Mesospheric wind semidiurnal tides within the Canadian Middle Atmosphere Model Data Assimilation System

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116,, doi: /2011jd015966, 2011 Mesospheric wind semidiurnal tides within the Canadian Middle Atmosphere Model Data Assimilation System X. Xu, 1 A. H. Manson, 1 C. E. Meek, 1 C. Jacobi, 2 C. M. Hall, 3 and J. R. Drummond 4 Received 18 March 2011; revised 9 June 2011; accepted 13 June 2011; published 2 September [1] The horizontal wind data from the standard version of Canadian Middle Atmosphere Model Data Assimilation System (CMAM DAS) for the years are analyzed to obtain the global structure and seasonal variability of the semidiurnal tide (SDT) in the mesosphere. The modeled amplitudes and phases of the SDTs at single stations from middle/high northern latitudes are quite similar to those observed by radars. The primary nonmigrating tides identified in both the meridional wind and zonal wind semidiurnal spectra at 88 km include the westward propagating wave numbers s = 1 (SW1), 3 (SW3), 4 (SW4), 6 (SW6), the standing s = 0 (S0), and the eastward propagating s = 2 (SE2). The migrating SDT (SW2) amplitude maxima usually occur at 40 N 60 N during December February and August September, and also at 40 S 60 S in April May, with the dominance of (2, 4) during October April and (2, 3) and (2, 5) dominance for other months. The CMAM DAS is quite successful in reproducing the dominance of SW1 in the Antarctic summer mesosphere. The modeled SW1 shows very good overall agreement in both amplitude and phase with wind measurements from UARS High Resolution Doppler Imager and Wind Imaging Interferometer (UARS HRDI/WINDII) and from TIMED Doppler Interferometer (TIDI). The CMAM DAS analyses for SW3, SW4, SW6, and S0 are also in reasonable agreement with those determined from the HRDI/WINDII or TIDI wind measurements. This work provides further evidence for the tidal forcing from below. Citation: Xu, X., A. H. Manson, C. E. Meek, C. Jacobi, C. M. Hall, and J. R. Drummond (2011), Mesospheric wind semidiurnal tides within the Canadian Middle Atmosphere Model Data Assimilation System, J. Geophys. Res., 116,, doi: /2011jd Introduction [2] The tidal dynamics in the MLT (mesosphere and lower thermosphere) region usually can be investigated in three ways: ground based observations, space based measurements and model simulations. The three categories of data have their respective advantages and disadvantages. Single ground based observations are able to provide the tidal climatology at individual locations due to continuous coverage of the local time, but cannot resolve the zonal structure of the tide. Space based measurements offer good geographical coverage, although aliasing and sampling problems may pose a challenge for extracting tidal components on less than seasonal time scales. Models allow the identification of zonal structure and possible sources for the tidal variability, although successful modeling requires a complex interplay of radiation, chemistry and dynamics considering complicated 1 Institute of Space and Atmospheric Studies, University of Saskatchewan, Saskatoon, Saskatchewan, Canada. 2 Institute for Meteorology, University of Leipzig, Leipzig, Germany. 3 Tromsø Geophysical Observatory, University of Tromsø, Tromsø, Norway. 4 Department of Physics and Atmospheric Science, Dalhousie University, Halifax, Nova Scotia, Canada. Copyright 2011 by the American Geophysical Union /11/2011JD tidal structures and variability in the middle atmosphere. Therefore, to better understand the tidal behaviors the various data sets are also combined for use, especially for the identification of migrating and nonmigrating tides. [3] The semidiurnal tide (SDT) usually dominates above 40 latitude (maxima near 50 ) with winter and late summer early fall (LSEF) maxima in the Northern Hemisphere (NH) [Manson et al., 2002a, 2006, 2010]. Using winds from 6 radars located at similar latitudes, Jacobi et al. [1999] showed the dominance of the migrating SDT (wave number s =2) near 90 km at northern midlatitudes. This was also found to be true for a wide range of middle to high northern latitudes (40 N 70 N) and below 90 km [e.g., Riggin et al., 2003; Portnyagin et al., 2004; Manson et al., 2006, 2010]. Cierpik et al. [2003] identified the presence of the westward propagating nonmigrating s = 1 and s = 3 semidiurnal tides at 95 km for 50 N 55 N by the combination of UARS (Upper Atmosphere Research Satellite) and multiradar winds. Manson et al. [2009, also Arctic tidal characteristics at CANDAC PEARL (80 N, 86 W) and Svalbard (78 N, 16 E) for : Observations and comparisons with the model CMAM DAS, submitted to Annals of Geophysics, 2011] provided evidence for the presence of the s = 1 SDT at 78 N 80 N in the boreal winter and spring. The nonmigrating semidiurnal tides are evidently stronger in the Southern Hemisphere (SH) than in the NH during their respective summer months. In their Introduction, Xu et al. [2009a] listed a series of papers 1of20

2 [Hernandez et al., 1993; Forbes et al., 1995; Portnyagin et al., 1998, 2000; Riggin et al., 1999; Baumgaertner et al., 2006; Murphy et al., 2006] wherein the dominance of s = 1 SDT was observed at the South pole and high southern latitudes in austral summer. Angelats i Coll and Forbes [2002] employed the UARS data with a spectral model to provide evidence for the nonlinear interactions between the semidiurnal migrating tide and the stationary planetary wave (SPW) as a possible source for the nonmigrating semidiurnal tides occurred in the SH. [4] With the launch of satellite missions like UARS and TIMED (Thermosphere Ionosphere Mesosphere Energetics and Dynamics satellite) that have high inclination and slowprecession orbits, remote sensing instruments onboard these satellites have made possible the global analyses of wind and temperature tides. The HRDI (High Resolution Doppler Imager) [e.g., Burrage et al., 1996; Manson et al., 2002c, 2004] and WINDII (Wind Imaging Interferometer) [e.g., Forbes et al., 2003] systems of UARS, and the TIDI (TIMED Doppler interferometer) [e.g., Oberheide et al., 2006, 2007; Iimura et al., 2009, 2010] have been dominant in measuring the global characteristics and zonal wave structures for the tidal winds; while the MLS (Microwave Limb Sounder)/ UARS and SABER (Sounding of the Atmosphere using Broadband Emission Radiometry)/TIMED measurements have yielded a series of analyses for the global tidal temperature field [e.g., Forbes and Wu, 2006; Forbes et al., 2006]. [5] Middle atmosphere general circulation models (GCMs) have provided new possibilities for the study of the tides in the middle atmosphere and related chemical and dynamical coupling processes [e.g., Miyahara and Miyoshi, 1997; McLandress, 1997, 2002a, 2002b; Hagan and Roble, 2001; Ward et al., 2005]. Herein, our focus is on the semidiurnal tides from CMAM (Canadian Middle Atmosphere Model). The remainder of this section describes the relevant work on this topic to date. The CMAM is a three dimensional chemistry climate GCM extending from the ground to the MLT, which means that the tidal and planetary waves are naturally forced by the model [Beagley et al., 1997; de Grandpré et al., 2000; Scinocca et al., 2008]. Manson et al. [2002a, 2002b] used the seasonal outputs from CMAM without interactive chemistry to analyze the gravity wave (GW) effects and tidal characteristics, which were compared with multiyear medium frequency (MF) radar observations between 2 N 70 N. The SDT amplitudes from CMAM without interactive chemistry were generally similar to those observed; while the SDT phases were in good agreement for CMAM and MF radar with longer/shorter vertical wavelengths generally at lower/higher latitudes. However, one significant weakness of these comparison studies was the absence of longitudinal dependence for the model tides. Manson et al. [2006] demonstrated that CMAM with interactive chemistry provided the main characteristics of the radar observed SDT in the NH, and quite detailed agreement with observations for the SDT from middle latitudes. However, the CMAM output values were saved and made available only for the locations of several MF radars, which did not allow for the zonal wave number decomposition. Du et al. [2007] presented the migrating and several nonmigrating semidiurnal tides extracted from the 3 h archived data from the extended CMAM which can reach an approximate height of 210 km [Beagley et al., 2000]. Their results showed that there was overall agreement in latitudinal structure of the nonmigrating tidal amplitude between the extended CMAM and TIDI measurements, with better agreement for the westward propagating s = 4 nonmigrating SDT. [6] A 3D Var data assimilation scheme has been coupled to CMAM [Polavarapu et al., 2005]. In the standard version of the CMAM Data Assimilation System (CMAM DAS), standard meteorological observations from radiosondes, aircraft, surface measurements, and satellites are assimilated below 1 hpa. Between 10 and 1 hpa, the only measurements are temperatures from the Advanced Microwave Sounding Unit A (AMSU A) instrument. No measurements are assimilated above 1 hpa. Vertical correlations are constructed to ensure negligible analysis increments above the stratopause. Therefore, the mesospheric response is entirely due to vertical connections in internal model dynamics. This provides an opportunity to examine whether the analysis increments applied in the troposphere and stratosphere can modulate the mesospheric tidal variability. Xu et al. [2011] compared the mesospheric mean winds from the CMAM DAS analyses with radar measurements, demonstrating that the CMAM when constrained by tropospheric and stratospheric observations can provide a realistic description of mesospheric background wind variability at middle and high northern latitudes, especially in winter. The effects on the mesospheric mean winds and tides of the assimilation of tropospheric and stratospheric observations are expected to be different due to different sources of variability for the mesospheric mean winds and tides. Therefore, it will be also interesting to examine the consistency between the CMAM DAS tides and observations. [7] In the present study, the semidiurnal tides within the CMAM DAS will be discussed in detail based upon a multiyear data set ( ). In section 2, the model and radar data used are briefly described. The consistency between the CMAM DAS analyses and radar measurements for the semidiurnal tides at single stations from middle and high northern latitudes is explored in section 3. The migrating and nonmigrating tidal analyses are provided in section 4 where consistency between CMAM DAS and satellite measurements is also demonstrated. In section 5, we discuss the Hough modes of the migrating semidiurnal tide and possible contributions from the refractive effects. This is followed by a comparison of variability of the Antarctic westward propagating s = 1 nonmigrating tidal amplitude with the Arctic stationary planetary wave number 1 (SPW1). Finally, a summary of results and conclusions are provided. 2. Data Description [8] As in work by Xu et al. [2011, also Mesospheric wind diurnal tides within the Canadian Middle Atmosphere Model Data Assimilation System, submitted to Journal of Atmospheric and Solar Terrestrial Physics, 2011], the hourly spaced data from the standard version of CMAM DAS (Canadian Middle Atmosphere Model Data Assimilation System) run are used in this study. The CMAM DAS is based on the CMAM, a spectral model with T47 truncation and 71 levels extending from the ground to the lower thermosphere ( hpa), coupled with a 3D Var assimilation scheme [Polavarapu et al., 2005]. In this system, standard meteorological observations from radiosondes, aircraft, surface 2of20

3 measurements, and satellites are used below 1 hpa. Between 10 and 1 hpa, the only measurements are temperatures from the AMSU (Advanced Microwave Sounding Unit) A instrument. No measurements are assimilated above 1 hpa. Vertical correlations are constructed to ensure negligible analysis increments above the stratopause. A digital filter is applied to reduce the spurious waves produced due to inserting the analysis updates. A detailed description of this version of CMAM DAS was provided by Polavarapu et al. [2005]. [9] The observed radar tides for this study are taken from medium frequency (MF) radars located at Tromsø (70 N, 19 E) and Saskatoon (52 N, 253 E), and a meteor wind radar at Collm (51 N, 13 E). A common analysis [Manson et al., 1999] has been used at each radar location to obtain the tidal amplitudes and phases. First, the mean hourly winds were formed for a given temporal window: calendar months are used here. A superposition of a mean wind and the diurnal and semidiurnal tides was then fitted to the time sequences of the mean hourly winds at 24 local hours. It is required that for each least squares fit there be data for 16 or more local hours out of 24 for each height over the window. The same method was also used on the location specific CMAM DAS hourly data to obtain the modeled tides for the radar locations, thus allowing a direct comparison with radar measurements. To compare with radar observations, the model data are interpolated into altitude surfaces using the geopotential height field. The highest altitude, which allows full global coverage for the tidal winds and wave number spectra, is 88 km for this version of CMAM DAS, which is different from the extended CMAM version that can reach an approximate height of 210 km [Beagley et al., 2000; McLandress, 2002a, 2002b; Ward et al., 2005; Du et al., 2007]. At this time, the DAS is not available for the latter. 3. Semidiurnal Tides at Single Stations [10] The SDT is usually dominant at latitudes higher than 40, so here the comparison between the CMAM DAS and radar observed SDTs will be emphasized for middle to high latitudes. We provide in Figure 1 the monthly amplitudes and phases of semidiurnal winds from CMAM DAS and radar observations at 88 km and 82 km above Tromsø, Saskatoon, and Collm. For each month, a 3 year ( ) average is shown for both data sets. The seasonal variations of the SDT amplitudes at Tromsø (Figure 1a) are quite similar for the CMAM DAS and radar results, except in spring and early summer months where the modeled (observed) values show local maxima (minima). The observed late summer and early fall (LSEF) SDT wind amplitude maximum is very well captured by the model. The CMAM DAS and radar SDT phases are usually in very good agreement (within 1 h), although the phase differences are sometimes slightly larger (within 2 3 h), for example, in spring and early summer when the amplitude differences are evident, or in March and November when the amplitudes are very small. The 3 h phase differences between the zonal wind and meridional wind are quite clear for both data sets, indicating that the SDT is generally circular at this location. [11] For Saskatoon (Figure 1b), the modeled and observed seasonal features of the SDT amplitudes are generally similar, with the exception of March and November where the modeled amplitude variations are quite different from observed. The LSEF SDT amplitude maximum is clearly provided by both the observations and modeling, although the modeled amplitudes are generally larger than the observations (by about 10 m/s for 82 km). The modeled and observed amplitude values are quite close to each other in summer and fall at 88 km, but the former are much larger in March and November. The phase agreement of the CMAM DAS with the observations is generally excellent but with possible large differences in the regions of small radar amplitudes (e.g., April and November). As observed or modeled, the phases are internally consistent between the two wind components. Turning next to Collm (Figure 1c), both radar and model show amplitude maxima in LSEF and winter, but the observed winter maximum is evidently larger than the modeled at 88 km. In addition, the radar observed a May peak in the zonal wind SDT amplitude over Collm while the modeled amplitude exhibits a local peak in June (Figure 1c). The phase agreement between the CMAM DAS and radar is again excellent at Collm, except when the amplitudes are very small. As expected, the phases of the two wind components are in quadrature for both data sets. [12] In order to examine the tidal vertical variations, Figure 2 provides contour plots for the meridional wind SDT amplitudes and phases at Saskatoon, obtained from the 3 year averages ( ) of MF radar measurements and CMAM DAS. The radar and CMAM DAS contour plots (3 year means) of Figure 2 are generally similar to those for year 2006 [Manson et al., 2010, Figure 2]. The seasonal structures of the SDT amplitude contours from CMAM DAS are broadly similar to those observed: both radar and CMAM DAS show the winter and LSEF tidal wind amplitude maxima, the model values reaching to lower heights ( 75 km). The SDT phase values, with respect to both height and month, are remarkably similar for radar and CMAM DAS. The vertical wavelength of the SDT is usually short in winter (monthly means km) and longer or evanescent in summer. The CMAM DAS provides stronger summer (June August) vertical phase gradients, i.e., smaller vertical wavelengths of 35 km, below circa 76 km. The SDT amplitudes from radar and the CMAM DAS also show some differences: at Tromsø (Figure 1a) the observed LSEF amplitude maxima occur at the upper heights, with the modeled maximum peaking at 82 km and having 3 months duration; while at Saskatoon the modeled amplitude maxima are stronger (Figure 2). The positions of the winter amplitude maxima also differ at Saskatoon: the observed feature extends from December March, while the modeled is broader and extends from November to April. In the Arctic (80 N), differences between CMAM DAS and radar observations are even more marked (Manson et al., submitted manuscript, 2011). 4. Migrating and Nonmigrating Semidiurnal Tides 4.1. Spectral Analysis of Zonal Wave Number Structure [13] The CMAM DAS offers a global coverage (up to 88 km), which allows the zonal wave number decompositions. First, at each altitude, latitude and longitude, a least squares fit was performed to obtain the tidal amplitudes and phases. For each altitude and latitude, a Fourier transform was 3of20

4 Figure 1. Monthly mean (top) amplitudes and (bottom) phases (local time) of the semidiurnal (solid lines) meridional and zonal (hatched lines) winds at (left) 88 km and (right) 82 km above (a) Tromsø (70 N, 19 E), (b) Saskatoon (52 N, 253 E), and (c) Collm (51 N, 13 E), from the 3 year ( ) of CMAM DAS data (blue lines) and radar observations (red lines). The error bars for both model and radar correspond to the standard deviations (or root mean square errors) for vector averaged values over the 3 years. then applied to the longitudinal sequence of complex tidal values to perform the zonal wave number decomposition for s = 7 tos = +7. Note that the frequency and zonal wave number fields also can be obtained simultaneously by a twodimensional Fourier transform analysis in time and longitude over the model outputs at each altitude and latitude, with no change in the results. Here, to make it easier to distinguish various tidal components, following Forbes and Wu [2006], we utilize the notation SWs (SEs) to denote a westward (eastward) propagating semidiurnal tide, with zonal wave number s. The standing (wave number s = 0) semidiurnal tide is denoted as S0. The stationary planetary wave (SPW) with zonal wave number m is expressed as SPWm. [14] Figure 3 provides latitude wave number contour plots of the SDT amplitudes within CMAM DAS at 88 km for each month (3 year averages). Positive (negative) wave number represents westward (eastward) propagation. The annual mean spectral analysis of the zonal wave number structure is 4of20

5 Figure 2. Height seasonal contour plots of (left) the amplitudes and (right) phases (local time) of the meridional wind semidiurnal tides at Saskatoon (52 N, 253 E), obtained from the 3 year (years 2006 to 2008) averages of (top) MF radar observations and (bottom) CMAM DAS. provided in Figure 4 where the migrating tidal component is excluded. Compared to the diurnal tide (Xu et al., submitted manuscript, 2011, Figures 5 and 6), there are a larger number of weak spectral features for the SDT. Such was also the case for the UARS HRDI wind spectra [Manson et al., 2004]. There is no doubt that the migrating component (SW2) is significant at any time for any latitude (Figure 3). The primary nonmigrating SDTs, which are identified in the 88 km spectra (Figures 3 and 4) include s = 1 (SW1), 0 (S0), 2 (SE2), 4 (SW4), 3 (SW3), and 6 (SW6) for both wind components. The amplitudes for SE2, S0, and SW1, relative to those for SW3, SW4, and SW6, are larger overall. All these tides exhibit their individual latitudinal and seasonal variations (Figure 3), which will also be discussed later. The presence of these nonmigrating SDTs is consistent with two mechanisms: nonlinear interactions between the migrating tide and SPWm [Angelats i Coll and Forbes, 2002], and zonally asymmetric thermal forcing [Hagan and Forbes, 2003]. A nonlinear process of SW2 with SPWm (mainly wave numbers m = 1 and 2) thus would force the SW1 and SW3 pair, and the S0 and SW4 pair, while modulation of SW2 solar energy absorption, which includes such processes as deep convection and latent heating, by the dominant topographic structures at low latitudes (mainly structure numbers S = 4 and 1) would lead to the SE2 and SW6 pair, and again the SW1 and SW3 pair. Note that these tides, depending on their vertical wavelengths, may or may not reach the upper atmosphere with significant amplitudes. The possible factors for the presence of the zonal wind SE6 and SE3 in Figures 3 and 4 are unknown at this moment, moreover they were little reported by the satellite measurements, so no further analysis will be attempted for them. The spectra for the 96 km HRDI winds [Manson et al., 2004] indicated the existence of SE2, S0, and SW4; while the primary nonmigrating components revealed in the TIDI 95 km meridional wind semidiurnal spectrum for a 60 day window centered on 15 February 2004 include SW3, SW1, SE2, S0, and SW4 [Oberheide et al., 2007, Figure 1] Latitudinal and Seasonal Structures [15] Latitude versus month contours at 88 km of the meridional wind and zonal wind amplitudes of SW2, SW1, S0, SE2, SW3, SW4, and SW6 are now provided in Figures 5 and 6. The meridional wind SW2 (Figure 5a) usually has four amplitude maxima along latitudinal direction: at latitudes poleward of ±40, as well as at tropics (±5 25 ). The corresponding phases poleward of ±40 appear to be dominantly symmetrical about the equator in May to September, and antisymmetrical for other months (not shown). In contrast to the meridional wind, the zonal wind SW2 (Figure 5b) amplitude maxima occur mainly at high midlatitudes (±40 60 ), with clear seasonal variations. The zonal wind SW2 phase differences between the two hemispheres (poleward of ±40 ) are usually small (0 3 h) in October to April, and larger (3 6 h) for the rest of the year (not shown). At 88 km the zonal wind SW2 also has a lesser maximum around the equator mainly in January to March, with usually little phase change across the equator. For the SW2 amplitude at 88 km (Figures 5a and 5b), both horizontal wind components show similar seasonal variations: the major amplitude maxima (at middle latitudes) occur in the local winter (centered in February) and late summer early fall (August September) seasons for the NH, and are broader in the SH, centered in austral autumn (April May); the secondary amplitude maxima (at low latitudes) occur mainly in January to March for both hemispheres. The extended CMAM free run [Du et al., 2007] provided similar seasonal and latitudinal structures of the SW2 amplitude and phase, although with the amplitude maxima centered at higher heights (in the thermosphere) in the extended CMAM. [16] From inspection of Figures 3, 5c, and 5d, one of the outstanding features is the dominance of SW1 (larger than SW2) in Antarctic and high latitudes in the local summer, which is consistent with the earlier observational results [Hernandez et al., 1993; Forbes et al., 1995]. Maximum 5of20

6 Figure 3. Amplitude of the semidiurnal tide at 88 km for both the meridional (V SDT) and zonal (U SDT) wind components as a function of latitude and zonal wave number (positive for westward propagation) for each month, obtained from the 3 years ( ) of CMAM DAS data. 6of20

7 Figure 4. Annual mean amplitude of the semidiurnal tide at 88 km for both the (top) meridional and (bottom) zonal wind components as a function of latitude and zonal wave number (positive for westward propagation), obtained from the 3 years ( ) of CMAM DAS data. The migrating tidal component (s = 2) is excluded in order to highlight the nonmigrating components. monthly mean amplitude values near the South Pole are of order 18 m/s. In contrast, the NH SW1 is much weaker (within 3 m/s for monthly mean) throughout all seasons (Figures 5c and 5d). SW1 is usually nonsymmetric about the equator, except for approximately antisymmetric structures in phase for the meridional wind during the equinoxes (not shown). For S0 (Figures 5e and 5f), the dominant amplitude maxima occur in the local summer seasons in the polar regions of both hemispheres; this tide also evidences significant amplitudes at middle/high latitudes (40 70 ), e.g., in the NH (February to April) and in the SH (April to June). The corresponding phases do not show hemispheric symmetry in the occurrence. The SE2 tide (Figures 5g and 5h) generally occurs in hemispheres during their respective summer months, maximizing at middle latitudes ( 40 ) for the zonal wind component, and at lower latitudes ( 20 ) for the meridional wind, although the NH maximum is stronger than the SH counterpart. Correspondingly, the phases are usually available only in one hemisphere, resulting in a lack of information on symmetry/antisymmetry about the equator. [17] The SW3, SW4, and SW6 tides (Figure 6) are usually weaker, with monthly amplitudes in the range of 0 6 m/s between 70 and 88 km. Large SW3 (Figures 6a and 6b) usually occurs at 30 S 60 S during April October and at 30 N 60 N in the autumnal equinox (October), with maximum values of order 6 m/s. The 100 km simulations from the GSWM model alone [Oberheide et al., 2007, Figure 8] provided the presence of SW3 in the SH during March to October; while TIME GCM alone yielded SW3 at middle latitudes in both hemispheres around September October [Oberheide et al., 2007, Figure 8]. The only tidal forcing for the nonmigrating tides is latent heat release due to deep tropical convection in the GSWM [Hagan and Forbes, 2003], while the TIME GCM nonmigrating tides are driven primarily by the interaction between the migrating tide and stationary planetary waves [Hagan and Roble, 2001]. This may suggest that the SH SW3 is driven mainly by latent heat release due to raindrop formation, while the SW3 amplitude symmetry about the equator around the equinox is due largely to the SPW1 SW2 interaction. The combined effect leads to the results [Oberheide et al., 2007, Figure 8] which are very similar to those shown in Figures 6a and 6b. For SW4 at 88 km (Figures 6c and 6d), large amplitudes occur between 30 N and 60 N in October March and between 20 S and 50 S in April September, showing a clear annual variation. The SW4 around 105 km from GSWM alone showed similar features [Oberheide et al., 2007, Figure 7], which may imply that forcing by latent heat release is more important than the wave wave nonlinear interaction for the presence of SW4 within the CMAM DAS. The SW6 amplitude (Figures 6e and 6f) shows similar latitudinal and seasonal distributions, but with somewhat weaker values as compared to SW4. The nonmigrating semidiurnal tides as presented above usually do not appear symmetric in both hemispheres, except for the equinoxes. Because of this, Hough mode assignment is difficult or impossible for these tides. In this study, the Hough mode decomposition is not performed for the nonmigrating tides. The Hough modes for SW2 will be shown later. [18] Note that due to the longitudinal variation of the nonmigrating tidal phase the tidal amplitude modulation in the presence of nonmigrating tides is different for stations at different longitudes. For 88 km and October, for example, the SDT amplitude was 20 m/s at Saskatoon (Figure 1b) and 10 m/s at Collm (Figure 1c), while the wave numbers dominating the same latitude/month/height regime are SW2, SW1, SW3, and SW4 (Figure 3). The amplitudes in October for SW2, SW1, SW3, and SW4 at 88 km are 9 m/s (Figure 5a), 4 m/s (Figure 5c), 5 m/s (Figure 6a), and 4 m/s (Figure 6c), respectively. The Saskatoon SDT amplitude is approximately equal to (SW2 + SW1 + SW3 + SW4) while the Collm tide amplitude is close to the SW2 amplitude value. This indicates that the nonmigrating tides are all in phase with the local migrating tide at Saskatoon, leading to a larger amplitude, while at Collm they tend to cancel each other, or even sum up to an oscillation out of phase with the local migrating tide, resulting in a smaller amplitude. Such a longitudinal difference in the tidal amplitude modulation is associated with the longitude spacing between the two radars (120 ) Comparison With Satellite Measurements [19] Angelats i Coll and Forbes [2002] provided monthly mean meridional wind amplitudes and phases of SW2, SW1, and SW3 at 95 km based upon a multiyear data set ( ) of wind measurements from HRDI/WINDII on UARS; while Oberheide et al. [2007] presented monthly mean amplitudes and phases for the primary nonmigrating semidiurnal tidal components (SW1, SW3, and SW4) in the range of km and between 45 S and 45 N, derived 7of20

8 Figure 5. Latitude month contours of the (a, c, e, and g) meridional and (b, d, f, and h) zonal wind amplitudes (m/s) at 88 km for SW2 (Figures 5a and 5b), SW1 (Figures 5c and 5d), S0 (Figures 5e and 5f), and SE2 (Figures 5g and 5h), obtained from the 3 years ( ) of CMAM DAS data. 8of20

9 Figure 6. Latitude month contours of the (a, c, and e) meridional and (b,d, and f) zonal wind amplitudes (m/s)at88kmforsw3(figures6aand6b),sw4(figures6cand6d),andsw6(figures6eand6f), obtained from the 3 years ( ) of CMAM DAS data. from TIDI measurements during The two sets of measurements provided similar climatologies of monthly mean amplitudes of SW1 and SW3 at 95 km in low and middle latitudes [Oberheide et al., 2007]. More recently, Iimura et al. [2009, 2010] reported the TIDI ( ) nonmigrating semidiurnal tides (SW1, S0, and SW3) in the MLT over the Antarctic and Arctic. These observation results provide opportunity for the validation of the modeled migrating and nonmigrating tides Comparison With UARS HRDI/WINDII [20] Note that the tides derived from the combined HRDI/ WINDII measurements are available only at the single height of 95 km, where day and night data are available for both the HRDI and WINDII instruments [Angelats i Coll and Forbes, 2002]. The standard version of CMAM DAS used here cannot go as high as 95 km. Hence, here we compare the CMAM DAS tides at 88 km, which is the highest geopotential height for a full global coverage in the model, with those determined from the UARS HRDI/WINDII wind measurements at 95 km. We acknowledge that this height difference may add difficulty to comparison. However, this comparison is valuable since both data sets (88 km CMAM DAS and 95 km UARS) reflect the tidal variability in the upper mesosphere and lower thermosphere. A comparison for the latitudinal structures of annual mean amplitudes for s = 3 and 2 diurnal tides (Xu et al., submitted manuscript, 2011) demonstrated good overall agreement between the CMAM DAS 88 km analyses and the UARS and TIDI measurements at 95 km. [21] Figure 7 shows latitudinal variations of the meridional wind SW2 amplitudes from the CMAM DAS analyses at 88 km and UARS HRDI/WINDII 95 km measurements, taken from Angelats i Coll and Forbes [2002], for January, February, April and September. Here the model results are 9of20

10 The exception is large discrepancies in the amplitude values at middle and high northern latitudes for January and September. In addition, in September the modeled and observed latitudinal structures of SW2 amplitudes are different near the equator with the model showing an additional amplitude minimum. At this time, it is hard to assess quantitatively what caused these differences. Possible candidates calculated from the 60 day intervals (centered at the 15th of each month) to be consistent with the UARS measurements. The CMAM DAS 88 km analyses show that the amplitude maxima occur at 40 S 60 S in April and at 40 N 60 N in September. The HRDI/WINDII 95 km measurements provide similar behaviors. Overall, the modeled meridional wind SW2 amplitudes at 88 km are similar to the 95 km observations for these months, with better agreement in the SH. Figure 7. Comparison between the meridional wind amplitudes of the migrating semidiurnal tides (SW2) from the CMAM DAS analyses at 88 km ( , solid lines) and UARS HRDI/WINDII 95 km measurements ( , triangles, redrawn from Angelats i Coll and Forbes [2002]) for (top to bottom) January, February, April, and September. Months indicate the centers of the 60 day analysis windows. The model error bars (shown every 7.5 in latitude for clarity) indicate the standard deviations over the 3 years, and the HRDI/WINDII error bars are unknown. Figure 8. Meridional wind amplitude of the westward propagating nonmigrating s = 1 semidiurnal tide (SW1) from the CMAM DAS analyses at 88 km ( , solid lines) and UARS HRDI/WINDII 95 km measurements ( , triangles, redrawn from Angelats i Coll and Forbes [2002]) for (top to bottom) January, April, July, and October. Months indicate the centers of the 60 day analysis windows. The model error bars (shown every 7.5 in latitude for clarity) indicate the standard deviations over the 3 years, and the HRDI/WINDII error bars are unknown. 10 of 20

11 Figure 9. Same as Figure 8 but for the westward propagating s = 3 semidiurnal tide (SW3). include the model bias, the height difference (95 km versus 88 km), and the time lag between two data sets ( versus ). [22] Figure 8 compares the meridional wind SW1 amplitudes at 88 km from the CMAM DAS analyses ( ) with those at 95 km from UARS HRDI/WINDII ( ), taken from Angelats i Coll and Forbes [2002], for January, April, July, and October. The model tides are obtained from the 60 day intervals to be consistent with UARS. Good overall agreement between the CMAM DAS 88 km analyses and HRDI/WINDII 95 measurements can be seen, especially for April and October. The model yields larger amplitudes at 88 km than the 95 km observations at high southern latitudes during austral summer months, which may indicate that the CMAM DAS overestimates the magnitude of the Antarctic SW1 tide in the local summer. This will also be seen in a later comparison with the TIDI results at high latitudes. The counterpart of Figure 8 for the meridional wind SW3 is provided in Figure 9. As a whole, the modeled latitudinal structure at 88 km of the SW3 amplitude is very similar to observed from HRDI/WINDII at 95 km. The agreement is quite remarkable in the SH, with both the model and measurements providing the amplitude peaks occurring at 30 S 60 S for all the four months shown. In January, the HRDI/WINDII amplitudes at 95 km are much stronger than modeled at 88 km for SW3 poleward of 60 N. For October, the HRDI/WINDII 95 km measurements are in good overall agreement with the CMAM DAS 88 km analyses with regards to the latitudinal structure of the SW3 amplitude except that the model provides larger amplitudes at 30 N 60 N Comparison With TIDI [23] In Figure 10, we present climatologies of monthly mean meridional wind amplitudes for the nonmigrating semidiurnal tidal components SW1, SW3, SW4, SW6, S0, and SE2 at circa 88 km between 45 S and 45 N, from the CMAM DAS and the TIDI measurements. For both data sets, the tides are calculated from the 60 day intervals (centered at the 15th of each month). The TIDI results are provided by J. Oberheide and are derived from measurements during For TIDI, the migrating tides are usually removed in the process of tidal analysis and are therefore not available [Oberheide et al., 2006, 2007]. [24] Overall, the CMAM DAS is in agreement with TIDI for the meridional SW1, SW3, SW4, and SW6 amplitudes between 45 S and 45 N. A general consistency between the model and TIDI is found for the latitudinal structure of the meridional SW1 amplitude between 45 S and 45 N during October May (Figures 10a and 10b). The 45 S 45 N SW1 amplitudes are in the range of 0 6 m/s at 88 km for both data sets. For the meridional wind SW3 amplitude, the NH October maximum and the broad SH maxima between April July as observed by TIDI (Figure 10d) are generally reflected within the CMAM DAS (Figure 10c). Regarding the meridional wind SW4 (Figures 10e and 10f), the CMAM DAS and TIDI generally yield similar seasonal amplitude variations around 40 N with large amplitudes during October March. For the SH, the model provides a broad maxima of SW4 amplitude during April September, while the TIDI maximum amplitudes occur around 40 S during the equinoxes. Both the modeled and TIDI SW4 amplitudes are in the range of 0 6 m/s. Both data sets show that SW6 is relatively weak. The modeled SW6 amplitude maximum is of order 2 4 m/s at 88 km, which is consistent with TIDI (Figures 10g and 10h). In contrast, the modeled S0 and SE2 amplitudes between 45 S and 45 N are quite different from those derived from TIDI. For example, the modeled S0 amplitude maxima at N around March and at S in the local summer (Figure 10i) are not present in the TIDI measurements (Figure 10j). The NH summer maxima in the SE2 amplitude within the CMAM DAS (Figure 10k) are not evident in the TIDI analysis (Figure 10l). The modeled SE2 maximum amplitudes are of order m/s, in comparison with the TIDI amplitudes in the range of 0 4 m/s. [25] To further explore the consistency between the CMAM DAS and TIDI measurements at high latitudes, the TIDI nonmigrating tidal analyses over the Arctic and Antarctica reported by Iimura et al. [2009, 2010] are used here. In Figures 11 and 12, the CMAM DAS analyses of the 11 of 20

12 Figure 10. Latitude month contours of the meridional wind amplitudes (m/s) for (a and b) SW1, (c and d) SW3, (e and f) SW4, (g and h) SW6, (i and j) S0, and (k and l) SE2 from the CMAM DAS analyses ( ) at 88 km (Figures 10a, 10c, 10e, 10g, 10i, and 10k) and the TIDI measurements ( ) at 87.5 km (Figures 10b, 10d, 10f, 10h, 10j, and 10l). The latter are provided by J. Oberheide. Months indicate the centers of the 60 day analysis windows for both data sets. 12 of 20

13 Figure 11. Comparison between the SW1 amplitude profiles from the CMAM DAS analyses ( , circles) and TIDI measurements ( , squares, redrawn from Iimura et al. [2009, 2010]) for (a, b, c, and d) the meridional wind and (e, f, g, and h) zonal wind at 87 N for April and June (Figures 11a, 11b, 11e, and 11f) and at 87 S for October and December (Figures 11c, 11d, 11g, and 11h). Months indicate the centers of the 60 day analysis windows. The model error bars indicate the standard deviations over the 3 years. The TIDI error bars represent 1 standard deviation from the mean values (1 sigma confidence intervals on the estimated amplitudes). amplitude and phase of SW1 at 87 N and 87 S are compared with the TIDI results. Here the model results are calculated for the 60 day intervals labeled by the centered months in order to be consistent with the TIDI analyses by Iimura et al. [2009, 2010]. Note that CMAM does not utilize a polar filter and the modeled tides at high latitudes are therefore not affected (S. R. Beagley, personal communication, 2011). The model/ TIDI agreement is very good for both the meridional wind and zonal wind amplitudes of SW1 at 87 N during the two 60 day intervals shown (Figures 11a, 11b, 11e, and 11f). For the 87 S SW1 tide in October (Figures 11c and 11g), the TIDI amplitude maximum occurs at about 87 km, while the modeled maximum is located at about 5 km lower. The TIDI amplitude maximum is much larger than modeled for the meridional wind 87 S SW1 tide in October, but with improved agreement in magnitude for the zonal wind amplitude. The modeled SW1 tide near the South Pole is larger than the TIDI value observed in December (Figures 11d and 11h), which is similar to the comparison between the model and HRDI/ WINDII in January (Figure 8), but the vertical structures of the SW1 amplitudes from the model/tidi are in very good agreement (Figures 11d and 11h). Similar to the model/radar phase comparison (Figure 1), Figure 12 shows that there is remarkably good agreement between the CMAM DAS and TIDI regarding the SW1 phases in the polar latitudes. Both data sets evidently show that the meridional phase is leading the zonal phase by 3 h. [26] Iimura et al. [2010] showed that the TIDI measured Arctic S0 amplitude maxima occurred primarily in the yaw periods (60 day windows) centered on 15 February and 15 April. This is similar to the CMAM DAS analyses (Figures 5e and 5f), although the modeled Arctic summer maxima (Figures 5e and 5f) are not evident in the TIDI measurements. For Antarctica, the modeled S0 shows 13 of 20

14 Figure 12. Same as Figure 11 but for the SW1 phase (UT of maximum at 0 longitude). amplitude maxima in the local summer season (Figures 5e and 5f). Similarly, the TIDI measurements [Iimura et al., 2009] also revealed the Antarctic S0 amplitude maxima in the 15 December and 15 February yaw intervals. However, the TIDI observed Antarctic maxima in the yaw period centered on 15 August [Iimura et al., 2009] are not present in the CMAM DAS analyses. 5. Hough Modes of SW2 and Refractive Effects [27] The late summer/early fall (LSEF) amplitude maximum of the semidiurnal wind is a dominant annual feature in the mesosphere (75 95 km) for middle/high northern latitudes and is attributed largely to the migrating tides [Manson et al., 2010, and references therein]. Very good agreement between the amplitudes of the composite SDT at Saskatoon (Figure 2) and the migrating SDT at similar latitudes (not shown) confirmed the dominance of the migrating SDT at northern middle/high latitudes in LSEF and winter. This is also coincident with the estimation using multistation observations at similar latitudes [e.g., Jacobi et al., 1999; Manson et al., 2006; Xu et al., 2009a] and the HRDI/WINDII measurements (Figure 7). [28] The Hough mode decomposition is now performed for SW2. In practice, SW2 can be adequately represented by a sum of the first several modes since higher modes are typically very weak. Hence, only the first two symmetric and first two antisymmetric modes are considered for the fitting and their amplitudes are provided in Figure 13. Consistent with the latitudinal structures (not shown), Figure 13 shows that the mesospheric migrating SDT (SW2) is usually dominated by symmetric Hough mode (2, 4) during October to April and by antisymmetric modes ((2, 3) and (2, 5)) for other months. The latitudinal structure of the HRDI wind migrating SDT at 96 km [Manson et al., 2004] also suggested the dominance of symmetric modes in the boreal winter and the dominance of antisymmetric modes in the boreal summer. In the mesosphere, the mode (2, 2) is weaker than the higher order modes. This is not surprising since the (2, 2) easily becomes evanescent when it propagates upward from the excitation altitude [e.g., Forbes, 1995; Yuan et al., 2008]. The antisymmetric modes are often minimal at the equinoxes 14 of 20

15 Figure 13. Altitude versus time contours of the amplitudes (m/s) of the first two (left) symmetric and (right) antisymmetric Hough modes for the meridional (V12) and zonal (U12) wind migrating semidiurnal (SW2) tides, obtained from the 3 year averages ( ) of CMAM DAS. The amplitudes 3 m/s are shaded. (Figure 13), in agreement with established knowledge. The seasonal variability of mode (2, 5) (Figure 13) is very similar to that of SW2 from middle/high northern latitudes (not shown), with amplitude maxima occurring in the boreal winter and LSEF. Note that higher order modes usually result from mode coupling due to interactions of the fundamental modes with the zonal mean wind [Forbes, 1982]. [29] Riggin et al. [2003] suggested refraction effects due to changes in vertical profiles of temperature and zonal wind as an important factor resulting in the observed autumn amplitude maximum of the SDT at middle and high northern latitudes. Their evidence was that an enhancement in the vertical wave number (i.e., the shortening of the vertical scale) was also found around the observed autumnal SDT amplitude maximum. Here a similar method is applied to estimate the refractive effects based on the modeled wind and temperatures near Saskatoon (52 N, 253 E). The basic equations [Hines, 1974; Forbes and Vincent, 1989; Riggin et al., 2003] for the calculation of the vertical wave number m can be expressed as m 2 ¼ N 2 gh n 1 4H 2 ð1þ h n ¼ h n 1 þ U 4 ð2þ C 0 sin N 2 ¼ g T dt dz þ G d where N is the buoyancy frequency, g is the gravitational acceleration, H is the local scale height ( 7 km), h n is the equivalent depth of the tidal mode, which can be calculated or taken directly from Forbes [1995], U is the zonal wind speed ð3þ 15 of 20

16 Figure 14. Altitude versus time contours of (top) buoyancy frequency (N), (middle) zonal wind (U), and (bottom) vertical wave number (m) for the (2, 4) mode of migrating semidiurnal tide (SW2), obtained from the 3 year averages ( ) of CMAM DAS data (monthly mean temperatures, temperature gradients, and zonal winds) near Saskatoon (52 N, 253 E). See text for details. (eastward taken as positive), C 0 is the phase (the solar time of maximum eastward wind) speed of the SDT at the equator (circa 465 m s 1 ), is colatitude, T is the temperature, z is the height, and G d is the dry adiabatic lapse (about 9.5 K/km in the mesopause region). Based on equations (1) to (3), the vertical wave number m for different Hough modes should show similar vertical and seasonal variations for a given location, except for the differences in the m value, since m depends only on h n when N and U are fixed. The m 2 for a low order mode can easily become negative in the mesosphere since the corresponding h n and thus h n are large. Such relationships were illustrated in Figure 6 of Yuan et al. [2008]. [30] Figure 14 provides the contours of the buoyancy frequency (N), zonal wind (U), and vertical wave numbers (m) for the (2, 4) mode, calculated from the 3 year averages of the CMAM DAS data near Saskatoon (52 N, 253 E). The results from Figure 14 are in good agreement with Riggin et al. [2003] with regards to the magnitudes of the three parameters. In contrast to Riggin et al. [2003], however, Figure 14 shows that the maximum value of m for (2, 4) is centered in July, not in August September. As mentioned above, the m parameter for other Hough modes (not shown) should provide similar seasonal structures except for the differences in the m value since m depends only on h n when N and U are fixed. Hence, our calculations suggest that the refractive effects may not account for the autumnal enhancement in the SW2 amplitude at middle/high northern latitudes, in contrast with the claim of Riggin et al. [2003]. [31] To find out the possible reasons for this apparent discrepancy (the time lag in the m maximum) between Riggin et al. [2003] and the present study, it is appropriate to comment on the results from two studies. For a given h n, a large negative U and a large N would cause a large m based upon equations (1) to (3). Figure 14 provides reasonable agreement between the three parameters. Riggin et al. [2003] found that large negative U values occurred in June August (their 16 of 20

17 Figure 15. Comparison between the amplitudes of the meridional wind westward propagating s = 1 nonmigrating semidiurnal tide (SW1) at 88 km, 80 S (blue lines; scale on left axis), and the temperature stationary planetary wave number 1 (SPW1) at 10 hpa, 70 N (red lines; scale on right axis), centered for the four austral summers (top to bottom) 2005/2006, 2006/2007, 2007/2008, and 2008/2009. Both the tidal and PW amplitudes are obtained, over a window of 4 days stepped by 1 day, from the CMAM DAS. The labels on the x axis indicate the first day of each month. 17 of 20

18 Figure 11), so the corresponding m values are supposed to be also large in June August if N is fixed. However, the calculated m (middle of their Figure 12) showed maxima in July September (note that the range of the time axis of their Figure 11 is different from those for their Figures 10 and 12). A similar problem is also found with their m contours (bottom of their Figure 12) when considering the variations in both N and U. Large N values were seen in July August (their Figure 10) and large negative U values were centered in July (their Figure 11), while the calculated maximum in m occurred in August September (bottom of their Figure 12). This might indicate that the m contour plots shown by Riggin et al. [2003] contain an overall time shift. 6. Comparison of Variability of the Antarctic SW1 With the Arctic SPW1 [32] Consistent with the radar and satellite measurements, the CMAM DAS provides a strong and dominant westward propagating s = 1 nonmigrating semidiurnal tide (SW1) in the MLT over Antarctica during austral summer. There is growing evidence that the presence of this tidal component might be attributed to the SPW1 and SW2 nonlinear interaction occurring in the NH stratosphere. The model study of Yamashita et al. [2002] verified that the nonlinear interaction forcing between SPW1 and SW2 could provide SW1. Their related calculations with this tide indicated that, once excited in the NH winter hemisphere, it could propagate to the polar MLT region of the SH and result in the local summer enhancement of SW1. Aso [2007] confirmed the transequatorial propagation of SW1, which is forced in the opposite winter hemisphere using a linearized steady and explicit tidal model. Based on observations using two radars, located at Scott Base (78 S, 167 E) and at Halley (76 S, 26 W), Baumgaertner et al. [2006] showed that the Antarctic SW1 amplitudes were positively correlated with the NH SPW1 near 1 hpa during the SH summer months. Murphy et al. [2009] has used amplitudes of the Antarctic SW1 during summer and early fall, obtained from 4 radars, to demonstrate strong correlations with the planetary wave (S = 1) amplitudes at middle and high northern latitudes between 100 and 0.1 hpa and hence source regions in the NH. A physical interpretation of the interactions described above is that the migrating tide propagates around the earth and interacts with the planetary wave in the winter stratosphere and lower mesosphere. These interactions yield nonmigrating tides which also propagate upward and toward the summer hemisphere, and are observed in the Antarctic MLT region [Yamashita et al., 2002; Aso, 2007]. [33] Figure 15 compares the meridional wind SW1 amplitude at 88 km, 80 S with the temperature SPW1 amplitude at 10 hpa, 70 N, within the CMAM DAS. Both the tidal and PW amplitudes are obtained over a window of 4 days stepped by 1 day. A good correlation is quite evident between variability of the Antarctic summer SW1 amplitude and the Arctic winter SPW1 amplitude for all four years. The results are consistent with the interactions described above and further support the predication that the Arctic winter SPW1 drives SW1 in the Antarctic summer. In contrast, the SW1 tide is weak in the Arctic summer mesosphere (Figures 5c and 5d). The CMAM DAS (not shown) does not provide a clear correlation between variability of the Antarctic winter SPW1 amplitude and the Arctic summer SW1 amplitude. This again agrees with the claims of Xu et al. [2009a, 2009b], indicating the existence of asymmetry in the interhemispheric SPW tidal coupling between two hemispheres. Note that the examination of the correlation between the tidal and PW fields provides only indirect evidence of the possible mechanism because the CMAM DAS involves complicated nonlinear feedbacks. 7. Summary and Conclusions [34] This study has examined the semidiurnal wind characteristics in the mesosphere simulated by the standard version of CMAM DAS which absorbs meteorological observations in the troposphere and stratosphere using a 3D variational assimilation technique. Emphasis is placed on the CMAM DAS analyses at 88 km which is the highest model height with a minimum height difference for the model/ satellite comparison. Three years ( ) of model data are analyzed. The main results from this study are as follows. [35] 1. The main seasonal features of amplitudes from CMAM DAS are generally similar to those observed for the semidiurnal tides at single stations from middle/high northern latitudes, while the corresponding phases are often in excellent agreement for CMAM DAS and radar, usually with short vertical wavelength in winter (monthly means km) and longer or evanescent in summer. [36] 2. The prominent wave numbers of nonmigrating semidiurnal tides for both the meridional wind and zonal wind include s = 1 (SW1), 0 (S0), 2 (SE2), 4 (SW4), 3 (SW3), and 6 (SW6). The presence of these nonmigrating tides is consistent with two mechanisms: nonlinear interactions between the migrating tide and SPWm, and zonally asymmetric thermal forcing. [37] 3. For the migrating semidiurnal tide (SW2), the major amplitude maxima usually occur at 40 N 60 N during December February and August September, and also at 40 S 60 S in April May, with the dominance of (2, 4) during October April and the (2, 3) and (2, 5) dominance for other months. This is shown to be in good overall agreement with the UARS HRDI/WINDII 95 km measurements. Our calculations show that the refractive effects may not account for the autumnal enhancement in the SW2 amplitude at middle/high northern latitudes, in contrast with the claim of Riggin et al. [2003]. [38] 4. Consistent with many radar and satellite measurements, the CMAM DAS shows that the SW1 tide is strong and dominant at the SH polar and high latitudes in the local summer season, but is much weaker in the NH. There is good overall agreement between the CMAM DAS 88 km analyses and HRDI/WINDII 95 km measurements for latitudinalseasonal variations of the SW1 amplitude. Very good agreement is seen between model and TIDI measurements for the SW1 amplitude in the MLT near the North Pole. The model and TIDI provide similar vertical profiles for the SW1 amplitude near the South Pole during the local summer, except for the modeled amplitude being larger. There is remarkably good agreement between the CMAM DAS and TIDI regarding the SW1 phase in the polar latitudes of both hemispheres. A strong correlation between variability of the SW1 amplitude in the MLT near the South Pole and the stratospheric SPW1 amplitude over the Arctic is found within 18 of 20