Explicit IMF By‐Dependence in Geomagnetic Activity: Quantifying Ionospheric Electrodynamics

Geomagnetic activity is mainly driven by the southward (Bz) component of the interplanetary magnetic field (IMF), which dominates all solar wind coupling functions. Coupling functions also depend on the absolute value of the dawn‐dusk (By) component of the IMF, but not on its sign. However, recent studies have shown that for a fixed level of solar wind driving, auroral electrojets in the Northern Hemisphere (NH) are stronger for By > 0 than for By < 0 during NH winter. In NH summer, the dependence on the By sign is reversed. While this By sign dependence, also called the explicit By‐dependence, is very strong in the winter hemisphere, it is weak in the summer hemisphere. Moreover, the explicit By‐dependence is much stronger in the westward electrojet than in the eastward electrojet. In this study, we study how the explicit IMF By‐dependence is coupled with large‐scale field‐aligned currents (FACs) by using FAC measurements from the Active Magnetosphere and Planetary Electrodynamics Response Experiment and an empirical ionospheric conductance model. We model the complete ionospheric electrodynamics by solving the current continuity equation, and show that during periods of elevated solar wind driving (Bz < 0), the IMF By component modulates Regions 1 and 2 FACs in the dawn sector of the winter hemisphere. This leads to an explicit By‐dependence in ionospheric conductance and the westward electrojet. We also show that the By‐dependence of FACs and conductance is weak in the dusk sector, which explains the earlier observation of the weak By‐dependence of the eastward electrojet.

the dawn-dusk (B y ) component is included in the Newell function, its effect is symmetric with respect to its sign, that is, the value of dΦ MP /dt does not depend on the sign of B y . The same applies to all other solar wind coupling functions (Borovsky & Birn, 2014;Kan & Lee, 1979;Milan et al., 2012;Perreault & Akasofu, 1978;Vasyliunas et al., 1982). However, several studies based on measurements by ground-based magnetometers (Friis-Christensen et al., 1975;Laundal et al., 2016;Murayama et al., 1980) and polar-orbiting satellites (Friis-Christensen et al., 2017;Smith et al., 2017) have shown that auroral electrojets in the Northern Hemisphere (NH) winter are stronger in both hemispheres for B y > 0 than for B y < 0. During NH summer, the dependence on the sign of B y is reversed. Holappa and Mursula (2018) quantified this explicit B y -effect in detail and showed that for a fixed value of dΦ MP /dt the NH westward auroral electrojet (measured by the AL-index) is about 40%-50% stronger for B y > 0 than for B y < 0 during NH winter solstice. This explicit B y -effect is much weaker during NH summer solstice, when the AL-index is only about 10% greater for B y < 0 than for B y > 0.
The physical mechanism of the explicit B y -effect has so far remained poorly understood. It is well known that IMF B y plays important roles in several magnetospheric and ionospheric dynamics. For example, IMF B y affects the geometry of the dayside reconnection (Park et al., 2006;Sonnerup, 1974;Trattner et al., 2012) and exerts magnetic stress on opened field lines, leading to dawn-dusk asymmetries in magnetospheric convection (Cowley et al., 1991;Tenfjord et al., 2015). This is also reflected in the distribution of fieldaligned currents (FACs), which are known to be modulated by IMF B y especially in the dayside (Iijima et al., 1978;D. L. Green et al., 2009;Weimer, 2001). The above mechanisms also affect the ionospheric parameters, where IMF B y is known to modulate the spatial distribution and amplitude of ionospheric electric fields (Heppner & Maynard, 1987;Ruohoniemi & Greenwald, 1996;Weimer, 1995). Several radar-based studies (Pettigrew et al., 2010;Ruohoniemi & Greenwald, 2005;Thomas & Shepherd, 2018) have shown that the cross-polar cap potential is larger for B y > 0 (B y < 0) in NH winter (summer) which is consistent with the explicit B y -dependency found in geomagnetic activity.
Recently, Reistad et al. (2020) showed that the size of the polar cap in both hemispheres (estimated from the latitudinal extent of Regions 1/2 FAC current system) exhibits an explicit B y -dependence: For a fixed level of solar wind coupling, the polar cap size is greater for B y > 0 (B y < 0) under negative (positive) dipole tilt angle. Thus, the B y -dependence of the polar cap size is similar to that of geomagnetic activity, implying that there is an explicit B y -dependence in the dayside reconnection rate. This mechanism obviously contributes to ionospheric currents by modulating the global magnetospheric convection. However, it cannot explain why IMF B y affects geomagnetic activity strongly only in the winter hemisphere, or why the B y -effect is stronger in the westward than in the eastward electrojet (Holappa & Mursula, 2018). Recently, Holappa et al. (2020) showed that there is an explicit B y -effect also in the flux of energetic (>30 keV) electrons precipitating into the atmosphere (measured by NOAA POES satellites). The B y -modulation of particle precipitation has a much higher relative contribution to ionospheric conductance in the winter hemisphere than in the summer hemisphere (Shue et al., 2001), which explains the above summer-winter difference in the B y -effect. Moreover, Holappa et al. (2020) found that the B y -effect in the electron fluxes is weaker in the dusk sector than in the midnight-dawn sector, which explains the relatively weak B y -dependence of the eastward electrojet.
The first goal of this study is to quantify the B y -dependence in ionospheric conductance in detail. Estimating ionospheric conductance from particle precipitation is not a straightforward task because, in addition to the number of fluxes of precipitating electrons and protons, ionization also depends on their energy spectra (Robinson et al., 1987). Global distribution of auroral particle precipitation could be estimated by empirical models, such as the OVATION model (Newell et al., 2014), based on empirical relations between DMSP particle precipitation measurements and dΦ MP /dt. However, because the OVATION model is driven by a coupling function, which is symmetric with respect to the sign of IMF B y , it cannot be used to study the explicit B y -dependence of particle precipitation (or conductance).
Conductance driven by particle precipitation is also known to be closely related to FACs (Lyons, 1980;Wallis & Budzinski, 1981). This connection is most direct in regions of upward FAC, which are co-located with regions of downward precipitating electrons Robinson et al., 2018). Regions of downward FAC are also associated with increased conductances, but a downward FAC corresponds to a smaller conductance than an upward FAC of the same amplitude. Thus, ionospheric conductance can be estimated from FACs using empirical relations (that are different for upward and downward FAC regions) which is also done, for example, in global MHD models (Mukhopadhyay et al., 2020;Ridley et al., 2004;Wiltberger et al., 2009). In this study, we use a statistical model (Robinson et al., 2020) relating ionospheric height-integrated conductances to FACs, measured globally by the Active Magnetosphere and Planetary Electrodynamics Response Experiment (AMPERE), based on the magnetic measurements made by Iridium satellites (Anderson et al., 2008(Anderson et al., , 2000Waters et al., 2001). In addition to particle precipitation, ionospheric conductance is affected by seasonally varying solar photoionization (A. E. S. Green et al., 1964;Moen & Brekke, 1993), which we take into account using empirical relations, parametrized by the solar radio flux F10.7 (Robinson & Vondrak, 1984;Robinson et al., 2020).
The second goal of this study is to quantify the B y -dependence in the auroral electrojets in different MLT sectors. We reach this goal by determining horizontal (Hall and Pedersen) ionospheric currents from a continuity equation, which connects the horizontal currents to measured FACs and modeled conductances using the method recently developed by Robinson et al. (2021). There are two main advantages of using the AMPERE measurements and the Robinson et al. (2020) conductance model. First, the spatial distribution of modeled conductance is consistent with the spatial distribution of FACs, which is known to be strongly dependent on IMF B y (D. L. Green et al., 2009;Iijima et al., 1978). Second, the high temporal coverage of AMPERE measurements yields sufficient statistics for extracting the B y -dependence.
This study is organized as follows. In Section 2, we describe the AMPERE data, the conductance model and the method of solving the current continuity equation. In Section 3, we present the statistical analysis of FACs. In Section 4, we use the average FAC patterns to derive conductances based on statistical analysis of simultaneous and coincident observations of FACs and conductances from Robinson et al. (2020). These average FAC conductance patterns are then used to determine the intensity and location of auroral electrojets. We discuss our results and the future outlook in Section 5. Finally, Section 6 presents our conclusions.

AMPERE Measurements
We use FAC measurements from AMPERE, based on the constellation of polar-orbiting commercial Iridium satellites in 2010-2017 (Anderson et al., 2000). The Iridium satellites are equipped with magnetometers, measuring magnetic perturbations caused by FACs with a 10-min cadence. The Iridium satellites are flying along six orbital tracks almost equally spaced in MLT (Anderson et al., 2000). Due to a large number (over 70) of Iridium satellites, there are multiple satellites flying over polar regions of both hemispheres at any one time. Due to precession of the orbital tracks, the MLT separation of the satellites varies between 1 and 3 h, which also sets the MLT resolution at which FACs can be estimated (Murphy et al., 2012). Instantaneous point measurements are converted into global 2D-maps of magnetic perturbations, using spherical harmonic fits with a resolution of 1° in latitude and 1 h in MLT. Finally, FACs are derived by using Ampere's law, that is, by calculating the curl of each magnetic perturbation map. The AMPERE data processing methodology is described in detail in Waters et al. (2001Waters et al. ( , 2020. For a review of AMPERE research, see Coxon et al. (2018).

Conductance Model
In addition to AMPERE data, we use an empirical conductance model (Robinson et al., 2020), based on the statistical relationship between height-integrated Hall and Pedersen conductances (Σ H and Σ P ) measured by the Poker Flat incoherent scatter radar and AMPERE-derived FACs. The statistical relationships between FAC and conductance employed in the model are MLT-dependent and also different for upward and downward currents, reflecting different physics of these regions. The response of conductance to a given FAC density is strongest in the midnight and dawn upward (R2) current regions and weakest in the dusk sector. To model the contribution of solar photoionization, we use empirical relations, parametrized by the F10.7 radio flux (Robinson et al., 1987).
Having estimates for both field-aligned (J ‖ ) currents and conductances allows us to estimate also the horizontal ionospheric electric field E ⊥ and (Hall and Pedersen) currents (J ⊥ ) by solving the current continuity equation is the height-integrated conductance tensor. (Note that the unit of J ‖ is A/m 2 while the unit of J ⊥ is A/m).
To solve Equation 2, we determined the Hall and Pedersen conductances from the FAC at each location using the statistical relations from Robinson et al. (2020). With E = -∇ϕ, Equation 2 becomes a second-order differential equation relating the electrostatic potential ϕ to J ‖ . We solve this equation by numerical integration, starting at a circle of zero potential assumed to be at 40° magnetic latitude. Smoothing of the potential is required to ensure the solution at each magnetic local time converges to the same value at the pole. At each 1° latitude step in the integration, the potential is smoothed over 3° in latitude and 3 h of magnetic local time. The integration extends to a latitude circle of 72°, above which FACs are small and there are no significant gradients in conductance. Above that latitude, we calculate the potential by interpolating values between diametrically opposite points along the 72° latitude circle. The smoothing process used to constrain the solution tends to dampen the potential values. To compensate for this dampening, we multiply the resulting potential by a sequence of amplification factors. After each factor is applied, we use the resulting potential distribution to calculate the electric fields, horizontal currents, and FACs. We select the amplification factor that minimizes the root mean square deviation between the input and output FAC distribution weighted by the absolute magnitude of the FAC density. This optimization procedure ensures that the derived electrojet intensities and locations are consistent with the statistically derived FACs from which they were derived. Plots comparing the input and output FACs (and modeled electrostatic potentials) for different seasons and signs of IMF B y are included in Figures S1-S8. The methodology used here is described in detail in Robinson et al. (2021).

Northern Hemisphere Winter
In this section, we study the possible explicit IMF B y -dependence of FACs during the two solstices (NH winter and NH summer  Iijima and Potemra (1976) and more recent statistical studies based on AMPERE data (Anderson et al., 2008D. L. Green et al., 2009 Figures 2c and 2d show that the B y -effect in the R1 and R2 FACs is in the same direction, but much stronger for more intense B y , especially in the dawn sector.
The amplitudes of the dawn R1 and R2 currents are clearly greater for B y > 5 nT and B y > 0. However, the difference in the amplitudes of the dawn R1 and R2 currents for B y < − 5 nT and B y < 0 is small and not statistically significant. Thus, only strongly positive IMF B y seems to significantly affect the dawn FACs.
In the dawn sector, the R2 current density is strongly correlated with the energy flux of auroral particle precipitation, dominated by downward precipitating electrons (Robinson et al., 2018). Thus, the B y -effect in the (upward) dawn R2 current strongly supports the results of Holappa et al. (2020) who found a similar explicit B y -effect in the flux of (downward) precipitating energetic electrons in the dawn sector. Interestingly, the dusk R2 current does not show statistically significant explicit B y -dependence (see Figures 2b and 2d). In the dusk sector, the downward R2 current is probably carried by both downward precipitating protons and upward moving secondary electrons (Robinson et al., 2018). Thus, Figure 2 implies that the explicit B y -dependence of precipitating electrons is stronger than that of protons.

Northern Hemisphere Summer
Figures 1c and 1d show the average NH FAC patterns in NH summer (NH summer solstice ± 30 days) for B y < 0 and B y > 0. The overall amplitude of FACs is greater in summer than winter, due to greater conductance in summer than winter (see, e.g., Carter et al., 2016; A. E. S. Green et al., 1964;Moen & Brekke, 1993;Ridley, 2007). Figures 1c and 1d also show the well-known modulation of the dayside R1 currents: The upward currents are located at higher latitudes for B y > 0 than for B y < 0, in agreement with earlier statistical results from the AMPERE data set (Carter et al., 2016;D. L. Green et al., 2009). This dependency arises from a dawn-dusk asymmetry in the magnetospheric convection controlled by IMF B y (Tenfjord et al., 2015). However, in this study, we are primarily interested in the explicit B y -dependence of the dawn and dusk sector FACs, closely connected to auroral electrojets.
The amplitudes of the dawn and dusk R2 currents in NH summer are not as strongly affected by B y as in NH winter. This is quantified in Figures 3a-3d, which show NH FAC densities for NH summer at 6 and 18 MLT with their standard errors, respectively. Figure 3a shows that the R2 current in the dawn sector is only slightly stronger for B y < 0 than for B y > 0 in the latitude range 67°-69°, but this difference is not statistically significant. This is also seen for large values of B y in Figure 3c. Figures 3b and 3d show that the B y -dependence is even weaker in the dusk R2 currents.

Southern Hemisphere
Figures 4a-4d are similar to Figures 1a-1d, but show the average FAC patterns for the Southern Hemisphere (SH). In SH winter (Figures 4a and 4b), the dawn R2 current is clearly stronger for B y < 0 than for B y > 0, while there are hardly any B y -effects in the dusk sector FACs. In SH summer (Figures 4c and 4d), the dawn and dusk R2 currents are only weakly modulated by IMF B y . These results show that the explicit B y -dependence of R2 currents is stronger in the dawn than the dusk sector in both hemispheres, and that this dependency is clearly seen only in the winter hemisphere.

Conductance
Using the average FAC maps (Figures 1 and 4) as an input, we derive estimates for ionospheric conductance using the Robinson et al. (2020) conductance model. Figures 5a-5d show the modeled height-integrated Hall conductances (Σ H ) for the NH. The conductances include contributions from solar illumination, parametrized by the solar radio flux intensity with F10.7 = 70 solar flux units, representing solar minimum conditions. The patterns represent illumination conditions in NH winter and summer solstices (December 22 and June 22) at 12 UT. Outside the auroral oval, where FACs are below 0.2 μA/m 2 , Hall and Pedersen's conductances are set to 2 S, which is a typical value for zenith angles over 90° (Ridley, 2007).
Figures 5a and 5b show that in NH winter Σ H is considerably higher in the postmidnight and dawn sectors for B y > 0 than for B y < 0. This is due to the strong B y -dependence of the dawn R2 currents (see R1 current in the premidnight sector for B y < 0. However, when averaged over the whole dusk sector (12-24 MLT), Σ H is only 4% higher for B y < 0 than for B y > 0. This is explained by the relatively weak B y -dependence of the dusk R2 currents. Also the coupling between R2 currents and conductance is weaker at dusk than at dawn in the Robinson et al. (2020) model (a modulation of R2 current density at dusk leads to a weaker modulation of conductance at dusk than at dawn).
During NH summer (see Figures 5c and 5d), the overall level of conductance is much higher than in NH winter, due to the large contribution by solar illumination. This also leads to a relatively weak B y -dependence of conductance. However, the B y -dependence of the dawn R1 current leads to a modulation of Σ H visible in Figures 5c and 5d. The B y -dependence would be relatively even weaker for greater (solar maximum) values of F10.7. Figures 6a-6d show the Hall conductance patterns for the SH. Similarly, as for NH, there is a strong B y -dependency in conductance during SH winter (Figures 6a and 6b), when the average auroral region Hall conductance in the dawn sector is 41% higher for B y < 0 than for B y > 0. Again, the B y -effect is relatively much weaker in SH summer (Figures 6a and 6b). Thus, the conductance model shows highly similar B y -effects (with expected dependencies on B y sign) in both hemispheres, giving strong evidence that the results are robust.

Auroral Electrojets
By using the FAC and conductance patterns presented in earlier sections, we model the horizontal ionospheric currents by solving the current continuity Equation 2. We concentrate on the East-West horizontal currents, which are dominant in the auroral electrojets. Figures 7a and 7b show the modeled East-West component of the ionospheric current in the NH in NH winter for B y < 0 and B y > 0, respectively. These HOLAPPA ET AL.     figures show that the westward electrojet (blue shaded area) is more extensive in latitude and stronger at dawn for B y > 0 than for B y < 0. The westward electrojet is most dramatically affected in the postmidnight sector (about 0-6 MLT), due to the strong B y -dependence of the conductance in this sector (see Figures 5a  and 5b). The peak westward current density in the dawn sector from 60° to 75° is 66% stronger for B y > 0 than for B y < 0. In NH winter, the eastward electrojet is only weakly modulated by IMF B y , the peak current density of the dusk sector being only 11% stronger for B y > 0. These results are in excellent agreement with Holappa and Mursula (2018), who found a strong explicit B y -dependency in the AL-index and a weak explicit B y -dependency in the AU-index in NH winter.
In NH summer (Figures 7c and 7d), the westward and eastward electrojets are stronger than in NH winter due to greater conductance. In NH summer, the explicit B y -dependence in latitudinal extent and amplitude of the westward electrojet is much weaker than in NH winter. However, the peak westward current density in the dawn sector is 29% stronger for B y < 0 than for B y > 0 (the opposite sign dependence as in NH winter). This is caused by the B y -dependence of the dawn R2 current, which is slightly stronger for B y < 0 than for B y > 0 (see Figure 3a). Thus, the explicit B y -dependence of the westward electrojet in NH summer is weaker than in NH winter, which is in excellent agreement with Holappa and Mursula (2018), who found that the AL-index is slightly stronger for B y < 0 than for B y > 0 in NH summer. Figures 7c and 7d show that the eastward electrojet is clearly more extensive in MLT for B y < 0 than for B y > 0 in NH summer. The eastward current is stronger for B y < 0 especially around noon. However, the peak eastward current density in the dusk sector is only 11% stronger for B y < 0 than for B y > 0. This is consistent with the weak B y -dependence of the AU-index (Holappa & Mursula, 2018), because the AU-index measures, by definition, the maximum magnetic disturbance induced by the eastward ionospheric currents.
Figures 8a-8d show the modeled East-West component of the ionospheric current in the SH. Similarly, as in NH, the explicit B y -dependence of the westward electrojet (blue shaded area) in the SH is much stronger in local winter than in local summer, in agreement with the earlier results based on the K-index of the Antarctic Syowa station (Holappa & Mursula, 2018). (Note the opposite dependence on the B y sign in SH winter than in NH winter.) However, the peak westward current in the dawn sector is about 109% stronger for B y < 0 than for B y > 0 in SH winter. Also, similarly as in NH, the eastward electrojet is more extensive in HOLAPPA ET AL. MLT for B y > 0, but the peak eastward current is only 2% stronger for B y > 0, which is within the statistical error. In SH summer, the westward electrojet is about 11% stronger for B y < 0 than for B y > 0 (opposite B y sign dependence as expected). However, this is not statistically significant because the B y -dependence in the SH dawn R2 currents is not statistically significant: The peak R2 current at 6 MLT is only about one standard error (0.3 μA/m 2 ) greater for B y > 0 than for B y < 0.

Discussion
Using AMPERE measurements, we have shown in this study that IMF B y strongly modulates the amplitude and spatial distribution of FACs and auroral electrojets. In this study, we have focused on the large-scale R1 and R2 currents in the dawn and dusk sectors, and how FACs are connected to auroral electrojets in these sectors. We showed (see Figure 1) that in NH winter, the R1 and R2 FACs are stronger for B y > 0 than for B y < 0 in the dawn sector for the same level of solar wind driving (the Newell coupling function dΦ MP /dt). Thus, FACs exhibit explicit B y -dependence, which is not explained, for example, by differences in solar wind driving for different signs of B y . These results are consistent with earlier statistical studies of FACs, based on Dynamics Explorer 2 (Weimer, 2001) and Iridium measurements (D. L. Green et al., 2009) which have also found stronger dawn R2 currents for B y > 0. We found a similar explicit B y -effect in the dawn R1 and R2 currents also in the SH in SH winter, but with a reversed dependence on the B y sign. While the B y -dependence of the dawn R1/R2 currents is strong in both hemispheres in local winter, there is only negligible B y -dependence in the dusk R1/R2 currents. Many studies have shown that the flux of precipitating electrons is strongly correlated with upward FACs, especially in the dawn sector and close to midnight (Juusola et al., 2016;Knight, 1973;Lyons, 1980;Robinson et al., 2018). Thus, the strong B y -dependence of the dawn R2 currents is in excellent agreement with Holappa et al. (2020), who found a similar, strong B y -dependence in the flux of energetic precipitating electrons in the midnight-dawn sector.
We have also used an empirical conductance model (Robinson et al., 2020), based on statistical relationships between FACs and Hall and Pedersen conductances. We showed that ionospheric conductance is strongly modulated by IMF B y in the dawn sector, especially below the regions of upward R2 currents. This reflects the above-mentioned B y -dependence of the precipitating electron flux in the dawn sector (Holappa et al., 2020). The modeled height-integrated Hall conductance averaged over the dawn sector of NH auroral zone is about 50% greater for B y > 0 than for B y < 0 in NH winter, when the solar wind coupling function is 2-3 times greater than its average value (2 < dΦ MP /dt/〈dΦ MP /dt〉 < 3). We found an equally strong B y -effect in the SH auroral zone conductance in SH winter, but with an opposite dependence on the B y sign, as expected from the observed B y -dependency of FACs. We note that the above range of the solar wind coupling function corresponds to southward B z of about −3.2 nT, on average. Thus our results are valid only for periods of southward B z .
Using the modeled ionospheric conductances and observed FACs, we also estimated horizontal ionospheric electric fields and currents. We showed that the peak current density of the NH westward electrojet is about 66% stronger for B y > 0 than for B y < 0 in NH winter, due to the strong B y -dependence of R2 currents and the ionospheric conductance in the dawn sector. In NH summer the peak current density of the westward electrojet is 29% stronger for B y < 0 than for B y < 0, because the dawn R2 currents are slightly stronger for B y < 0. We also showed that the B y -dependence in the peak current density of the eastward electrojet is weak (about 11%) in both NH winter and summer.
Even though it is well known that IMF B y modulates FACs (Anderson et al., 2008;D. L. Green et al., 2009;Iijima et al., 1978;Milan et al., 2015;Weimer, 2001) and ionospheric electric fields (Heppner & Maynard, 1987;Ruohoniemi & Greenwald, 1996;Thomas & Shepherd, 2018;Weimer, 1995), relatively few studies on the role of IMF B y in modulating the auroral electrojets have been made. Earlier evidence for the strong explicit B y -dependence of the westward electrojet in the winter hemisphere is based on geomagnetic indices (Friis-Christensen et al., 1972;Holappa & Mursula, 2018;Holappa et al., 2019;Murayama et al., 1980) and equivalent currents derived from ground-based magnetometers (Laundal et al., 2016). Recent studies (Friis-Christensen et al., 2017;Smith et al., 2017), based on magnetic perturbations measured by polar-orbiting satellites, provide additional independent support. While the existence of the explicit B y -effect in the westward electrojet is now well established, previous studies have not shown how IMF B y modulates different parts of the global magnetospheric-ionospheric electric circuit. The results of this paper are in excellent agreement with the above studies based on ground-based observations. As a novel contribution, we have shown how these hemispheric and MLT-asymmetries in auroral electrojets arise from the B y -dependence of FACs. In particular, we have shown that the B y -dependence of FACs also explains the summer-winter difference in the explicit B y -dependence of the westward electrojet and the weak explicit B y -dependence of the eastward electrojet.
In this study, we have demonstrated the usefulness of the novel methodology of deriving global distribution of ionospheric electrodynamical parameters using AMPERE measurements of FACs (Robinson et al., 2020). A similar approach has been successfully applied recently by Mukhopadhyay et al. (2020) to derive ionospheric currents from MHD-modeled FACs. Future validation of these models will undoubtedly improve their accuracy in simulating the strength of the electrojet currents, but the relative intensities should not differ significantly from the results shown here. These are primarily determined by the FAC densities in the oppositely directed current sheet pairs in the dawn and dusk sectors. Since our conclusions are based on the relative strengths of the auroral electrojets, they are not sensitive to the absolute accuracy of the auroral electrodynamic model.
Even though we only investigated statistical patterns of ionospheric parameters here, the same methodology can be applied in the future for studying individual events. Event-based studies are needed for better understanding of the chain of physical processes occurring, for example, during different substorm phases. This paper also further demonstrates that the B y -effect is important for space weather predictions. Our results show that as none of the current solar wind coupling functions take the explicit B y -dependence into account, they over/underestimate geomagnetic activity, depending on season and IMF B y . Follow-on investigation is ongoing to study if different space weather prediction assets, for example, global MHD models, are able to reproduce the By-dependence of FACs and auroral electrojets quantified here.

Conclusions
In this study, we showed that for a fixed level of solar wind driving, R1 and R2 FACs are stronger for B y > 0 than for B y < 0 in NH winter. The B y -dependence is reversed in SH winter. Using a conductance model (Robinson et al., 2020), based on empirical relations between FAC and ionospheric conductance, we showed that the B y -dependence of FAC leads to a modulation of ionospheric conductance in the dawn sector. This modulation of conductance is relatively much stronger in the winter hemisphere than in the summer hemisphere. By solving a current continuity equation connecting FAC, conductance and auroral electrojets, we showed that the NH westward electrojet is about 66% stronger for B y > 0 than for B y < 0 in NH winter. These results emphasize the importance of the explicit B y -effect in space weather prediction.