgments

[42] The National Center for Atmospheric Research is sponsored by the National Science Foundation. This study was supported by the NASA Sun-Earth Connection Theory Program.

[43] Shadia Rifai Habbal thanks the referee for his assistance in evaluating this paper.

Ancillary