Wind measurements in the mesosphere and lower thermosphere obtained with a medium frequency (MF) radar in Hawaii (22°N, 160°W) spanning ∼16 years are employed to examine the intra-annual and interannual variability of the mean and tidal motions at altitudes between 84 and 94 km. Intra-annual periodicities range from ∼3 to 12 months, with significant coherence in altitude and between the zonal and meridional components of each motion field. Interannual variations confirm the dominant periodicities identified previously at this site and elsewhere, in particular, the significant diurnal, and less significant semidiurnal, tidal responses at periods of ∼28 and ∼48 months. Amplitudes of these long-period oscillations of the diurnal tide increase with altitude below 92 km and are larger than the amplitudes of the 12 month oscillations above ∼90 km. Phases of the ∼28 and ∼48 month oscillations show a downward progression with a slightly larger altitude variation in the meridional diurnal tide for the ∼28 month oscillation and a significantly larger altitude variation in the zonal diurnal tide for the ∼48 month oscillation. The long and nearly continuous Hawaii data set also enables characterization of the responses of the wind fields to the 11 year solar cycle. Both wind and tidal fields exhibit this periodicity, but these responses display interesting and different relations to the phase of the solar cycle. The 11 year oscillation of the meridional wind is nearly in phase with the solar cycle, while the 11 year oscillation of the zonal wind is in approximate quadrature with the solar cycle. The 11 year oscillations of the semidiurnal tidal amplitudes and the meridional diurnal amplitude are all in approximate quadrature with the solar cycle (with the tidal amplitudes leading by ∼29 to 37 months), while the 11 year oscillation of the zonal diurnal amplitude is somewhat nearer an antiphase than a quadrature relation to the solar cycle (leading by ∼54 months).
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Long-term variations in atmospheric circulation and structure (on time scales of decades or less) have been of research and societal interest for many years. The most obvious potential sources of such variations external to the atmosphere are the ∼11 year variations in solar flux, episodic volcanic activity, and anthropogenic changes induced by greenhouse gases and other emissions. Sources of interannual variability internal to the atmosphere or the coupled atmosphere-ocean system include the El Niño–Southern Oscillation (ENSO) of the equatorial Pacific, the quasi-biennial oscillation (QBO) at equatorial latitudes, the North Atlantic Oscillation, and the Pacific Decadal Oscillation.
 Variations in the density, composition, circulation, and structure of Earth's thermosphere due to solar cycle variations in UV and EUV radiation are large and have been recognized for decades (see Dickinson  and Fuller-Rowell and Rees  for reviews of earlier studies). Similar variations are also suspected to occur throughout the lower and middle atmosphere in response to solar cycle variations in long- and short-wave radiation, but compelling evidence of such responses has proven more elusive to researchers.
 Related studies have addressed long-term trends and interannual variability of mean temperatures and mean and tidal winds in the mesosphere and lower thermosphere (MLT). These may occur via direct responses to variable solar inputs in the MLT or as responses to external or internal influences at lower altitudes via upward coupling of wave energy and momentum. Temperature measurements at northern midlatitudes between 41°N and 56°N from 1923 to 1995 show negative trends of approximately −0.1 to −0.9 K yr−1 from 25 to 100 km [Lysenko et al., 1999]. Wind measurements at Molodezhnaya, Mawson, and Davis, Antarctica from 1970 to 2006 [Merzlyakov et al., 2009] exhibit a negative trend in the eastward wind during summer and a positive trend in the southward wind during winter while the westward wind decreases during summer after 1993. Merzlyakov et al.  also report negative trends of the meridional semidiurnal tide during winter and summer. At northern midlatitudes, the annual mean zonal wind shows a decreasing trend until the 1980s while meridional wind shows an increasing trend until the 1990s [Portnyagin et al., 2006; Merzlyakov and Portnyagin, 1999]. In the equatorial region, the annual mean meridional wind shows a gradual change from northward to southward from 1993 to 2005 [Sridharan et al., 2007]. Keuer et al.  observed an altitude variation of the long-term trend of winds at Juliusruh from 1990 to 2005. They found negative (positive) trends of the eastward wind in summer (winter) below ∼85 km, with these trends reversing above ∼85 km. The northward wind, on the other hand, exhibited a positive trend, but with decreasing amplitude with increasing altitude, from ∼70 to 80 km, during summer, with a similar trend during winter at ∼70 km. These trends also reversed and increased in amplitude at higher altitudes.
 There is also an interannual variability of winds that is apparently not consistent between hemispheres [Portnyagin et al., 1993a]. Amplitudes of the meridional diurnal tide from High Resolution Doppler Imager (HRDI) measurements show a larger interannual variability prior to June 1993 than subsequently [Burrage et al., 1995]. Lieberman et al.  examined the diurnal tidal amplitude from 1997 to 1998 when amplitude anomalies were observed at Kauai (22°N, 154°W), Hawaii, and Christmas Island (2°N, 157°W). They found water vapor heating to provide strong forcing of the migrating diurnal tide and latent heating to be a major source of the nonmigrating diurnal tidal modes. Using a primitive equation model, they also demonstrated that interannual variations of the diurnal tidal amplitude are related to water vapor heating. Semidiurnal tidal amplitudes also exhibit long-term amplitude decreases [Jacobi et al., 2005; Portnyagin et al., 2006; Merzlyakov and Portnyagin, 1999] and large (a factor of ∼2) interannual variations [Fritts and Isler, 1994; Burrage et al., 1995], but again with interannual variability that is not consistent between hemispheres [Portnyagin et al., 1993b].
 Much of the interannual variability in mean and tidal winds remains unexplained at present. Of the influences that have been diagnosed, the most apparent signatures appear to exhibit solar cycle and/or QBO periodicities. Sprenger and Schminder  analyzed ionospheric drift measurements at Kühlungsborn (54.1°N) and Collm (51.3°N), Germany, from 1957 to 1968 and found the mean zonal wind to be positively correlated with the solar cycle (i.e., larger zonal winds at solar cycle maxima), but with the semidiurnal tidal amplitude negatively correlated with the solar cycle.
 More recent studies employing data collected at a variety of sites by Dartt et al. ; at Scott Base, Antarctica, by Baumgaertner et al. , Fraser et al. , and Fraser ; at Saskatoon, Canada, by Namboothiri et al. [1993, 1994]; at Juliusruh, Germany, by Keuer et al. ; and in the UK by Pancheva et al.  yielded trends that often differed qualitatively. Several factors may have contributed to these differences, including relatively weak responses, insufficient data and significance, differing data intervals, influences of long-term oscillations, or instrument changes over the course of the observations. Thus, there is a need for further studies employing longer and coincident data sets where such measurements are possible.
 Our purpose in this paper is to assess the long-term oscillations of the mean winds and the semidiurnal and diurnal tides determined using MF radar wind measurements at Kauai (22°N, 160°W), Hawaii, performed nearly continuously from October 1990 to August 2006. The wind measurements at Kauai and our analysis approach are described in section 2. Results presented in section 3 include climatological mean seasonal variations, interannual variations, and long-term oscillations. Mean winds and the diurnal and semidiurnal tides all exhibit expected annual and semiannual periodicities, though with significant variations in altitude. The diurnal tide displays variability between 12 and 24 months and strong responses at ∼28 and ∼48 month periods that are much less pronounced or absent in the semidiurnal tidal response. Section 4 focuses on apparent 11 year oscillations of these wind fields, and the diurnal and semidiurnal tides are both seen to exhibit responses that increase in altitude but vary in relative phase. A summary of these results is presented in section 5.
2. Wind Measurements and Analysis Approach
 A medium frequency (MF) radar system was installed at Kauai (22°N, 160°W), Hawaii, in September 1990. The system utilized a spaced antenna technique to determine horizontal wind velocities from partial reflection drift signals from the D and lower E regions of the ionosphere [Murayama et al., 2000; Vincent and Lesicar, 1991]. The Kauai MF radar operated at a frequency of 1.98 MHz with a peak power of 25 kW, transmitting 25 μs pulses. Details of the Kauai MF radar and an initial assessment of the horizontal winds were described by Fritts and Isler . The Kauai MF radar had a time resolution of 2 min and a height resolution of ∼4 km oversampled at 2 km from 60 to 98 km. Only altitudes from 84 to 94 km were used in this analysis, however, because the data were most continuous in this altitude range.
Figure 1 shows measurement periods by the Kauai MF radar. The system operated from 28 September 1990 to 10 September 2006, almost continuously, and additionally from 3 February 2007 to 19 March 2007. For our study of long-period oscillations, Kauai MF radar data from October 1990 to August 2006 were used. The radar did not operate for two intervals greater than a month in duration from 27 July 1993 to 11 September 1993 and from 1 August 1996 to 2 September 1996. Additionally, the radar did not operate for more than half of the months of September 1992, January and February 1994, February 1996, July 1997, and December 2001.
 Winds were measured at 2 min intervals in each 2 km altitude bin and averaged to create hourly means where 10 or greater 2 min wind measurements were available. Spectral analyses (not shown) revealed the same dominant periodicities in the zonal and meridional wind fields, with somewhat stronger meridional responses in general. The diurnal tide was the dominant feature in the wind fields, with maximum amplitudes occurring typically during March and August. The semidiurnal tide was also significant and achieved maxima typically during January and July.
 Hourly mean winds at each altitude were fitted with 12 h and 24 h sinusoids in 4 day intervals for which at least 75% of the hourly means were available. This procedure yielded 4 day mean winds and amplitudes and times of maxima for the semidiurnal and diurnal tides. Hereafter, the times of the maxima will be referred to as phases.
 The resulting 4 day mean winds and tidal amplitudes of the zonal and meridional components during a calendar month were averaged separately to determine monthly means. Tidal phases were averaged in the same manner, but weighted by the 4 day amplitudes. The resulting monthly mean tidal amplitudes and phases were compared with the results of vector means determined from the tidal amplitudes and phases and confirmed that similar results were obtained. Monthly mean winds and tidal amplitude and phases estimates were also assessed with 4 day interval fits including and excluding 48 h (2 day) and 96 h (4 day) motions, and separately from monthly composite days. These tests confirmed that the various estimation methods yielded monthly mean winds and tidal fits that were robust and agreed closely among different methods. Finally, monthly values for months having less than half the total data were determined from a cubic spline interpolation.
3.1. Seasonal Variations
 Seasonal variations of the mean winds and tidal and 2 day amplitudes during initial observations were presented by Fritts and Isler [1992, 1994] and Isler and Fritts . Mean seasonal variations of the winds over a longer period from 70 to 90 km were presented by Gavrilov et al. . The present paper describes seasonal and longer variations of mean winds and tidal motions spanning the entire data set.
Figure 2 shows monthly mean winds and semidiurnal and diurnal tidal amplitudes for altitudes from 84 to 94 km averaged from 1990 to 2006. Figures 2a, 2c, and 2e show zonal components, and Figures 2b, 2d, and 2f show meridional components. The zonal wind exhibits a larger altitude variation with a clear semiannual oscillation. Eastward winds maximize from December to January below 88 km and during June above 90 km. Maxima are ∼22 ms−1 at 86 km in January and ∼14 ms−1 at 92 km in June. Zonal winds are westward from March to April with amplitudes increasing downward and maximum of ∼14 ms−1 at 86 km.
 Seasonal and altitude variations of the meridional wind are less clear. Southward winds dominate from May to July at all altitudes, maximizing at 94 km in June at ∼7 ms−1. On the other hand, northward winds dominate from November to February, maximizing at 86 km in February at ∼5 ms−1. The largest altitude variation of the meridional wind is in March, with a transition from southward to northward between 92 km and 94 km. Northward winds are typically a winter phenomenon while southward winds are a summer phenomenon.
 The semidiurnal tide is enhanced from June to August in the zonal component with a maximum of ∼11 ms−1 at 94 km in August. While the amplitude of the semidiurnal tide decreases with altitude in June and July, the amplitude increases with altitude in August. However, the seasonal and altitude variations of the zonal semidiurnal tidal amplitude are generally small. While the semidiurnal tide is enhanced during June in both components, the meridional semidiurnal tide maximizes from November to February, with a maximum of ∼13 ms−1 at 92 km in December. Therefore, a semiannual oscillation of the semidiurnal tide is seen more clearly in the meridional component.
 The diurnal tide exhibits somewhat similar structure between the zonal and meridional components, with both maximizing in March and August. Altitudes of the maxima are higher in March than in August, with maxima in both March and August larger in the meridional than in the zonal component. Both maxima in the zonal component are ∼25 ms−1, while both maxima in the meridional component approach ∼30 ms−1. We note, however, that these maxima may be skewed toward lower amplitudes and/or altitudes due to the tendency for MF radars to underestimate wind amplitudes above ∼90 km [Cervera and Reid, 1995; Hocking and Thayaparan, 1997].
Figure 3 shows monthly mean phases of the semidiurnal and diurnal tides as functions of altitude between 84 and 94 km. The earliest phases of the semidiurnal tide occur at 90 km in autumn in both zonal and meridional components. These occur at 2.7 h UT for the zonal component and −2.3 h UT, or equivalently 9.7 h UT, for the meridional component. With the latest phase of the meridional component at 90 km in August, there is a significant phase retardation by ∼6.3 h from August to September. Typically, the phase of the semidiurnal tide is earlier when the amplitude is smaller and later when the amplitude is larger.
 The phase of the diurnal tide exhibits nearly uniform downward progression between 84 and 94 km throughout the year. The earliest phases occur in December in both components and the latest phases occur in July. The longest vertical wavelength in the zonal component at altitudes between 84 and 94 km is observed in May, at ∼62 km, and the shortest is observed in September, at ∼41 km. The vertical wavelength in the meridional component is longer than that in the zonal component, especially from November to January, with a maximum of ∼74 km. These wavelengths are much longer than those for the (1, 1) mode predicted theoretically [Forbes, 1995]. The meteor radar wind measurements at Maui (20.8°N, 156.4°W), Hawaii, from 2002 to 2007 exhibit a much smaller vertical wavelength, 32±3 km, between 85 and 95 km, as an annual average. The phases of the diurnal tide observed by the Kauai MF radar agree well with the phases of the migrating diurnal tide predicted by the Global-Scale Wave Model (GSWM) [Hagan et al., 1999, 2001] below 90 km throughout the year. However, the phases determined from the Kauai MF radar measurements are earlier than the GSWM predictions above 90 km, especially during winter. Possible explanations include (1) tidal phase determinations with the MF radar may contain systematic errors above ∼90 km and (2) the diurnal tide over Hawaii likely represents a mixture of migrating and nonmigrating components.
3.2. Interannual Variations
 Interannual variations of mean winds over Hawaii determined with the MF radar were presented by Gavrilov et al.  in terms of variances. However, variances are typically large when and where the mean amplitude is large. Interannual variations of the mean winds and semidiurnal and diurnal tidal amplitudes are presented in Figure 4, focusing on 90 km in a manner similar to Baumgaertner et al.  and Dowdy et al. . Monthly means for each year are shown in gray and the means over the period from 1990 to 2006 are shown in black. It is apparent that interannual variations are seen more clearly in the zonal wind than in the meridional wind. Maximum standard deviations of the multiyear mean winds are ∼15 ms−1 in April in the zonal wind and ∼7 ms−1 in January in the meridional wind. The monthly averages of the zonal wind in April range from a minimum of ∼−26 ms−1 in 1993 to a maximum of ∼11 ms−1 in 2004.
 Standard deviations of the monthly semidiurnal tidal amplitudes are similar between the zonal and meridional components. Maximum standard deviations for the multiyear averages are ∼5 ms−1 in March for both zonal and meridional components. Comparatively, large standard deviations are observed from November to February in the meridional component when the amplitude is large.
 The diurnal tide shows a significant semiannual oscillation with maxima in March and August. However, the largest standard deviations are in September, ∼12 ms−1 for both zonal and meridional components. Amplitudes during this period range from ∼12 ms−1 in 1991 to ∼51 ms−1 in 1997 for the zonal component and from ∼9 ms−1 in 1991 to ∼48 ms−1 in 1997 for the meridional component. Interannual variations of monthly diurnal tidal amplitudes of the zonal and meridional components are also presented by Lieberman et al. .
3.3. Long-Period Oscillations
 Time series of monthly mean winds from 1990 to 2001 were shown previously by Gavrilov et al.  at altitudes between 70 and 90 km. Fourier transforms of monthly mean winds and semidiurnal and diurnal tidal amplitudes from January 1991 to December 2005 from 84 to 94 km are shown in Figure 5, with a 90% significance level for each component. Clear oscillations occur at 6 and 12 month periods in both zonal and meridional winds, decreasing in amplitude with altitude in the zonal wind and increasing in the meridional wind. Four month (terannual) and longer-period oscillations are also seen in the zonal wind and also decrease in amplitude with altitude in each case.
 The semidiurnal tide exhibits clear semiannual and annual oscillations in both zonal and meridional components. Amplitudes of the semiannual oscillations of the semidiurnal tide increase with altitude. However, the amplitudes of the annual oscillations of the semidiurnal tide decrease with altitude in the zonal component and increase with altitude in the meridional component. Terannual oscillations also appear, increasing in amplitude with altitude. Amplitudes of longer-period oscillations than 12 months are typically smaller at lower altitudes in both components.
 The diurnal tide exhibits clear semiannual oscillations in both zonal and meridional components, increasing in amplitude with altitude. Annual oscillations are smaller than the semiannual oscillations, especially at higher altitudes. Longer-period oscillations than 12 months are also seen clearly in both zonal and meridional components, with dominant periods of ∼2.4 and 4 years, amplifying with altitude.
 Based on the spectral analyses in Figure 5, the monthly mean winds and semidiurnal and diurnal tidal amplitudes from October 1990 to August 2006 were fit to sinusoids with periods of 3, 4, 6, 12, 28, 48, 96 and 132 months to determine mean amplitudes and phases of these long-period oscillations simultaneously during the measurement period. A period of 28 months was chosen as an average of the quasi-biennial oscillation (QBO) while the true period of the QBO varies from ∼26–30 months [Xiong et al., 1995; Rasmusson et al., 1990]. Individual fits to each oscillation were also performed and essentially the same amplitudes and phases were obtained with much larger standard deviations than the simultaneous fits. Only fits to the long-period oscillations of mean winds and tidal amplitudes were performed because long-period oscillations of tidal phases were below 90% significance levels at all altitudes. We confirmed that the same results for the long-period oscillations were obtained with and without detrending the monthly data.
 Results for the mean winds are shown in Figure 6 including 90% significance levels. Because longer-period oscillations than 12 months in the wind variations are not clear in Figure 5, Figure 6 shows amplitudes and phases of 4, 6, and 12 month periods, as well as a period of 132 months to examine possible variations with the 11 year solar cycle. The phases of these oscillations are presented as time lags of the maxima from 1 January 1990. Amplitudes of the terannual and semiannual oscillations are larger in the zonal than in the meridional component. The amplitude of the terannual oscillation of the zonal wind is nearly constant with an amplitude of ∼4 ms−1 and is less than 1 ms−1 below ∼94 km in the meridional component, but increases slightly with altitude. The amplitudes of the semiannual oscillations of the zonal and meridional winds decrease with altitude from ∼11 and 4 ms−1 at 84 km, respectively. The amplitude of the annual oscillation of the zonal wind decreases sharply with altitude below 92 km. Thus, the semiannual oscillation is dominant at 92 km in the zonal wind field. The amplitude of the annual oscillation of the meridional wind increases gradually with altitude and dominates the meridional wind field above 90 km.
 Phases of the terannual oscillations of the mean winds exhibit slow, but opposite, phase variations in altitudes between the zonal and meridional winds. The maximum of the meridional amplitude lags that of the zonal amplitude by ∼2–3 months. Phases of the semiannual and annual oscillations are nearly constant below ∼90 km in both components, with the zonal lagging the meridional by ∼3 months in the semiannual oscillations and the meridional lagging the zonal by ∼1 month in the annual oscillations. Above ∼90 km, the meridional (zonal) semiannual (annual) oscillation exhibits a phase variance with altitude of ∼2 (∼5) months.
 Amplitudes and phases of the 4, 6, 12, 28, and 132 month oscillations of the semidiurnal tidal amplitude are shown in Figure 7. Because the semidiurnal tidal amplitude itself is small, the amplitudes of these long-period oscillations are likewise small, typically ∼1–2 ms−1 or less. Above 92 km, amplitudes of all oscillations are greater in the meridional component than in the zonal component because the amplitudes typically increase with altitude in the meridional component while they are constant or decrease with altitude in the zonal component. Surprisingly perhaps, the amplitudes of the ∼28 and 132 month oscillations are comparable to the terannual, semiannual, and annual oscillations at higher altitudes.
 Despite their small amplitudes, the phases of the various long-period oscillations of the semidiurnal tidal amplitudes are reasonably well defined in the meridional component. Both the terannual and semiannual oscillations exhibit upward phase motions, though at different rates between the two components. The annual oscillations have nearly uniform phase with altitude, but a lag of the meridional component by >4 months. The phases of the ∼28 and 132 month oscillations exhibit altitude variations that agree between components within measurement uncertainties.
 Similar results to those just discussed for the semidiurnal tide, but for the diurnal tide, are displayed in Figure 8. Amplitudes of these oscillations are typically larger than for the semidiurnal tide by ∼2 to 5 times, except for the annual oscillation at higher altitudes, because of the larger diurnal tidal amplitudes. Each oscillation, except the annual, increases with altitude, with a larger meridional than zonal response for all but the terannual and semiannual oscillations.
 Phases of the long-period oscillations of the diurnal tidal amplitudes (see Figure 8) agree well between the zonal and meridional components for the semiannual, annual, and ∼28 month periods, except for the annual oscillations above ∼90 km where the meridional amplitude is very small. Phases for these oscillations also progress downward in each case, in contrast to the upward (or no) phase progressions seen for these periods in the semidiurnal tidal fits in Figure 7. The 48 month fits also exhibit clear phase descents in each component, but at different rates and with greater uncertainties than seen at shorter periods or in the semidiurnal tides. The greater phase uncertainties are likely an indication of a range of long-period influences (at different or varying periods) that contribute to these fits. Where the amplitudes of the 11 year fits are large, these fits likewise exhibit reasonably uniform phase variations with altitude, with the meridional component lagging the zonal by ∼20 to 30 months. Variations at lower altitudes are likely not reliable because of the very small amplitudes at these locations.
 Maxima and minima of the 11 year oscillations of the mean winds and semidiurnal and diurnal tidal amplitudes averaged between 86 and 94 km are shown in Figure 9, together with the maxima and minimum of the solar 10.7 cm radio flux (Solar Geophysical Data, NOAA, Boulder, USA). These show the meridional mean motion to be nearly in phase with the 11 year solar cycle lagging by ∼11 months, with the zonal mean motion nearly in quadrature, with the zonal mean maximum leading the solar cycle by ∼39 months. However, a test of the significance level for the correlation coefficient between the solar flux and the meridional wind shows no significance at the 90% confidence level.
 The 11 year oscillations of both components of the semidiurnal tide and the meridional component of the diurnal tide are in approximate quadrature with the solar cycle, with tidal amplitude maxima leading the solar cycle by ∼29 to 37 months. The zonal component of the diurnal tide, on the other hand, is approximately midway between an antiphase and a quadrature correlation with the solar cycle, leading the solar cycle by ∼54 months. The correlation coefficient between the zonal diurnal tidal amplitude and the solar radio flux is −0.19 at 94 km while the correlation coefficient is −0.08 for the meridional diurnal tide.
 Our analysis of long-term trends revealed that they may exhibit significant variations with altitude. A positive trend in the westward wind was observed between 86 and 94 km maximizing at 90 km. On the other hand, a negative trend of the northward wind was observed between 84 and 94 km, with the largest negative trend at 86 km. Both semidiurnal and diurnal tidal amplitudes exhibited negative trends at ∼84 km, with positive trends above ∼88 km.
 In our analysis, long-term oscillations of the mean winds and tides were obtained with and without the detrended data. Thus, long-term oscillations of the mean winds in the MLT over Hawaii comprise primarily the semiannual and annual oscillations. Mean wind oscillations at 4 month and 11 year periods are comparatively small. They nevertheless exhibit consistent phase variations with altitude, suggesting that they are real.
 The solar 10.7 cm radio flux from February 1947 to September 2008 exhibits a sinusoidal fit at 11 years with maxima in April 1991 and 2002 during the Kauai MF radar measurements. These correspond approximately to our inferred maximum westward and northward winds, as noted previously by Dartt et al.  and seen in Figure 9. Previous studies of the correlations between mean winds and the solar cycle by Bremer et al. , Fahrutdinova et al. , Greisiger et al. , and Namboothiri et al. [1993, 1994] have reported clearer correlations in the zonal than in the meridional wind. A positive correlation with the meridional wind was reported by Sprenger and Schminder , which is consistent with our results. The smaller meridional wind amplitude of the 11 year oscillation found by Sprenger and Schminder  and in our study may indicate why a meridional wind correlation has not been seen by others. However, a positive correlation of the meridional wind and the solar cycle was observed by Keuer et al.  in both summer and winter. This may indicate that the phase of the 11 year cycle of the meridional wind is nearly in phase with the 11 year solar cycle, as also seen in our study.
 Both positive and negative correlations of the zonal wind with the solar flux have been reported, however, likely as a consequence of different measurement periods [Greisiger et al., 1987]. Given that the solar flux maximized in April 2002, it can be expected that the eastward wind decreased and the northward wind increased above the Hawaii MF radar from June 1998 to April 2002, based on the trends noted above. At middle latitudes in the UK (52°N) [Middleton et al., 2002], Saskatoon (52°N), Collm (52°N), and Obninsk (55°N) [Portnyagin et al., 2006], however, a positive trend was observed in both eastward and northward velocities during this period. At the lower latitude of Tirunevelli (8.7°N) [Sridharan et al., 2007], in contrast, a negative trend in the northward wind was observed while a weak positive trend was observed in the eastward wind. Thus, varying long-term trends at different latitudes may influence short-term correlations with the 11 year solar cycle.
Baumgaertner et al.  attempted to explain the negative correlation between the semidiurnal tidal amplitude and solar flux through amplitude variations of the wave number 1 planetary wave in the NCEP/NCAR reanalysis, based on assumptions that (1) the semidiurnal tide observed at Scott Base was mainly a westward propagating zonal wave number 1 mode [Murphy et al., 2006] and (2) the westward propagating zonal wave number 1 semidiurnal tidal mode at high latitudes is excited by nonlinear interaction between the migrating semidiurnal tidal mode and the stationary planetary wave zonal wave number 1 [Angelats i Coll and Forbes, 2002; Forbes et al., 1999]. As noted above, however, the negative correlation of the semidiurnal tide has also been observed at low and middle latitudes where the semidiurnal tide is dominated by the migrating mode [Forbes, 1982a, 1982b; Hagan and Forbes, 2003] and the explanation by Baumgaertner et al.  cannot be applied at these latitudes.
 The semidiurnal tide is mainly excited by O3 insolation absorption of ultraviolet (UV) light in the stratosphere and propagates upward and grows in amplitude as neutral atmospheric density decreases due to conservation of mass and energy [Chapman and Lindzen, 1970; Forbes and Garrett, 1979; Holton, 1973]. According to this theory, increases of temperature and ozone in the stratosphere should cause larger amplitudes of the semidiurnal tide. Temperature and ozone are found to be significantly correlated with the solar cycle [Remsberg, 2008; Li et al., 2008] and can be expected to yield larger semidiurnal tidal forcing in the stratosphere during solar maximum [Huang and Brasseur, 1993]. Higher stratospheric temperatures accompanying solar maximum are expected to lead to somewhat higher mean densities at higher altitudes [Schmidt et al., 2006], suggesting slower growth of the semidiurnal tidal components with altitude. However, it is not clear at this stage which of the various competing influences most strongly impacts semidiurnal tidal amplitudes at this time.
 In contrast to other studies, Pancheva et al.  report a positive correlation between the solar flux and semidiurnal tidal amplitudes for measurements from January 1989 to May 1993. Referring to our results in Figure 9, we note that both semidiurnal tidal components were decreasing from inferred maxima prior to our first monthly estimates in October 1990, based on the phases of these fits, while the solar flux decreased from approximately April 1991 until approximately October 1996. Thus, the large majority of the interval examined by Pancheva et al.  is also found to exhibit positive correlations in our data set. This suggests caution in the inference of long-term oscillations (or trends)employing data intervals less than the period of the oscillation being examined.
 The diurnal tide exhibits significant semiannual, quasi-biennial, quasi–4 year, and 11 year oscillations that increase in amplitude with altitude. Only the annual oscillation exhibits a decrease of amplitude with altitude. The amplitudes of the semiannual, quasi-biennial, and quasi–4 year oscillations are somewhat similar in the zonal and meridional components. However, the amplitude of the 11 year oscillation is twice as large in the meridional component as in the zonal component. In contrast to the semidiurnal tide, correlations of the diurnal tidal amplitude with the solar flux have not been reported previously. One possible reason is that long-term measurements have typically been made at middle and high latitudes [Baumgaertner et al., 2005; Namboothiri et al., 1993], where the semidiurnal tide is dominant [Forbes and Garrett, 1979; Hagan et al., 2001; Zhang et al., 2006]. Another reason is that it is difficult to examine the 11 year oscillation of the diurnal tide because of the comparable amplitude of the quasi-biennial oscillation [Fraser et al., 1989], as also seen in our analysis.
 Referring to the phase of the 11 year oscillation of the diurnal tidal amplitude in Figure 9, we see that the maximum occurred in October 1997 in the zonal component and in July 1999 in the meridional component. Thus, as noted above, the 11 year oscillation of the meridional component of the diurnal tide is in approximate quadrature with the solar cycle (with the tidal oscillation leading) whereas the zonal component of the diurnal tide is more nearly in antiphase with the solar cycle (with the tidal amplitude minimum lagging the solar cycle maximum by ∼1 year).
 The diurnal tide originates largely from H2O insolation absorption of the near infrared (IR) in the troposphere and lower stratosphere [Lieberman et al., 2007] and significant variations in tropospheric temperatures and water vapor are in phase with the solar cycle [Gleisner et al., 2005]. While the diurnal tide is also excited by O3 UV heating, the UV diurnal tide will suppress the IR diurnal tide due to the phase difference between them [Hagan, 1996; Hagan et al., 1999]. Thus, solar cycle minima and maxima exciting the IR and UV diurnal tides with different efficiencies and/or phases seem a plausible cause for the phase relations displayed in Figure 9.
 Finally, we note that an ∼5 year oscillation was observed in the semidiurnal tide at Halley, Antarctica [Hibbins et al., 2007], and in column ozone in the extratropics [Tung and Yang, 1994], potentially as a result of the modulation of the QBO [Mayr et al., 2000, 2007]. This is sufficiently close to our quasi–4 year oscillation to speculate that the modulations of the semidiurnal and diurnal tides in the MLT may be related to these other observations and another indication of possible links between the various periodicities noted in our study and other studies.
 We employed MF radar wind measurements at Kauai, Hawaii, over 16 years to assess long-period oscillations of mean winds and the semidiurnal and diurnal tidal amplitudes in the MLT. In addition to the well-known annual and semiannual oscillations, mean winds displayed terannual oscillations, especially in the zonal component. Mean winds did not exhibit significant longer-period oscillations than 12 months, except for weak responses at an 11 year period having no significant altitude variations.
 Semidiurnal tidal amplitudes exhibited long-period oscillations having primarily semiannual, annual, quasi-biennial, and 11 year periodicities. The observed phases of these oscillations were generally consistent with observations at other sites for various observing intervals. In particular, both components were seen to be in approximate quadrature with the solar cycle for reasons that are not understood at this time.
 Diurnal tidal amplitudes exhibited long-period oscillations having primarily semiannual, quasi-biennial, ∼5 year, and 11 year periodicities. Correlations with the solar cycle varied from approximate quadrature in the meridional component to more nearly an antiphase relation in the zonal component, perhaps as a result of the varying solar influences on tropospheric IR forcing and stratospheric UV forcing.
 This work has been supported by the National Science Foundation by grants OPP-0438777 and ATM-0634650 and NASA grant NNH05CC69C.