Geomagnetic lunar and solar daily variations during the last 100 years

This paper describes long‐term changes in the geomagnetic lunar (L) and solar (S) daily variations. We analyze the eastward component of the geomagnetic field observed at eight midlatitude stations during 1903–2012. The amplitude and phase for the semidiurnal component of the L and S variations are examined. Both L and S amplitudes correlate with the solar activity index F10.7, revealing a prominent 11 year solar cycle. In both cases, the correlation is slightly better with F10.7 than F10.7. The sensitivity of the L variation to solar activity is comparable with that of the S variation. The solar cycle effect is also found in the phase of the S variation but not apparent in the phase of the L variation. The ratio in the amplitude of the L to S variation shows a long‐term decrease (approximately 10% per century), which may be due to a reduction in lunar tidal waves from the lower atmosphere to the upper atmosphere in association with climate change.


10.1002/2014JA020203
In order to determine long-term Sq changes other than those caused by solar activity, some authors attempted to "remove" the effect of solar activity by fitting solar activity indices, such as sunspot number, and subtract the fit from the original data. Sellek [1980] and Schlapp et al. [1990] examined the residual Sq amplitude after subtraction of the linear fit to the sunspot number. They found significant trends in the residual at some stations, but the trends were not always consistent among stations. Elias et al. [2010], conducting a similar analysis, showed increasing trends in the residuals at Apia (13.8 • S, 188.2 • E), Fredericksburg (38.2 • N, 282.6 • E), and Hermanus (34.4 • S, 19.2 • E) during . They argued that the observed increasing trends are partly due to an increase of CO 2 as well as a secular decrease in the main geomagnetic field strength. Earlier, Rishbeth and Roble [1992], using a numerical model, predicted that an increase in the CO 2 concentration would lead to an increase of the plasma density at dynamo region altitudes. Jarvis [2005], on the contrary, found a long-term decrease in the diurnal and semidiurnal spectral power at Lerwick (61.1 • N, 358.8 • E), Niemegk (52.1 • N, 12.7 • E), and Tucson (32.2 • N, 249.3 • E) after removing components correlated with the sunspot number and aa index. The long-term reduction was attributed to a long-term reduction in the solar diurnal and semidiurnal tides from the lower atmosphere, which was reported earlier by Bremer et al. [1997] based on wind measurements. The reduction in tidal forcing from the lower atmosphere was predicted to occur by Ross and Walterscheid [1991] in connection with a depletion of stratospheric ozone. de Haro Barbas et al. [2013], by carefully comparing the residual Sq trends at Apia, Fredericksburg, Hermanus, Bangui (4.4 • S, 18.6 • E), and Trivandrum (8.5 • N, 77.0 • E) with model results, suggested that the secular variation of the main geomagnetic field is the primary source of long-term Sq changes at those stations.
Besides solar daily variations, ground geomagnetic perturbations also show lunar spectral components. The main component is the semidiurnal lunar variation with a period of 12 h in lunar time, or ∼12.42 h in solar time. The geomagnetic lunar (L) variation is usually much smaller than the geomagnetic solar (S) variation, approximately one tenth in magnitude [Yamazaki et al., 2011]. Atmospheric lunar tides are responsible for the ionospheric currents that produce the L variations on the ground [Maeda and Fujiwara, 1967;Tarpley, 1970b]. Lunar tidal waves are excited near the surface due to the gravitational forcing of the Moon acting on the lower atmosphere, solid Earth, and ocean [Lindzen and Chapman, 1969;Hollingsworth, 1971]. Those waves can propagate vertically upward into the ionospheric dynamo region, just as upward propagating solar tides do. In the dynamo region, the lunar semidiurnal tide in the wind attains the amplitude of up to 10-15 m s −1 [Stening et al., 1994;Zhang and Forbes, 2013;Forbes et al., 2013], which is smaller than the solar semidiurnal tide by a factor of 3-5 at the same level. Despite its relatively small amplitude, atmospheric lunar tides have been extensively studied in the past. The primary reason for this is the fact that the lunar tidal forcing is, in principle, very well known unlike forcing for other atmospheric waves. Therefore, for example, observations of the atmospheric response to lunar tidal periods serve as an excellent test for general circulation models that incorporate the lunar tidal forcing [Vial and Forbes, 1994;Pedatella et al., 2012a].
Lunar tidal effects on the upper atmosphere have received renewed attention lately, in the context of vertical atmospheric coupling during stratospheric sudden warming (SSW) events. Observations have indicated enhanced semidiurnal lunar tidal perturbations in the geomagnetic field during SSW events [Fejer et al., 2010;Park et al., 2012;Yamazaki et al., 2012aYamazaki et al., , 2012b. At the magnetic equator, the amplitude of L variations during SSWs is more than twice as large as that during non-SSW periods [Yamazaki, 2013]. Significant changes were also found in S variations at midlatitudes [Yamazaki et al., 2012c]. Numerical studies have shown that changes in the background atmosphere during SSWs provide conditions favorable for amplification of lunar and solar semidiurnal tidal waves from the lower atmosphere [e.g., Stening et al., 1997;Pedatella et al., 2012bPedatella et al., , 2014Forbes and Zhang, 2012;Jin et al., 2012;Wang et al., 2014]. Those studies demonstrated how changes in the lower atmosphere can drive upper atmospheric variability through lunar and solar semidiurnal tidal waves, which transfer energy and momentum from the lower atmosphere to the upper atmosphere.
In contrast to the S variation, long-term changes in the L variation have not been studied much. The dependence of the L variation on solar activity is known as in the S variation. Chapman et al. [1971] and Malin et al. [1975] reported that the sensitivity of the L variation to the sunspot number is significantly smaller than that of the S variation, from which they argued that ionospheric currents responsible for the L and S variations may flow at different levels of the ionosphere. They considered that L currents may flow in the F region. Stening and Winch [1979] noted that the effect of solar activity on the L variation varies for different stations and for different periods. Butcher [1980] and Matsushita and Xu [1984] discussed that L currents flowing in the F region is improbable, and both L and S currents should flow primarily in the ionospheric dynamo YAMAZAKI AND KOSCH ©2014. The Authors. region. Later, De Meyer [2003], carefully analyzing L and S variations at Dourbes (50.1 • N, 4.6 • E), concluded that the influence of the solar cycle on L and S variations is nearly the same. Changes in the L variation on time scales longer than a solar cycle are far less known.
In this study, geomagnetic data during the last 100 years are analyzed to shed light on long-term changes in both L and S variations. We examine the semidiurnal component of L and S variations, focusing on long-term changes on time scales of a solar cycle and longer. Similarities and differences between the two variations are discussed.

Data and Methods
Hourly mean data of the geomagnetic field were obtained from the World Data Center for Geomagnetism (Edinburgh). The data from eight midlatitude stations located within ±60 • latitude during 1903-2012 were analyzed. The location of the stations and their coordinates are shown in Figure 1 and Table 1, respectively. We use the eastward Y component. In previous studies of long-term Sq variations, the northward X component (or horizontal H component) was preferred for no obvious reason [Jarvis, 2005;Elias et al., 2010;de Haro Barbas et al., 2013]. At midlatitudes, the vertical Z component is the most robust to geomagnetic disturbances among the three components of the geomagnetic field, X, Y, and Z [e.g., Yamada, 2002]. The Z component is, however, strongly subject to the effect of currents induced in the ocean and conducting Earth, which depends on the exact location of observatories [e.g., Kuvshinov and Utada, 2010]. Takeda [2013a] demonstrated how long-term variations in Sq(Z) can be influenced by relocation of an observatory. The effect of geomagnetic activity is most evident in X at midlatitudes. Moreover, Sq(X) is very sensitive to the center position of the Sq current whorl, which varies significantly from day to day [Stening et al., 2005a[Stening et al., , 2005b and season to season [Campbell andSchiffmacher, 1985, 1988]. Sq(Y) is much more robust to spatial changes of the overhead current system as well as changes in geomagnetic activity. Hence, the Y component is used in this study. The X or H component could be also used if the effect of magnetospheric ring currents is properly evaluated using the Dst index and removed [e.g., Takeda, 1984]. However, the Dst index can go back only about 50 years, and thus, we did not attempt to analyze the X or H component in this study.
A 60 day running median value was subtracted from each hourly data point, which removes the secular variation of the main geomagnetic field as well as the effect of magnetospheric currents. The latter can change seasonally [Campbell, 2003;Maus and Lühr, 2005]. The residual was defined as ΔY. ΔY was visually  [Richmond, 1995b]  inspected, and erroneous data were removed. By erroneous data, we mean extreme values that are certainly artifacts. Also, ΔY that exceeds five standard deviations was automatically rejected for each year, which eliminates the data during severe geomagnetic disturbances. For each solar day, a 60 day median daily variation was determined (i.e., the 60 day running median at each solar hour) and subtracted from ΔY, which removes a substantial portion of the S variation. The residual was denoted as ΔY ′ . The procedure described above is illustrated in Figure 2 using an example of the Y component geomagnetic field at Kakioka during September-October 2007. The main geomagnetic field in the Y component is approximately −3700 nT at Kakioka, which can be seen in Figure 2b. The subtraction of the 60 day median baseline gives ΔY, which varies around zero as shown in Figure 2c. Figure 2c also shows the 60 day median daily variation. The deviation of ΔY from the 60 day median daily variation is ΔY ′ , which is plotted in Figure 2d. Figure 2a presents geomagnetic activity index Kp. The robustness of the Y component to geomagnetic activity can be seen. L and S semidiurnal variations were derived from ΔY ′ and ΔY, respectively, in the following way. For each year, ΔY and ΔY ′ were grouped into three seasons: D months, E months, and J months. The D months are from November of the previous year to February. The E months are March, April, September, and October. The J months are from June to August. This seasonal binning was needed because the phase of L and S variations varies with seasons. For each seasonal group of the data, L and S semidiurnal variations were derived from ΔY ′ and ΔY, respectively, using a least squares fitting technique. The L and S semidiurnal variations are expressed as follows: (1) where is lunar time in hour; t is solar time (or local time) in hour; A L and A S are the amplitude of L and S semidiurnal variations in nano tesla, respectively; and P L and P S are the corresponding phases (i.e., lunar or solar time of maximum in hour). In equations (1) and (2) phase of the semidiurnal variations were averaged over running 5 years including the preceding and following 2 years. This time window is long enough to reduce the standard error of the small L variation sufficiently below its amplitude. We do not examine long-term changes for different seasons separately, which would reduce the amount of data and thus require a longer time average.

Results
Figure 3a shows percent changes in the amplitude of the L semidiurnal variation from 1905 to 2010. By percent change, we mean the relative change from the average for the whole period. The average L semidiurnal amplitude varies between stations, taking values between 1.1 and 2.0 nT (inclusive). The green lines indicate the results from each station, and the red line is their average. The variations are largely similar at different stations, indicating that long-term variations are dominated by global changes. This is also the case for the amplitude of the S variation, depicted in Figure 3b. The average S semidiurnal amplitude varies between 8.9 and 12.9 nT (inclusive) for different stations.
The 11 year solar cycle can be clearly seen in both Figures 3a and 3b. The L and S amplitudes change approximately ±20% due to the solar cycle variation. Given that our results are 5 year averaged, the actual year-to-year response of the L and S variations to the solar cycle would be greater. It is important to note that the sensitivity of the L variation to the solar cycle is comparable with that of the S variation, which does not support the claim by previous researchers that the S variation is 2-3 times more sensitive to the solar cycle [Chapman et al., 1971;Malin et al., 1975;Olsen, 1993]. Indeed, a later study found that the sensitivity of L and S variations to the sunspot number is nearly the same [ De Meyer, 2003]. It can be seen in the figures that changes in the L and S amplitudes differs in detail. The difference will be discussed in section 4.
We plot in Figure 3c the 5   1905-2010. The F 10.7 index is a good proxy for the solar extreme ultraviolet (EUV) radiation flux, which creates ionospheric plasma. The data before 1948 are derived from the sunspot number using the following formula given by Tapping et al. [2007]: where N S is annually averaged values of the sunspot number. The F 10.7 index was provided by the Herzberg Institute of Astrophysics for 1948 and onward. The correlation is remarkable between the F 10.7 index and the S variation amplitude. The square of the correlation coefficient R 2 gives a measure of the goodness of the linear fit. We found R 2 = 0.94 for F 10.7 and A S (average of all stations). The correlation is slightly better when √ F 10.7 is used, which gives R 2 = 0.95. This is the case for all stations, i.e., the correlation is better when √ F 10.7 is used, although the difference in R 2 is small. For the amplitude of the L variation (average of all stations), we obtained R 2 = 0.68 for F 10.7 and R 2 = 0.70 for √ F 10.7 . The use of √ F 10.7 yields a better correlation than F 10.7 at all stations. We will discuss later in section 4 why √ F 10.7 tends to fit better than F 10.7 to both L and S amplitudes. Figure 3d shows the 5 year average of the interplanetary electric field BV, which dominates the high-latitude electric field. B is the magnitude of the interplanetary magnetic field (IMF), and V is the solar wind speed. Strictly speaking, the high-latitude electric field is controlled by the IMF clock angle as well [Shepherd et al., 2002]. However, the dependence on the IMF clock angle becomes negligible on time scales of yearly average [Lockwood et al., 2013], which justifies the use of BV as a measure of the solar wind effect on the high-latitude electric field. The solar wind parameters before 1965 are estimated from geomagnetic activity by Svalgaard andCliver [2007, 2010], while they are from satellite data in the OMNI database [King and Papitashvili, 2005] for 1965 and onward. (It is noted that an updated version of the BV estimate from  Table 2. geomagnetic activity was recently made available by Lockwood et al. [2014].) The variations in F 10.7 and BV are similar not only for the 11 year solar cycle but also for a longer-period variation, which is characterized by a steady increase from the beginning through 1950s, a local minimum around 1965, and a declining trend since 1980s. Those long-term variations can be found in the L and S amplitudes as well. A correlation analysis reveals R 2 = 0.58 for BV and A S , and R 2 = 0.57 for BV and A L .
Long-term changes in the phase of the L and S variations are depicted in Figure 4. It is noted that the results are 5 year YAMAZAKI AND KOSCH ©2014. The Authors. averaged, so that the actual year-to-year variability would be somewhat larger. The figure shows only deviations from the average value at each station. It can be seen that the phases are fairly stable during the period analyzed. For the results averaged for all stations (red), the maximum deviation is less than 1 h. The 11 year solar cycle is visible in the phase of the S variation. The results for the L variation are more noisy, and the solar cycle variation is not apparent. It is known that the focus of the Sq current system moves to later local times during solar maximum [Olsen, 1993], which produces the solar cycle variation in the phase of the S variation. The physical mechanism is yet to be investigated. Figure 5 shows the ratio in the amplitude of L to S semidiurnal variation during 1905-2010. It can be seen that the amplitude ratio tends to increase with decreasing solar activity. Besides the 11 year solar cycle, the amplitude ratio also reveals a persistent decrease throughout the period. The lunar-to-solar ratio in 2010 is approximately 10% lower than that of a hundred years ago. The long-term decrease in the lunar-to-solar ratio is largely monotonic (i.e., there is no apparent change in the rate of the decrease), and it is distinct  Table 3. from the long-term trends in F 10.7 and BV, which show rises and falls (Figures 3c and 3d). The long-term trends at individual stations as well as the trend in the average are listed in Table 2, which includes the 95% confidence interval derived using a bootstrap method with 1000 iterations. The long-term reduction is significant (>95%) at all the stations except Niemegk, where the trend is small and insignificant. The ratio of the reduction is found to be greater in the Southern Hemisphere than in the Northern Hemisphere.
To provide insight into whether the long-term reduction in the lunar-to-solar ratio is due to changes in the L variation or S variation (or both), we analyzed the long-term trend in the L and S amplitudes separately. Since we found that both A L and A S correlate slightly better with √ F 10.7 than F 10.7 , we use the linear fit of √ F 10.7 to A L and A S as a measure of the solar activity influence on the L and S variations, respectively. The fitting was made separately for A L and A S at each station. We subtracted YAMAZAKI AND KOSCH ©2014. The Authors. those fits from the original A L and A S . Variations in the residuals are presented in Figure 6. The green lines indicate the results from each station, and the red line is their average. The trends at individual stations and the trend of the average can be found in Table 3. At most stations, the declining trend in the L amplitude is significant. Meanwhile, the trend in the S amplitude is positive at most stations, and some of them are significant (>95%). These results indicate that the long-term reduction in A L /A S is mainly due to the reduction in A L . The increase in A S also contributes but to a smaller extent.

Discussion
The interpretation of geomagnetic variations is sometimes complicated by the fact that various processes are involved in changes of ionospheric currents. We take a rather simplified view of Takeda [2013b] in order to understand our observational results. Takeda [2013b] used a very simple expression for the magnitude (A) of the Y component geomagnetic daily variation on the basis of Fukushima's [1979] model: where 0 is the permeability of the vacuum (= 4π × 10 7 ), Σ is the effective ionospheric conductivity (S), U is zonal neutral wind speed (m s −1 ), and B z is the vertical geomagnetic field strength (nT). The constant 3/4 arises from the fact that the geomagnetic daily variation is produced not only by ionospheric currents but also by currents induced in the conductive Earth, which accounts for approximately 25% of the total effect. In light that the midlatitude ionospheric current system is subject to the Cowling effect [Takeda, 1991], the effective ionospheric conductivity is given as follows: where Σ P and Σ H are height-integrated Pedersen and Hall conductivities, respectively. Takeda [2013b] computed Σ P and Σ H using the International Reference Ionosphere model [Bilitza, 1990;Bilitza et al., 2011]. He pointed out that the solar cycle variation in A is almost solely due to changes in Σ.

Correlation of F 10.7 With L and S Variations
It is found that both lunar and solar variations correlate slightly better with √ F 10.7 than F 10.7 . According to Takeda [2013b], Pedersen and Hall conductivities at a given height are proportional to the electron number density N e . The electron number density tends to scale with √ F 10.7 in the dynamo region where the ionospheric plasma is largely in photochemical equilibrium. The dominant plasma species is O + 2 , which is produced by photoionization at a rate J (s −1 ) and lost through recombination with electrons at a rate (s −1 ). That is, where represents an EUV photon. Photochemical equilibrium implies the following:

Differences in the Solar Cycle Response of L and S Variations
We found differences in the solar cycle response of L and S variations (Figures 3a and 3b). There are solar minimum periods when the L amplitude does not decrease as much as the S amplitude does, e.g. , 1923, 1944, and 1996. It is possible that the L amplitudes for these years are dominated by the effect of SSWs, which significantly enhances L variations. Although the occurrence of SSWs is basically limited to the northern winter periods, the effect on L variations is strong enough to affect its yearly average amplitude [Yamazaki, 2013]. Yamazaki [2013] identified the occurrence of SSWs for the years 1994, 1995, 1997, and 1998. Since we use 5 year averaged amplitudes, A L around the year 1996 should be under significant influence of these SSWs. We could not confirm whether the L amplitudes for 1923 and 1944 are also affected by SSWs, because there is no stratospheric data to identify the occurrence of SSWs for these periods. There is indirect evidence to suggest the occurrence of an SSW for 1923. Bartels and Johnston [1940] reported an extraordinary large L variation at the magnetic equator for January 1923. An abnormally enhanced L variation at the magnetic equator is often observed during a strong SSW event. SSWs also affect S variations but in a different way [Yamazaki et al., 2012c]. Relative changes in S during SSWs are, on average, much smaller compared to those in L (Y. Yamazaki, Solar and lunar ionospheric electrodynamic effects during stratospheric sudden warmings, submitted to Journal of Atmospheric and Solar-Terrestrial Physics, 2014).
There may be other atmospheric processes that modulate year-to-year variability of the L amplitude. For example, Pedatella and Forbes [2009] and Pedatella and Liu [2013] presented evidence that upward propagation of some tides is affected by the El Niño-Southern Oscillation (ENSO). Similarly, the quasi-biennial oscillation (QBO) in the stratosphere and mesosphere has some influence on tidal propagation [Hagan et al., 1992[Hagan et al., , 1999Xu et al., 2009]. Indeed, Olsen [1994] and Jarvis [1996,1997] reported small but significant QBO signatures in the S variation. The effect of ENSO and QBO on the lunar tide, however, has not been studied.
It is likely that the solar cycle effect on the L variation presented by earlier researchers is more or less affected by the effect of SSW (and possibly by other atmospheric effects). This is a significant issue when one attempts to determine the sensitivity of the L variation to solar activity from a relatively small data set, e.g., a few years respectively from a solar maximum and solar minimum [e.g., Chapman et al., 1971;Malin et al., 1975].

Solar Cycle Effect on the Lunar-to-Solar Amplitude Ratio
We have shown that the amplitude ratio in the L to S variation increases with decreasing solar activity ( Figure 5). In light of equation (4), L and S semidiurnal amplitudes may be assumed to be proportional to the effective ionospheric conductivity Σ, zonal neutral wind speed U, and geomagnetic field strength B z . Designating U L and U S as the wind speed for lunar and solar semidiurnal tides, the amplitude ratio in the L to S variation can be expressed as follows: where the asterisks (*) denote the effect of upward propagating tides from the lower atmosphere, while the circle (•) represents the contribution of the tides locally generated in the dynamo region. Enhanced solar heating during high solar activity periods would lead to greater solar heating in the dynamo region and hence a larger U • S . Also, there is evidence in the literature [Bremer et al., 1997;Baumgaertner et al., 2005] that solar semidiurnal tides from the lower atmosphere decrease with increasing solar activity. The mechanism for this is not fully understood, but changes in the background atmosphere, and thus changes in the propagation conditions, may be a reason. If that is the case, a similar solar cycle dependence would be expected in upward propagating lunar semidiurnal tides. Therefore, the negative response of A L ∕A S to solar activity can be qualitatively understood as arising from the increased in situ forcing (thus increased U • S ) and decreased lower atmospheric forcing (thus decreased U * L and U * S ) during high solar activity periods.

Long-Term Changes in L and S Variations
Our results have revealed a long-term reduction in the amplitude ratio A L ∕A S ( Figure 5 and Table 2). This is probably due to a reduction of the lunar tidal wind in the dynamo region. A separate analysis of L and S variations ( Figure 6 and Table 3) has also indicated a negative trend in A L , while the trend tends to be positive for YAMAZAKI AND KOSCH ©2014. The Authors.

10.1002/2014JA020203
A S . The global-scale centurial increase in the S range was also reported by Svalgaard [2009]. The long-term increase in the S variation can be attributed to an increase in ionospheric conductivities associated with the decrease of the geomagnetic dipole moment. Takeda [1996] numerically studied how the strength of the main geomagnetic field affects the ionospheric wind dynamo. Cnossen and Richmond [2013], also using a numerical model, described the effect of secular changes in the Earth's magnetic field on Sq variations; a comparison with observations was presented later by de Haro Barbas et al. [2013]. The L variation should be subject to the same effect from the increasing ionospheric conductivities, but a reduction in the lunar tidal wind is probably dominant, so that the net effect produces a declining trend. It is not clear what causes the long-term reduction in the lunar tidal wind.
Climate change in the lower atmosphere has an impact on the upper atmosphere through various mechanisms [Cnossen, 2012]. Long-term changes in the atmospheric composition, such as an increase of CO 2 , can affect the dynamics of the lower and middle atmosphere, through which lunar tidal waves propagate. Numerical studies will be necessary to gain a better understanding for long-term changes in the lunar tidal wind.

Conclusions
We have examined geomagnetic lunar (L) and solar (S) semidiurnal variations at eight midlatitude stations during 1903-2012. The L and S variations are a consequence of electric currents flowing in the ionospheric dynamo region between approximately 90 and 150 km, driven by lunar and solar tidal winds, respectively. The following are the main results of the present study: 1. Five year averaged amplitudes of the L and S semidiurnal variations show prominent 11 year solar cycle variations of approximately ±20%. The sensitivity of the L variation to solar activity is comparable with the sensitivity of the S variation to solar activity, consistent with the conclusion of De Meyer [2003]. Changes in the L amplitude can be contaminated by the effect of the stratospheric sudden warming (SSW), as well as other atmospheric effects. 2. Both L and S amplitudes correlate with the solar activity index F 10.7 . In both cases, the correlation is slightly better with √ F 10.7 than F 10.7 . This is probably due to the fact that the effective ionospheric conductivity tends to be proportional to the electron number density, which scales with the square root of the plasma production rate under photochemical equilibrium. 3. The solar cycle variation is also present in the phase of the S semidiurnal variation but not apparent in the L variation phase. 4. The ratio in the amplitude of L to S semidiurnal variation decreases with increasing solar activity. This can arise from an increased in situ forcing and decreased lower atmospheric forcing during high solar activity periods. 5. The lunar-to-solar amplitude ratio also shows a long-term reduction of approximately 10% per century. This is due to a reduction in the L amplitude, as well as an increase in the S amplitude to a smaller extent. The long-term increase in the S amplitude is attributed to an increase in ionospheric conductivities due to the secular variation of the geomagnetic field. The long-term decrease in the L amplitude is probably due to a decrease of the lunar tidal wind in the dynamo region, which is possibly linked to climate change in the lower atmosphere.