Quasi‐biennial oscillation of the ionospheric wind dynamo

The interannual variation of the ionospheric solar quiet (Sq) current system is examined. A dense magnetometer network over Japan enables the accurate determination of the central position of the northern Sq current loop or the Sq current focus, during 1999–2015. It is found that the Sq focus latitude undergoes an interannual variation of ±2° with a period of approximately 28 months, similar to the quasi‐biennial oscillation (QBO) in the tropical lower stratosphere. The QBO‐like variation of Sq is particularly evident during 2005–2013. No corresponding interannual variability is found in solar extreme ultraviolet radiation. Comparisons with tidal winds, derived from a whole‐atmosphere model, reveal that the QBO‐like variation of the Sq current focus is highly correlated with the amplitude variations of migrating and nonmigrating diurnal tides in the lower thermosphere. The results suggest that the stratospheric QBO can influence the ionospheric wind dynamo through the QBO modulation of tides.


Introduction
Solar quiet (S q ) daily variations of the geomagnetic field are primarily due to electric currents flowing in the dynamo region of the ionosphere (95-150 km) (see a review by Yamazaki and Maute [2016]). In the dynamo region, the neutral wind U moves the electrically conducting ionosphere across Earth's main magnetic field B, which produces an electromotive force U × B. The associated current density J can be expressed as wherêis the ionospheric conductivity tensor and E is the electric field. The neutral wind at dynamo region heights is dominated by atmospheric tides. The dynamo action by those tides leads to the formation of a global-scale ionospheric current system, which is often referred to as S q current system. A typical pattern of the dayside S q current system is illustrated in Figure 1a. The S q current system is normally composed of a counterclockwise vortex in the Northern Hemisphere and a clockwise vortex in the Southern Hemisphere. The S q current system effectively disappears during nighttime because of low ionospheric conductivities.
The strength and shape of the S q current system change on various timescales. The day-to-day and hour-to-hour variations are mostly due to the variability of atmospheric tides and other waves that propagate into the dynamo region from the lower layers of the atmosphere [Kawano-Sasaki and Miyahara, 2008;. An extreme example of the meteorological impact on the S q current system can be found during major stratospheric sudden warming events [Yamazaki et al., 2012a[Yamazaki et al., , 2012b. The S q current system also shows seasonal variability [Takeda, 2002;Chulliat et al., 2016], which is due to the effects of both ionospheric conductivity and neutral wind. On longer time scales, the solar cycle effect dominates the variability of the S q current intensity. The S q current intensity during solar maximum is higher than during solar minimum by a factor of 2 or so owing to enhanced ionospheric conductivities [Takeda, 1999[Takeda, , 2013.
The present study focuses on the interannual variation of the S q current system. Recent numerical studies showed that the interannual variation of atmospheric tides in the lower thermosphere could be affected by the quasi-biennial oscillation (QBO) [Liu, 2014;Gan et al., 2014;Miyoshi et al., 2017] and the El Niño-Southern Oscillation (ENSO) Liu, 2012, 2013]. The question remains whether the QBO and ENSO have any measurable impact on the ionosphere. This study aims to find out the importance of these meteorological sources in producing interannual variability in the ionospheric electrodynamics. We examine the S q current system, which is a direct consequence of the ionospheric wind dynamo in the lower thermosphere.  The year-to-year variation of the S q current intensity is primarily controlled by solar activity, which makes it difficult to detect small changes caused by atmospheric tides. We instead examine the latitudinal position of the S q current focus. By "S q current focus," we mean the center of the S q current loop (see Figure 1a). The accurate determination of the S q current focus is important in this study, which will be achieved by using a dense magnetometer network over Japan. The latitudinal position of the S q current focus is not sensitive to solar activity [Yamazaki et al., 2011], and its variability is not well understood.

Geomagnetic Data
Ground-based magnetometer data are obtained from 14 Japanese observatories; three stations are operated by the Japan Meteorological Agency and 11 stations by the Geospatial Information Authority of Japan. Figure 1b shows the location of the observatories. We first use the horizontal intensity (H) and the declination angle (D) of the geomagnetic field. The H component geomagnetic disturbances associated with the magnetospheric ring current are corrected by subtracting the Dst index multiplied by cos m , where m is the magnetic latitude. The corrected H field is denoted as H c . The northward (X) and eastward (Y) components of the geomagnetic field are then derived from H c and D. The magnetic perturbations due to the S q current system can be derived by subtracting the nighttime baseline, under the assumption that S q currents are negligible during nighttime due to low ionospheric conductivities. The magnetic perturbations in X and Y are designated as ΔX and ΔY, respectively, which will be used to determine the latitudinal position of the Northern Hemisphere S q current focus.
For the determination of the S q focus position, we basically follow the technique recommended by Stening et al. [2005]. This technique requires ΔX and ΔY data from a north-south chain of magnetometers at midlatitudes where the S q current focus usually appears. It relies on the fact that both ΔX and ΔY become zero under the focus of the S q current system. The application of the technique involves the following two steps: (1) determine the time when ΔY crosses the zero level and (2) plot ΔX at that time as a function of latitude to find the latitude where ΔX is zero. We determine the S q focus latitude on the monthly basis. We first calculate the average daily variations ΔX and ΔY for each month using the ΔX and ΔY data corresponding to the 10 quietest days of the month. We then apply the technique described above to ΔX and ΔY. The 10 quietest days are routinely selected and published by GFZ German Research Centre for Geosciences.  indicating that the S q current focus is located within the latitudinal range of the Japanese magnetometer array. As can be seen in Figure 2c, the ΔX values corresponding to ΔY = 0 smoothly changes with latitudes, from positive values at lower latitudes to negative values at higher latitudes. The latitude where ΔX = 0 gives the S q focus latitude. We used the polynomial function of degree n = 3 for the latitudinal interpolation of the ΔX data. The 1 error in the S q focus latitude was estimated by propagating uncertainty in the nighttime base line of X, though the fitting process for determining the latitude of ΔX = 0. The S q focus latitude was derived for each month from January 1999 to December 2015.

Ground-to-Topside Model of Atmosphere and Ionosphere for Aeronomy
We examine the interannual variability of tides in the dynamo region using the Ground-to-topside model of Atmosphere and Ionosphere for Aeronomy (GAIA). GAIA is a coupled atmosphere-ionosphere model extending from the ground to the exobase [e.g., Jin et al., 2011;Miyoshi et al., 2012;. The model consists of physical equations appropriate for various atmospheric processes in the troposphere, stratosphere, mesosphere, and thermosphere under the assumption of hydrostatic equilibrium. The horizontal resolution of the model is 2.8 ∘ in longitude and latitude, and the vertical resolution is 0.2 scale height.

Miyoshi et al.
[2017] performed a long-term GAIA simulation for the years 1997-2013. We use the same run, but the simulation was extended until March 2016. Following Jin et al. [2012], the lower part of the model, below 30 km, was constrained on the basis of a nudging technique using the Japanese 25 year Meteorological Reanalysis [Onogi et al., 2007]. This acts as external forcing that drives the QBO and ENSO in the model, along with other short-term and long-term atmospheric variability. The model also takes into account the variable energetic solar radiation. The F 10.7 solar activity index was used as a proxy of the solar EUV/UV, which is the primary heat source of the upper atmosphere. The model was run under geomagnetically quiet conditions for the entire duration of the simulation.
Neutral temperature, zonal and meridional winds were output for the altitude range of 100-150 km, corresponding to the dynamo region. Following Forbes et al. [2008], a tide was defined in the following form: where A n,s and n,s are the amplitude and phase, t is the time, Ω is the rotation rate of the Earth, is the longitude. n is the subharmonics of a day. The parameter n = 1, 2, 3 corresponds to oscillations with periods of 24 h, 12 h, and 8 h and are referred to as diurnal, semidiurnal, and terdiurnal tides, respectively. The variable s is the zonal wave number, indicating eastward propagating waves when s < 0 and westward propagating waves when s > 0. The Fourier decomposition technique  enables to determine the amplitude and phase of tides with different combinations of n and s. We examine the amplitudes of the migrating diurnal tide (n = 1, s = 1), nonmigrating diurnal tide with zonal wave number 3 (n = 1, s = −3), and migrating semidiurnal tide (n = 2, s = 2). In the rest of the paper, these tides are referred to as DW1, DE3, and SW2, respectively. DW1, DE3, and SW2 are known to have particularly large amplitudes in the dynamo region [e.g., Oberheide et al., 2011] thus have a potential to influence the S q current system.
For the validation of the tides simulated by GAIA, DW1, DE3, and SW2 in the temperature field at 100 km altitude are compared with those derived from the Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) instrument [Remsberg et al., 2008] on board the Thermosphere Ionosphere Mesosphere Energetics Dynamics (TIMED) satellite. The model-data comparison will be presented in section 3.2.
Although GAIA solves for electric fields and currents in the ionosphere, the model does not calculate the magnetic perturbations associated with the ionospheric currents, which are necessary for the determination of the S q focus position. Thus, we do not conduct model-data comparisons for S q . The purpose of using GAIA is to derive the interannual variability of tidal winds in the dynamo region, which we will compare with the observed S q variability. Figure 3a shows monthly values of the S q focus latitude over Japan from 1999 through 2015. The average latitude is 30.7 ∘ N, in agreement with previous studies [e.g., Stening et al., 2007]. The variations in the S q focus latitude are much greater than the estimated 1 error. The S q focus latitude occasionally exhibits a large northward displacement beyond 40 ∘ N. Such events occurred in February of 2006, 2008, and 2013. As will be seen later, these variations are in part due to the seasonal cycle superposed on the effect of QBO. The average seasonal variation of the S q focus latitude during 1999-2015 is presented in Figure 3b. The results show a rapid northward motion of the S q current focus from January to February. The S q current focus latitude is lowest during September, and it shows a secondary peak in November. These seasonal characteristics are largely consistent with those presented by Vichare et al. [2016] for the Indo-Russian region. The driving mechanism for the seasonal variation of the S q focus latitude is not well understood. The ionospheric conductivity at middle latitudes is generally highest during local summer and lowest during local winter, which does not explain a complex seasonal pattern of the S q focus latitude. Takeda [1990] and Kawano-Sasaki and Miyahara [2008] numerically showed that changes in the thermospheric winds can affect the latitudinal position of the S q current focus.

S q Focus Latitude
The anomaly in the S q focus latitude was calculated by subtracting the average seasonal variation (Figure 3b) from the original monthly data (Figure 3a). In Figure 4a, the black line shows monthly values of the S q focus latitude anomaly, revealing fluctuations on a timescale of a few months. The blue and red lines show the smoothed values calculated by applying 7 month and 13 month moving windows, respectively. The two results are in good agreement, indicating that the results are not very sensitive to the choice of the smoothing window. It can be clearly seen that the S q focus latitude oscillates by approximately ±2 ∘ on interannual timescales. The interannual variation is most evident during 2005-2013, which roughly corresponds to lowsolar flux periods. Figure 4b shows the monthly mean zonal wind measured at Singapore (1.2 ∘ N, 103.6 ∘ E), which represents the stratospheric QBO. The wind data, extended from Naujokat [1986], are provided by Freie Universität Berlin (FUB). The observations cover the region from 70 hPa (∼18 km) to 10 hPa (∼31 km), where the QBO is most prominent. It can be seen that the interannual variation of the S q focus latitude correlates with the phase of the stratospheric QBO. The S q focus latitude tends to be lower and higher during the easterly and westerly phases of the stratospheric QBO, respectively. The NINO.3 index, which represents ENSO activity, also shows significant interannual variability (Figure 4c), but the interannual variation of the NINO.3 index is not coherent with the interannual variation of the S q focus latitude. As discussed by Liu [2016], the stratospheric QBO has a very regular oscillation cycle around 28 months, while ENSO variability consists of longer-period oscillations (∼43 and ∼62 months). A spectrum analysis of the monthly values of the S q focus latitude anomaly revealed a peak period of ∼28 months. Figure 4d displays the EUV measurements (0.1-50 nm) by the Solar EUV Monitor (SEM) spectrometer [Judge et al., 1998] on the Solar Heliospheric Observatory (SOHO). The interannual variation of the EUV flux is dominated by the 11 year solar cycle. It is interesting to note that the period when the interannual variation of S q focus latitude was prominent (e.g., 2005-2013) roughly corresponds to the period of low EUV flux when the year-to-year change in the EUV flux is particularly small.
The interannual variation in the geomagnetic activity index Ap is shown in Figure 4e. It is noted that the overall geomagnetic activity level is low because our analysis is limited to geomagnetically quiet days. Geomagnetic activity peaked in 2003 during the declining phase of solar cycle. However, there is no corresponding variation in the S q focus latitude. Similar to the EUV data, the interannual variation is small during the solar minimum, when the interannual variation of the S q focus latitude is large.

Tides in the Lower Thermosphere
As we showed in the previous section, the focus position of the S q current system shows a periodic oscillation similar to the stratospheric QBO. In this section we investigate the interannual variation of atmospheric tides in the lower thermosphere, where S q currents are driven through the ionospheric wind dynamo mechanism. Our focus is on these tidal components: DW1, DE3, and SW2, which are known to have large amplitudes at dynamo region heights [e.g., Oberheide et al., 2011].

TIMED/SABER-GAIA Comparisons
We first present comparisons between the temperature tides derived from TIMED/SABER data and GAIA simulation. Figures 5a and 5b compare the average seasonal variations in the amplitude of the migrating diurnal tide DW1 at 100 km derived from TIMED/SABER and GAIA, respectively. The model-data agreement is very good. It is known from previous studies [e.g., Burrage et al., 1995;Forbes et al., 2008] that the DW1 amplitude in the mesosphere and lower thermosphere is subject to a semiannual modulation with equinoctial maxima.
Conducting numerical experiments, McLandress [2002a] demonstrated that the latitudinal shear in the zonal mean wind plays a role in producing seasonal variability of the migrating diurnal tide.  Figure S1 in the supporting information for the seasonal climatology of DW1.) The interannual variation of the DW1 amplitude is presented in Figures 5c and 5d for TIMED/SABER and GAIA, respectively. The anomaly was computed in the same way as for the S q focus latitude. That is, we first subtracted the average seasonal variations from the original data and then applied the 13 month running average to the residual data. The results clearly show that the interannual variation of DW1 is dominated by a QBO-like oscillation. The QBO modulation of the migrating diurnal tide in the mesosphere and lower thermosphere has been reported by earlier researchers [e.g., Hagan et al., 1999;Forbes et al., 2008;Wu et al., 2008;Mukhtarov et al., 2009;Xu et al., 2009]. McLandress [2002b] attributed the QBO modulation of DW1 to the change in the zonal circulation. Mayr and Mengel [2005] showed that the mechanism suggested by McLandress [2002b] is effective only below 50 km altitude, and the QBO modulation of DW1 above 80 km is mainly due to the momentum deposition from small-scale gravity waves.
The GAIA model reproduces the interannual variation of DW1, but the amplitude of the QBO oscillation is somewhat smaller compared to the TIMED/SABER observations. Figure 5e compares the stratospheric QBO at 10 hPa with the interannual variation of the DW1 amplitude. The results are presented for the average over 10 ∘ S-10 ∘ N where the interannual variation of DW1 is relatively large. It can be seen that the DW1 amplitude tends to be greater during the westerly phase of the stratospheric QBO. It is noted that the phase of the interannual variation of DW1 is shifted to later years during 2009-2014 with respect to the phase of the stratospheric QBO. The reason is unclear. Figure 6 compares the amplitudes of the eastward propagating nonmigrating diurnal tide with wave number 3, or DE3, at 100 km derived from TIMED/SABER and GAIA in the same format as Figure 5. The GAIA model reproduces main characteristics of seasonal and interannual variability of DE3. The QBO effect is evident in the amplitude anomaly (Figures 6c and 6d), consistent with previous reports [e.g., Oberheide et al., 2009;Häusler et al., 2013]. The QBO modulation of DE3 weakens toward the end of the period, which can be seen in the GAIA results as well as in the TIMED/SABER data. As shown in Figure 6e, the DE3 amplitude tends to be greater during the westerly phase of the stratospheric QBO, similar to the DW1 results.
As shown in Figure 7, the model-data agreement is not as good for the semidiurnal migrating tide SW2. The seasonal and latitudinal patterns of SW2 are only in rough agreement between the TIMED/SABER measurements and GAIA simulation (Figures 7a and 7b). Akmaev et al. [2008] encountered a similar problem when they compared SW2 from TIMED/SABER with the Whole Atmosphere Model (WAM). It was considered that the difference in data sampling between observations and simulations could be a part of the reason for the disagreement. The amplitude anomaly of SW2 shows a complex latitudinal pattern (Figures 7c and 7d). The QBO modulation of the SW2 amplitude is visible in the TIMED/SABER data (Figure 7e), which is partially reproduced by GAIA. The SW2 amplitude tends to be greater during the easterly phase of the stratospheric QBO, when the DW1 and DE3 amplitudes become small, which is consistent with previous studies [e.g., Forbes et al., 2008;Pancheva et al., 2009]. The mechanism for the opposite QBO responses in DW1 and SW2 is still to be understood.

QBO Modulation of Tidal Winds
Next, we examine the interannual variation of tidal winds in GAIA. The seasonal climatology was first determined for DW1, DE3, and SW2 in the zonal and meridional winds at 100-150 km (see Figures S1-S3 in the supporting information). Amplitude anomalies were then derived as the deviation of monthly tidal amplitudes from the seasonal climatology. Figures 8a and 8b show the amplitude anomaly in DW1 at 100 km for zonal and meridional winds, respectively. The QBO effect is evident, accounting for the amplitude anomaly of up to ±3 m/s in the zonal wind and ±5 m/s in the meridional wind. Given that the GAIA model underestimates the interannual variability of DW1 in temperature (Figure 5), the actual QBO effect on the tidal winds is likely to be greater. The QBO modulation of DW1 winds is mostly confined within ±40 ∘ latitudes. The peak modulation occurs at ±10-30 ∘ latitudes, indicating the dominance of the (1,1) Hough mode of classical tidal theory [Lindzen and Chapman, 1969]. The QBO modulation of DW1 can also be seen at 110 km (Figures 8c and 8d) but with smaller amplitudes. At higher altitudes (Figures 8e-8h), the solar cycle effect dominates the interannual variability of DW1 winds. It is known that DW1 in the dynamo region consists of the tide from the lower atmosphere and the tide locally excited by solar EUV/UV heating [Forbes, 1982;Hagan et al., 2001]. The strong solar cycle influence at high latitudes can be explained by the variability of DW1 locally generated in the thermosphere. Figure 9 presents the results for DE3 winds in a similar format as Figure 8. The QBO modulation of DE3 is evident in the zonal wind (±3 m/s) over the equator. The effect can be seen throughout the dynamo region. The vertical wavelength of DE3 is longer compared to DW1, which allows the wave to propagate to higher altitudes before being dissipated. Significant interannual variability can also be found in SW2 winds ( Figure 10). However, the QBO effect is not immediately obvious, indicating that contributions by other sources are also important for SW2. At 150 km, the solar cycle influence dominates the interannual variability of SW2 winds.

Comparison With S q Focus Latitude
We now examine the relationship between the interannual variability of the S q focus latitude and tides. In this section, we use a band-pass filter for periods between 20 and 40 months to extract the variations around the QBO periodicity (∼28 months), instead of the 13 month running mean filter used in the preceding sections. The band-pass filter substantially removes the signals associated with the ENSO (>40 months) and 11 year solar cycle. Figure 11a shows the band-pass-filtered anomaly in the S q focus latitude. As previously shown in a It is noted that the 20-40 month band-pass filter was applied to all the variables before calculating the correlation coefficients. Figure 4a, the S q focus latitude exhibits a QBO-like variation of ±2 ∘ , most notably during 2005-2013. We first compare the results with the stratospheric QBO. Table 1 gives the correlation coefficients for the interannual variability of the S q focus latitude over Japan and the mean zonal wind over Singapore. The band-pass filter was applied not only to the S q focus latitude but also to the mean zonal wind. Table 1 shows that the correlation coefficient depends on height, being positive at 10 hPa (∼31 km) and negative at 50 hPa (∼21 km). This is because the phase of the stratospheric QBO varies with height (see Figure 4b). The strongest correlation was obtained at 20 hPa (∼26 km) where the variations in the S q focus latitude and mean zonal wind are in phase. The correlation coefficient is as high as 0.93 when the analysis is limited to the period 2005-2013. Figure 11b shows the band-pass-filtered anomaly in the DW1 meridional wind amplitude at 18 ∘ N. Different colors correspond to different altitudes. The QBO influence is apparent at 100 and 110 km. These tidal variations are nearly in phase with the variation in the S q focus latitude, which is reflected in the high correlation coefficients: 0.91 at 100 km and 0.90 at 110 km during 2005-2013 (see Table 1). Figure 11c is the same as Figure 11b but for the DE3 zonal wind amplitude at 4 ∘ N. The QBO modulation of the DE3 wind is visible at all heights without any phase shift. A comparison with the S q focus latitude reveals high correlation coefficients throughout the dynamo region (Table 1). Figure 11d shows the band-pass-filtered anomaly in the SW2 meridional wind at 57 ∘ N, where the interannual variability of the tide is most pronounced (see Figure 10). The tidal variations are not well correlated with the S q focus latitude (Table 1) nor with the stratospheric QBO. Thus, the interannual variability of SW2 winds may be dominated by other sources than QBO.

Discussion
The speculation about the stratospheric QBO influence on the ionospheric wind dynamo has existed for many years without compelling evidence. Some studies found a weak geomagnetic variation at a period around 27 months [Stacey and Wescott, 1962;Yacob and Bhargava, 1968;Olsen, 1994;Jarvis, 1996Jarvis, , 1997, while other 10.1002/2016JA023684 studies did not find such a peak in the geomagnetic spectrum [London and Matsushita, 1963;Shapiro and Ward, 1964;Love and Rigler, 2014]. It has often been a matter of debate whether the quasi 2 year oscillation in the geomagnetic field is associated with the stratospheric QBO or the same period of oscillation in solar activity [e.g., Yacob and Bhargava, 1968;Sugiura and Poros, 1977]. In the latter case, the geomagnetic variation arises from changes in ionospheric conductivities rather than neutral winds. We showed that the QBO-like variation in the S q current system is evident during the solar minimum period when interannual variability of solar activity is small. Besides, the latitudinal position of the S q current focus is not sensitive to solar activity (see Figure 3). Based on these observations, we can rule out the possibility of the dominant solar contribution to the interannual variation of the S q focus latitude.
The S q current system can be regarded as a superposition of the current systems driven by different tides. Since different tides drive different patterns of the ionospheric current system, changes in the tidal composition would affect the shape and intensity of the S q current system [e.g., Richmond et al., 1976;Stening, 1989;Yamazaki et al., 2012b]. Using the GAIA model as well as TIMED/SABER measurements, we showed that the atmospheric tides DW1, DE3, and SW2 in the dynamo region are significantly influenced by the stratospheric QBO, supplementing previous observations and numerical results [e.g., Forbes et al., 2008;Liu, 2014]. We made direct comparisons between the interannual variations in the tidal wind amplitudes and the S q focus latitude, finding that the QBO-like variation of the S q current focus is highly correlated with the interannual variations in the diurnal tidal amplitudes (i.e., DW1 and DE3) in the dynamo region. These results suggest that the quasi 2 year variation of the S q current system is likely due to tidal variability associated with the stratospheric QBO.
It is beyond the scope of the present study to determine the relative contribution of different tides (DW1, DE3, SW2, and other tides) to the QBO modulation of S q . Further numerical experiments would be necessary to clarify which tide plays a dominant role in the QBO modulation of the ionospheric wind dynamo and how exactly the tide affects the latitudinal position of the S q current focus. Although the SW2 wind amplitude in GAIA did not clearly show the QBO influence, the possible contribution of SW2 cannot be excluded because of the limited ability of GAIA in reproducing the interannual variability of SW2 (see Figure 7).
More efforts are required to establish the morphology of the QBO effect on the ionospheric dynamo. Observations in different longitudes could provide insights into the role of nonmigrating tides. Also, it needs to be clarified whether the QBO effect on the S q focus latitude can be observed in the Southern Hemisphere.
Our results showed no obvious correlation between the interannual variations of the ENSO activity index and S q focus latitude ( Figure 4). However, it is possible that the ENSO activity affects the S q current system indirectly by modulating the stratospheric QBO. Studies have shown that the amplitude and phase of the stratospheric QBO depend on ENSO activity [Taguchi, 2010;Yuan et al., 2014;Geller et al., 2016]. The possible ENSO effect on the ionospheric wind dynamo should be further investigated.
The interannual variation of the S q focus latitude over Japan was most evident during 2005-2013, when the solar EUV flux was low. It is possible that the QBO modulation of the ionospheric dynamo is solar cycle dependent. A longer data set would be necessary to clarify the impact of solar activity. An important piece of information obtained from the GAIA simulation is that the QBO modulation of tidal winds occurred in the dynamo region throughout the period examined, regardless of solar activity. Thus, the apparent absence of the QBO signal during 1999-2004 is not due to the absence of the QBO variation in tides, but due to other mechanisms that make the QBO modulation of the S q current system undetectable. The numerical study by Liu and Richmond [2013] showed that the meteorological contribution to ionospheric variability is more significant in solar minimum conditions than in solar maximum conditions. During solar maximum, the ionospheric dynamo at F region heights (above 150 km) becomes important, thus the contribution by the E region dynamo, which is more responsive to meteorological forcing, is relatively small. More discussion on the role of the F region dynamo in the S q current system and its solar activity dependence can be found in Maute and Richmond [2016].
A natural question that arises from the present study is whether the QBO modulation of the ionospheric wind dynamo has a broader impact on the ionosphere. A number of studies have already reported on the quasi 2 year variation in the ionospheric plasma density [Chen, 1992;Kane, 1995;Echer, 2007;Tang et al., 2014;Zhou et al., 2016;Chang et al., 2016], but the association with the stratospheric QBO is yet to be established. Yamazaki and Richmond [2013] numerically showed that there are two mechanisms by which upward propagating tides in the lower thermosphere can affect the ionosphere. One is through the electrodynamic effect. That is, the electric field generated by the dynamo action of tides will modulate the plasma transport perpendicular to the geomagnetic field, which is dominated by the so-called E × B drift. The other mechanism is tidal mixing. The dissipation of tidal waves alters the mean circulation of the thermosphere, which in turn modulates the thermospheric composition that determines the production and loss rates of the ionospheric plasma (see also Jones et al. [2014aJones et al. [ , 2014b for detailed discussions on the tidal mixing mechanism). Chang et al. [2016] showed observational evidence that tidal mixing, along with the direct solar effect, is in play in the ionospheric QBO. More numerical work is required to determine the relative importance of different mechanisms for the ionospheric QBO.

Conclusions
The main results of the present study may be summarized as follows: 1. The latitude of the S q current focus, estimated using a dense magnetometer network over Japan for 1999-2015, shows an interannual variation of ±2 ∘ . 2. A quasi 2 year variation is found in the S q focus latitude during 2005-2013. The S q focus latitude tends to be higher and lower during the westerly and easterly phases of the stratospheric QBO, respectively. 3. No corresponding interannual variation is found in the ENSO activity index NINO.3, solar EUV flux, or geomagnetic activity index Ap. 4. The QBO-like variation of the S q focus latitude is highly correlated with the amplitude variations of DW1 and DE3 tidal winds in the dynamo region.
These results suggest that the variation of atmospheric tides due to the stratospheric QBO could be an important source for interannual variability of the ionospheric wind dynamo.