The equatorial electrojet during geomagnetic storms and substorms

The climatology of the equatorial electrojet during periods of enhanced geomagnetic activity is examined using long‐term records of ground‐based magnetometers in the Indian and Peruvian regions. Equatorial electrojet perturbations due to geomagnetic storms and substorms are evaluated using the disturbance storm time (Dst) index and auroral electrojet (AE) index, respectively. The response of the equatorial electrojet to rapid changes in the AE index indicates effects of both prompt penetration electric field and disturbance dynamo electric field, consistent with previous studies based on F region equatorial vertical plasma drift measurements at Jicamarca. The average response of the equatorial electrojet to geomagnetic storms (Dst<−50 nT) reveals persistent disturbances during the recovery phase, which can last for approximately 24 h after the Dst index reaches its minimum value. This “after‐storm” effect is found to depend on the magnitude of the storm, solar EUV activity, season, and longitude.


Introduction
During periods of elevated geomagnetic activity, equatorial ionospheric electric fields and currents undergo significant deviations from their quiet day patterns (e.g., Fejer [2002] for a review). Two mechanisms have been proposed to account for the generation of equatorial ionospheric electrodynamic effects for geomagnetically disturbed conditions. One is the prompt penetration of the high-latitude electric field to lower latitudes [Nishida, 1968;Gonzales et al., 1979], and the other is the ionospheric dynamo due to storm time thermospheric winds [Blanc and Richmond, 1980]. Simulation studies have shown that electric field perturbations produced by these two processes can be comparable at equatorial latitudes [Richmond et al., 2003;Maruyama et al., 2005].
The prompt penetration of the high-latitude electric field is most evident when the magnetospheric convection suddenly increases or decreases. Under steady magnetospheric conditions, the inner magnetosphere tends to be shielded from the magnetospheric convection field [Wolf, 1995]. In other words, the middle-and low-latitude ionosphere is largely shielded from the effect of the high-latitude electric field. When the magnetospheric convection abruptly increases, the middle-and low-latitude ionosphere is exposed to the influence of the enhanced dawn-to-dusk convection electric field in the high-latitude ionosphere until the magnetospheric configuration readjusts and a new state of shielding is established. In contrast, a rapid decrease in the magnetospheric convection causes a temporary excess of the dusk-to-dawn shielding electric field, which affects lower latitudes until a reestablishment of the shielding is attained. The time scale for the shielding processes is typically less than 1 h [Kikuchi et al., 2000;Peymirat et al., 2000], but it depends on magnetospheric conditions [Senior and Blanc, 1984;Maruyama et al., 2007]. During the main phase of a geomagnetic storm, the penetration electric field is sometimes observed to last for several hours without decay [Kelley et al., 2003;.
The ionospheric wind dynamo is an electrodynamic process that generates electric fields and currents as the electrically conducting atmosphere moves through the geomagnetic field [Richmond, 1995]. During a geomagnetic storm, the dynamo electric fields and currents are altered as the wind generated by high-latitude Joule heating and ion-drag forcing disturbs the normal quiet time thermospheric circulation. The theoretical basis of the storm time disturbance dynamo was established by Blanc and Richmond [1980]. The high-latitude forcing produces equatorward winds, which turns westward at middle and low latitudes due to the action of the Coriolis force. The westward wind drives equatorward currents, which build up positive charges near the equator. Poleward electric fields are thus set up, and they drive poleward Pedersen currents that substantially balance the wind driven equatorward currents. The poleward electric YAMAZAKI AND KOSCH ©2015. The Authors. fields also produce eastward Hall currents, which cause positive charge accumulation at the dusk terminator and negative charge accumulation at the dawn. The effect of the disturbance dynamo at equatorial latitudes is, therefore, a generation of a westward electric field on the dayside and eastward electric field on the nightside, which opposes the quiet time pattern. An additional mechanism was suggested by Fuller-Rowell et al. [2002], where storm time meridional wind surges drive eastward currents at middle latitudes, which would produce similar equatorial electric fields and currents as the Blanc and Richmond theory. This mechanism enables a rapid disturbance dynamo onset within an hour or two after the high-latitude energy input, because the slow buildup of westward winds due to the Coriolis force is not involved.
The low-latitude electrodynamic response to high-latitude geomagnetic activity has been extensively studied by B. G. Fejer and his colleagues using a large data set of vertical plasma drift measurements at Jicamarca (see reviews by Fejer [1981Fejer [ , 2002). In particular, Scherliess [1995, 1997] and Scherliess and Fejer [1997] introduced a technique to determine the average response of the equatorial vertical plasma drift to variable high-latitude forcing. By binning the data with respect to times of large increase or decrease in the auroral electrojet index AE, they were able to show how the disturbance electric field depends on the time history of auroral electrojet activity. Their approach was successful especially for the nighttime when disturbance signals were large. Daytime effects were, however, found to be small in the drift data, and the characteristics remained to be clarified.
Understanding the response of the daytime electric field to geomagnetic activity is important because it has a significant impact on the dayside ionospheric plasma distribution during geomagnetically disturbed periods [e.g., Mannucci et al., 2005]. Ground-based magnetometer data have been often used to study the daytime equatorial electrojet response to geomagnetic activity. A number of case studies have been made to provide evidence for the effect of the prompt penetration electric field and disturbance dynamo electric field [e.g., Sastri, 1988;Kikuchi et al., 2003Kikuchi et al., , 2008Veenadhari et al., 2010;Le Huy and Amory-Mazaudier, 2005;Zaka et al., 2009]. A weakness of case studies is that they do not consider the effect of quiet time day-to-day variations, which could be comparable to disturbance effects. The quiet time day-to-day variability of the equatorial electrojet arises primarily from irregular changes in the neutral wind caused by meteorological forcing from the lower atmosphere [Yamazaki et al., 2014a]. The neglect of the quiet time variability is a significant issue particularly for the disturbance dynamo effect, which is usually identified by merely comparing the data for an event day against a reference quiet day. Meanwhile, a statistical approach such as work by Fejer and Scherliess [1995] has an advantage that it can average out the contribution of the quiet time variability by using a large data set. In the present study, we statistically analyze long-term records of geomagnetic data to reveal characteristics of the equatorial electrojet during periods of enhanced geomagnetic activity. The main objective is to establish the average (or climatological) response of the daytime equatorial electrojet to geomagnetic storms and substorms.
Geomagnetic storms and substorms are separate phenomena, involving different magnetospheric processes [e.g., Kamide and Maltsev, 2007]. Both phenomena are initiated by the injection of solar wind energy into the magnetosphere, and thus, they occur under similar solar wind conditions. As a result, almost all geomagnetic storms are accompanied by substorms, and most intense substorms occur during geomagnetic storms. Nonetheless, their behaviors are sometimes very different. For example, multiple (5-10) substorms with a period of 3-4 h are often observed during a single event of an intense geomagnetic storm [e.g., Huang, 2005;Troshichev and Janzhura, 2009]. The present study does not attempt to completely separate the contributions of the two phenomena, which would require a statistical analysis of storm events unaccompanied by a substorm as well as a statistical analysis of substorm events without a geomagnetic storm. Both geomagnetic storms and substorms involve enhanced energy input into the high-latitude thermosphere/ionosphere that affects the equatorial electrodynamics but they act on different time scales. A typical duration of a substorm is a few hours, while that of a geomagnetic storm is several hours to days. In this paper, the terms "geomagnetic storms" and "substorms" are used only to distinguish their time scales. For substorms, we investigate hour-to-hour responses between the AE index and equatorial electrojet perturbations. Meanwhile, for geomagnetic storms, responses between the disturbance storm time (Dst) index and equatorial electrojet perturbations are examined on the basis of a 6 h integration of data, which is longer than a typical duration of a substorm.

Data
We used hourly ground magnetometer data to derive the strength of the equatorial electrojet in the Indian and Peruvian regions. Following standard practice, a proxy for the equatorial electrojet intensity was obtained by taking the difference in the magnitudes of the horizontal (H) component of the geomagnetic field at a dip-equatorial station, where the dip latitude is within ±3 • and an off-equatorial low-latitude station of the same longitude sector [e.g., Rastogi, 1989;Anderson, 2011]. This substantially removes the effect of magnetospheric currents from the data. The difference in H is denoted as ΔH. The baseline of ΔH was calculated as the average of the five-hourly nighttime values starting from midnight. We first determined the solar cycle, seasonal, lunar time, and local (solar) time-dependent quiet day (Kp < = 2) values ΔH Quiet using basically the methodology described by Yamazaki et al. [2011]. The ΔH Quiet is then subtracted from ΔH. The residuals ΔH − ΔH Quiet are due to the effect of the prompt penetration electric field and disturbance dynamo electric field, as well as due to quiet time day-to-day variability. It is assumed that the quiet time day-to-day variations are independent of geomagnetic activity, while the variations due to the prompt penetration electric field and disturbance dynamo electric field depend on geomagnetic activity and its time history. Therefore, the quiet time day-to-day variations can be removed by averaging ΔH − ΔH Quiet data under certain geomagnetic activity conditions.

Figure 1.
Equatorial electrojet perturbations in the Indian sector at the substorm times t 1 -t 8 as defined in Table 1. The circles indicate the average value at each local time with error bars representing the standard error for the average. The solid curves show a smooth fit of a local time function as given by equation (1).
Hourly values of the Dst index and the AE index are used to quantify activities of geomagnetic storms and substorms, respectively. We use the corrected Dst index (or D cx ) given by Mursula and Karinen [2005].

Response to Substorms
The response of the equatorial electrojet to rapid changes in auroral electrojet activity was examined by binning the ΔH − ΔH Quiet data according to the time history of the hourly AE index, similar to the method introduced by Fejer and Scherliess [1995]. The binning criteria are given in Table 1, where AE 0 is the AE value for the present hour and AE n (n = 1, 2, 3, or 4) is the AE value for n hours prior. The substorm times t 1 -t 8 represent different stages of auroral electrojet activity, as indicated in Table 1. Figure 1 illustrates the average response of the equatorial electrojet to auroral electrojet activity in the Indian sector. The eight panels show equatorial electrojet perturbations at substorm times t 1 -t 8 , which are defined in Table 1. In each panel, the circles indicate the average value of ΔH − ΔH Quiet at each local time.
The error bars have a length of twice the standard error for the average. A smooth fit of a local time function is also indicated. The fitting function used is where t is local time in hour and A k and B k are coefficients that can be determined by least squares fitting. The fitting was done to all the available data, not to the average data points at each local time.
The substorm time t 1 is when the AE index is suddenly increased after at least 3 h of quiet periods. It corresponds to substorm onset. (See Table 1 for the average AE indices for each substorm time.) Generally, the equatorial electric field perturbation after substorm onset could be either eastward or westward. Huang [2012] showed that the westward electric field perturbation is in many cases related to a northward turning of the interplanetary magnetic field (IMF), which often coincides with substorm onset. Huang [2012] also showed that the equatorial electric field perturbation after substorm onset is eastward when the IMF is continuously southward around the onset without a northward turning of the IMF. Our results for t 1 involve substorms both with and without a northward turning of the IMF. The results indicate that the average electrojet perturbation at substorm onset is dominated by an eastward electric field, which results from the prompt penetration of the dawn-to-dusk convection electric field to equatorial latitudes. The enhanced YAMAZAKI AND KOSCH ©2015. The Authors. eastward electric field fades away in 1 h, and a westward disturbance electrojet starts to dominate as geomagnetic activity remains high (t 2 and t 3 ). This decay of the eastward electrojet can be attributed to two processes. First, the development of the dusk-to-dawn shielding electric field reduces the effect of the dawn-to-dusk convection electric field. Second, a westward electric field due to the disturbance dynamo effect develops on the dayside low-latitude ionosphere.
The results for t 4 indicate that the westward disturbance electric field dominates when high auroral electrojet activity persists at least for four hours. Although the shielding process is completed at this stage, the enhanced high-latitude convection electric field leaks to lower latitudes and influences the equatorial electrojet. The westward disturbance electrojet at t 4 could be contributed by the eastward steady-state penetration electric field but be dominated by the westward disturbance dynamo electric field. The prompt penetration electric field under steady state conditions has been noticed in numerical simulations [Peymirat et al., 2000;Zaka et al., 2010], but often overlooked in interpreting observations. For example, Fejer and Scherliess [1995] and Fejer et al. [2008] attributed the disturbance electric field after four hours of high geomagnetic activity to the sole effect of the disturbance dynamo.
A sudden decrease in the AE index gives rise to a strong westward perturbation in the equatorial electrojet as indicated by the results at t 5 . It is known that a rapid northward turning of the IMF from a steady southward condition, and thus a rapid substorm recovery, is often followed by a transient augmentation of the westward electric field at equatorial latitudes [e.g., Rastogi and Patel, 1975;Fejer et al., 1979]. This is owing to the dusk-to-dawn shielding electric field that is left behind after the abrupt decrease of the dawn-to-dusk convection electric field . The temporal progression from t 5 to t 7 demonstrates a gradual decay of the westward disturbance electrojet due to attenuation of both shielding electric field and disturbance dynamo electric field. The electrojet perturbation is negligible at t 8 after at least 4 h of quiet geomagnetic conditions. Figure 2 shows the same as Figure 1, but for the Peruvian sector. The equatorial electrojet response to the changes in AE is largely consistent with the Indian sector results. Thus, our discussion on the Indian sector results is valid for the Peruvian sector results as well. The effect of the prompt penetration electric field at t 1 is larger in the Peruvian region than in the Indian region. This is probably because the effective ionospheric conductivity is greater in the Peruvian region due to weaker background geomagnetic field .
Our results shed light on the daytime effect, which Fejer and Scherliess [1995] found difficult to clearly resolve in their data set. The results can be interpreted in terms of the penetration electric field and disturbance dynamo electric field, but it is difficult to know which is how much. We believe that our results at t 1 -t 7 are more or less affected by both processes. The separation of the two contributions would be possible only through a comparison with numerical simulations that include both mechanisms.

Response to Geomagnetic Storms
For geomagnetic storm effects, ΔH-ΔH Quiet were binned according to the time history of the hourly Dst index. All the storm events with the minimum Dst index less than −50 nT were identified for the period 1957-2011 (total 1324 storm events), and the time when the Dst reaches its minimum value was assigned as t min . For multiple-onset storms, each phase with the minimum Dst index less than −50 nT was treated separately. Figure 3 shows the average time series of the Dst index as a function of time from t min . The multiple-onset storms were carefully divided into pieces of events so that the same Dst value would not be used more than once in averaging. The results illustrate a typical geomagnetic storm with the main phase magnitude of Dst(t min )=−94.7 nT. The recovery phase can be recognized as a gradual recovery in Dst after the main phase. The storm onset, which is often characterized by a storm sudden commencement, is not visible in Figure 3. This is because the data were sorted with respect to t min , not the onset time. Since we will focus on the equatorial electrojet response during the main phase and recovery phase, the exact time for the onset is not important for our results. We defined storm times T 1 -T 8 according to the time with respect to t min . T n is a 6 h interval from 6(n − 3)+1 + t min to 6(n − 2) + t min , so that T 1 and T 2 are in the developing phase of the storm while T 3 -T 8 are in the recovery phase. The time intervals for T 1 -T 8 are indicated in Figure 3. Figure 4 presents the response of the equatorial electrojet to geomagnetic storms in the Indian sector. Different panels show ΔH-ΔH Quiet at different storm times T 1 -T 8 . (See Figure 3 for the time intervals for T 1 -T 8 .) The results demonstrate the development and decay of the electrojet perturbation during the average geomagnetic storm. The electrojet perturbation is mainly westward throughout the period we investigate. The westward disturbance in the equatorial electrojet persists for approximately 24 h after t min (i.e., from T 3 to T 6 ). Such a long-lasting effect can result from the disturbance dynamo electric field, which has been predicted to last for many hours after high-latitude forcing ceases [Blanc and Richmond, 1980;. Indeed, in previous studies, equatorial electrojet perturbations during the recovery phase were often attributed to the disturbance dynamo effect [Le Huy and Amory-Mazaudier, 2005;Zaka et al., 2009]. It should be noted, however, that we cannot rule out possible contributions of the steady-state penetration electric field that slowly attenuates during the recovery phase. Figure 5 is the same as Figure 4 but for the Peruvian sector. The results for the Indian and Peruvian sectors show some similarities and differences. During the increase of storm activity at T 1 and T 2 , a westward disturbance develops in both regions. Equatorial electrojet perturbations at T 1 and T 2 may include the contribution of short-term eastward penetration electric field but be dominated by the westward disturbance dynamo electric field. The local time for the maximum westward disturbance slightly shifts to later local times at the transition from the main phase to the recovery phase (i.e., from T 2 to T 3 ) in both Indian and Peruvian regions, which is probably due to changes in the penetration electric field from high YAMAZAKI AND KOSCH ©2015. The Authors. latitudes. A marked difference in the results for the two regions is in the pattern of electrojet perturbations during the recovery phase. That is, the results for the Indian sector show a westward disturbance with a single peak around the noon, while the Peruvian sector results reveal a semidiurnal variation with a westward disturbance in the morning and eastward disturbance in the afternoon. The difference may result from a longitudinal dependence in the disturbance winds. The electrojet perturbation during the recovery phase persists for approximately 24 h in both regions. The duration of the effect (i.e., ∼24 h) may be  We now take a closer look at the "after-storm" effect, which is evident during the recovery phase. The results for T 4 and T 5 are combined at each longitude sector, and regrouped according to the magnitude of the storm, solar EUV activity, and season. Concretely, the storm magnitude binning is based on the Dst value at t min (i.e., the minimum Dst value), and binning criteria are Dst(t min ) > −70 nT for weak storms; −100 < Dst(t min ) ≤ −70 nT for moderate storms; and Dst(t min ) ≤ −100 nT for strong storms. For the solar EUV activity binning, we use the index P [Richards et al., 1994], which is defined as the average of the daily F 10.7 index and its 81 day mean. The binning criteria are P ≤ 80 sfu, 80 < P ≤ 180 sfu, P >180 sfu, where sfu denotes the solar flux unit, 10 −22 W m −2 Hz −1 . The seasonal binning is based on the three Lloyd seasons, i.e., D months consisting of November, December, January, and February; E months consisting of March, April, September, and October; and J months consisting of May, June, July, and August. Figure 6 depicts how the electrojet perturbation during the recovery phase depends on the magnitude of the storm. The results indicate that a larger storm leads to a stronger effect. In the Peruvian sector, not only the westward disturbance in the morning but also the eastward disturbance in the afternoon increases with increasing storm intensity. A larger storm involves more energy input to the high-latitude upper atmosphere, which would drive stronger and longer-lasting disturbance winds and resulting disturbance dynamo electric field. Figure 7 shows equatorial electrojet perturbations during the recovery phase for different solar EUV activity conditions. The after-storm effect tends to be more significant for higher solar activity. This is probably due to enhanced ionospheric conductivities during high solar flux periods. For quiet conditions, daytime ionospheric dynamo currents are approximately twice as strong during solar maximum in comparison with solar minimum [e.g., Takeda, 1999Takeda, , 2002. Besides, numerical experiments by Huang [2013] showed that the disturbance dynamo electric field is stronger for higher solar flux conditions for the same storm, although the mechanism was not examined.  Seasonal effects, presented in Figure 8, reveal longitudinal differences. In the Indian sector, the westward disturbance shows a strong annual modulation with maximum effect during the D months and minimum effect during the J months. The seasonal variation is barely visible in the Peruvian sector results, except that the eastward disturbance in the afternoon is largest during the D months. It is probable that storm-time disturbance winds vary with the season, so that the resulting disturbance dynamo electric field is also seasonal dependent. The dependence of disturbance winds on the season is not well understood. The most comprehensive empirical model, DWM07 [Emmert et al., 2008], does not include seasonal variations. Further studies will be required in order to understand the seasonal dependence of the disturbance equatorial electrojet, presented in Figure 8. It may be noted that the equatorial electrojet flows in the Northern Hemisphere at the Indian sector, while the Peruvian electrojet flows in the Southern Hemisphere. Such a difference in geographical conditions adds to the complexity in how the disturbance dynamo electric field changes with season at different longitudes.
It is interesting that the results for the J months show a semidiurnal perturbation in both regions, i.e., a westward disturbance in the morning and an eastward disturbance in the afternoon. A semidiurnal perturbation in the quiet time equatorial electrojet is often a manifestation of semidiurnal tidal forcing from the lower atmosphere [e.g., Yamazaki et al., 2014b]. It may be possible that the upward propagating semidiurnal tides are altered during geomagnetic storms due to changes in the background thermospheric wind. The feasibility of this mechanism is yet to be studied.

Summary
Using long time series of ground magnetic field measurements, we have examined the climatology of the equatorial electrojet response to geomagnetic storms and substorms in the Indian and Peruvian sectors. Substorm effects have been determined by sorting the equatorial electrojet perturbations according to the time history of the AE index. It has been shown that a sudden increase in the AE index leads to an enhancement of the daytime eastward electrojet, indicating the penetration of the high-latitude convection electric field to lower latitudes. The enhanced eastward electrojet fades away in 1 h, and as geomagnetic activity remains high, a westward disturbance electrojet starts to dominate, which is a manifestation of the disturbance dynamo electric field. A rapid reduction in AE is followed by an intensification of the westward disturbance electrojet, which can be attributed to the dusk-to-dawn shielding electric field. These results are in good agreement with previous studies based on F region equatorial vertical plasma drift measurements [e.g., Fejer and Scherliess, 1995].
Storm effects have been examined by arranging equatorial electrojet perturbations with respect to the time for the minimum Dst. The disturbance electrojet develops as the storm intensifies, and it persists for approximately 24 h during the recovery phase for the average storm with the minimum Dst value of −94.7 nT. The electrojet perturbations during the recovery phase are likely to be due to the disturbance dynamo electric field, which has been predicted to last for many hours after geomagnetic activity subsides [Blanc and Richmond, 1980;. In the Indian sector, the after-storm effect is characterized by a westward disturbance with a maximum around the noon, while in the Peruvian sector, the effect is more semidiurnal with a westward disturbance in the morning and an eastward disturbance in the afternoon. Further analysis has revealed that the after-storm effect is dependent on the magnitude of the storm, solar EUV activity, and season. That is, the amplitude of the electrojet perturbations during the recovery phase tends to increase with an increase in the storm magnitude and solar flux level. The after-storm effect in the Indian sector shows a strong annual modulation with maximum and minimum effects during northern winter and summer, respectively. Meanwhile, the seasonal variation is not so apparent in the Peruvian sector.