Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–30

Contents lists available at ScienceDirect

Journal of Atmospheric and Solar-Terrestrial Physics

1364-68

doi:10.1

$L.C.

083651

provisio

NNH05C

Nationa� Corr

E-m

journal homepage: www.elsevier.com/locate/jastp

Comparison of diurnal tide in models and ground-based observations duringthe 2005 equinox CAWSES tidal campaign$

L.C. Chang a,�, W.E. Ward b, S.E. Palo a, J. Du c, D.-Y. Wang b, H.-L. Liu d, M.E. Hagan d, Y. Portnyagin e,J. Oberheide f, L.P. Goncharenko g, T. Nakamura h, P. Hoffmann i, W. Singer i, P. Batista j, B. Clemesha j,A.H. Manson k, D.M. Riggin l, C.-Y. She m, T. Tsuda n, T. Yuan m

a Department of Aerospace Engineering Sciences, University of Colorado, Boulder, CO, USAb Department of Physics, University of New Brunswick, Fredericton, New Brunswick, Canadac Department of Applied Mathematics and Theoretical Physics, Cambridge University, Cambridge, United Kingdomd National Center for Atmospheric Research, Boulder, CO, USAe Institute for Experimental Meteorology, Obninsk, Russiaf Department of Physics and Astronomy, Clemson University, Clemson, SC, USAg MIT Haystack Observatory, Westford, MA, USAh National Institute of Polar Research, Tokyo, Japani Leibniz Institute of Atmospheric Physics at the Rostock University, Kuhlungsborn, Germanyj National Space Research Institute – INPE, Sa ~o Jose dos Campos, SP, Brazilk Institute of Space and Atmospheric Studies, University of Saskatchewan, Saskatoon, Saskatchewan, Canadal Colorado Research Associates, Division of NorthWest Research Associates, Boulder, CO, USAm Department of Physics, Colorado State University, Fort Collins, CO, USAn RISH, Kyoto University, Kyoto, Japan

a r t i c l e i n f o

Article history:

Received 22 October 2010

Received in revised form

17 December 2010

Accepted 24 December 2010Available online 31 December 2010

Keywords:

Diurnal tide

General circulation models

Model validation

CAWSES

Global tidal campaign

26/$ - see front matter & 2010 Elsevier Ltd. A

016/j.jastp.2010.12.010

Chang was supported by U.S. NSF award

8 under the direction of S.E. Palo at the

n of Rarotonga and Kauai radar data was s

C70. The National Center for Atmospheric R

l Science Foundation.

esponding author. Tel.: +886 921 952 794.

ail address: loren.chang@colorado.edu (L.C. C

a b s t r a c t

In this study, ground-based observations of equinox diurnal tide wind fields from the first CAWSES

Global Tidal Campaign are compared with results from five commonly used models, in order to identify

systematic differences. WACCM3 and Extended CMAM are both self-consistent general circulation

models, which resolve general climatological features, while TIME-GCM can be forced to approximate

specific conditions using reanalysis fields. GSWM is a linear mechanistic model; while GEWM is an

empirical model derived from ground-based and satellite observations. The models resolve diurnal

tides consistent in latitudinal structure with observations, dominated by the upward propagating (1,1)

mode. There is disagreement in the magnitudes of the tidal amplitudes and vertical wavelengths, while

differences in longitudinal tidal variability indicate differences in the nonmigrating tides in the models.

These points suggest inconsistencies in model forcing, dissipation, and background winds that must be

examined as part of a coordinated effort from the modeling community.

& 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The mesosphere and lower thermosphere (MLT) region of theEarth’s atmosphere extends between roughly 50–120 km altitude,and forms what is often considered to be the transition regionbetween the electrically neutral, turbulence driven lower atmo-sphere and the charged, rarefied gases of the ionosphere and near

ll rights reserved.

s ATM-0228026 and ATM-

University of Colorado. The

upported by NASA contract

esearch is sponsored by the

hang).

Earth space. The dominant global scale dynamical features of thisregion are the atmospheric tides – persistent global scale oscilla-tions with periods at harmonics of a solar day, forced by periodicinsolation of water vapor by IR radiation and latent heat release inthe troposphere, as well as UV absorption by ozone in thestratosphere (Hagan, 1996). Though they are relatively small inthe troposphere and stratosphere source regions, the tides thenpropagate upward, growing in amplitude as the atmosphericdensity decreases, until they are dissipated in the lower thermo-sphere by turbulent and molecular diffusion (Hagan et al., 1995;Ortland and Alexander, 2006).

The tides are categorized by period, with the diurnal tidehaving a period of 24 h, the semidiurnal tide having a period of12 h, and so on. Tides of these specific periods may be furthersubdivided into migrating and nonmigrating tides based upon

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–3020

their zonal wavenumbers and phase velocities. Migrating tideshave zonal wavenumbers such that they propagate westward atthe same rate as the apparent motion of the Sun. On the otherhand, nonmigrating tides have zonal wavenumbers such thattheir phase velocities are not Sun-synchronous. In a simplifiedatmosphere, the latitudinal structure of each tidal component canbe decomposed as being a sum of Hough modes – orthogonalsolutions to Laplace’s tidal equation. These modes may propagatevertically away from their source regions, or can be trapped(evanescent) modes, which decay away from their excitationregions.

The diurnal tide represents the response of the atmosphere tothe largest component of solar heating, and is the primary targetfor this study. From past observations (Hays and Wu, 1994;Burrage et al., 1995; Wu et al., 2006, 2008), it is known that thediurnal tide is dominated at low latitudes by the migratingdiurnal tide (zonal wavenumber s¼1), which itself is dominatedby the upward propagating diurnal (1,1) Hough mode, accountingfor a primary peak in tidal temperature fields at the equator, withsecondary peaks near 351 latitude; and peaks in horizontal windfields around 20–301 latitude in both hemispheres. Evanescentmodes dominate the diurnal tidal response at mid to highlatitudes. Additionally, diurnal tidal amplitudes exhibit a regularseasonal amplitude variation, maximizing at the equinoxes andminimizing at the solstices.

Though the migrating tides are usually dominant, nonmigrat-ing tides can also be important in certain times and regions,superimposing with the migrating tides to create longitudinaltidal variability when observed from ground stations at differentlongitudes along the same latitude circle. Using GSWM runsincorporating latent heat release due to deep tropical convectionin the troposphere, Hagan and Forbes (2002) found that inaddition to modulating the radiatively excited migrating diurnaltide, latent heat release was also responsible for forcing eastwardpropagating diurnal tides with wavenumbers 2 and 3 (s¼�2,�3),westward propagating diurnal wavenumber 2 (s¼2), as well asdiurnal standing (s¼0) oscillations. Various nonmigrating tides,such as the s¼2 nonmigrating diurnal tide (Oberheide et al.,2002), and the s¼1 nonmigrating semidiurnal tide may also beforced by nonlinear interactions between the migrating tides andstationary planetary waves (Chang et al., 2009). Similarly, numer-ical experiments by Hagan et al. (2009) have shown that migrat-ing and nonmigrating tides of the same period may also interactwith each other to produce stationary planetary waves.

In addition to reflecting the nature of forcing phenomenaoccurring at lower levels, the tidal response in the MLT regionalso has implications for thermosphere and ionospheric phenom-ena occurring far above the mesosphere and lower thermosphere.Tidal perturbations with longer vertical wavelengths are moreresistant to dissipation, and can penetrate into the lower thermo-sphere. The signatures of such tidal components can be trans-mitted further up into the upper thermosphere and exospherethrough interaction with the E-region dynamo, or through directupward propagation (Oberheide and Forbes, 2008). Evidence ofsuch tidally modulated phenomena has been found in observa-tions of equatorial F region anomaly (Immel et al., 2006), as wellas in exosphere temperatures derived from satellite drag data(Forbes et al., 2009).

Clearly, understanding the atmospheric tides observed at theirpeak amplitudes in the MLT region can help to shed light on theenergetics and dynamics of regions both above and below the MLT.This makes the understanding of tidal structure and variability inthe MLT region on seasonal and shorter day-to-day time scales,important to the understanding of the atmosphere as a whole. Suchan effort necessitates the use of observations that are global in scale,with local time coverage sufficient to unambiguously resolve the

tides over short time scales. At the present time, satellite-borne tidalobservations can provide global scale coverage in the MLT region,but require very long integration times due to their slow rates oflocal time precession. Conversely, ground-based instruments canprovide excellent resolution of day-to-day changes in the tide, butare limited in spatial coverage. As a result of these limitations, theuse of tidal observations from multiple ground-based locations, aswell as both space and ground-based observational platforms isdesired. Examples of such studies include Yuan et al. (2010), whichcombined observations from TIMED/SABER and sodium lidar; aswell as Ward et al. (2010), which examined the consistency betweentidal signatures in TIMED satellite data, ground-based observations,and the Extended Canadian Middle Atmosphere Model.

In the following study, we expand on the work of Ward et al.(2010) by comparing multiple ground-based observations of thediurnal tide from the CAWSES Global Tidal Campaigns to resultsfrom five different atmospheric models. The objective is to examinethe capabilities of current models in resolving the diurnal tide, andhighlight commonalities and shortcomings amongst the models.

2. Observations

The CAWSES Global Tidal Campaigns are intended to provideperiods of coordinated global scale observations of the atmo-spheric tides, involving multiple ground and space-based instru-ments. A total of four observational campaigns were conductedbetween 2005 and 2008, corresponding to the four seasons (Wardet al., 2010). In this study, we focus on ground-based observationsfrom the first campaign, which took place during the equinoxconditions of September/October 2005. In particular, we examineresults from 14 ground-based instruments in various regionsaround the globe, shown in Fig. 1 and listed in Table 1. Thesediurnal tidal observations are then compared with tidal resultsfrom five different atmospheric models, which will be furtherexamined in the following section.

The ground-based instruments utilized in this study includelidar, medium frequency (MF), and meteor (MR) radar stations,the characteristics of which are briefly summarized. MF and MRradars have long been utilized in studies of MLT wind fields, andfunction by measuring backscatter off of ionized structures in theMLT region, which are advected with the background wind fields,thus acting as tracers. MR utilizes ionization trails from ablatingmicro-meteors, while MF radars measure backscatter fromgradients and fluctuations in the ionospheric D region (Khattatovet al., 1996; Hocking and Thayaparan, 1997; Riggin et al., 2003).Barring hardware failures, these radar systems are capable ofcontinuous operation over long periods of time, making themattractive for long-term climatological studies.

Lidars are a more recent development in MLT observations.The CSU sodium lidar at Fort Collins is utilized in this study, andcan provide profiles of MLT winds and temperatures throughbackscatter from sodium atoms resulting from meteor ablation inthe region (She et al., 2002, 2003). The resulting wind andtemperature profiles have high vertical resolution (0.5 km forthe lidar dataset utilized here, compared to 2–4 km for the radardatasets), allowing for the study of smaller scale structuresand transient features, such as gravity waves. However, suchlidars are limited by weather conditions and have shorter con-tinuous operating times on the order of days to roughly one week(Yuan et al., 2010). In this particular case, the lidar was inoperation for roughly seven days. Generally speaking, MF andMR radars are capable of long duration observations, whilecurrent lidar systems provide higher vertical resolution over ashorter time period.

Table 1Observational sites utilized in comparison.

Site Instrument Latitude Longitude

Tromso MF 701N 191E

Andenes Meteor 691N 161E

Saskatoon MF 521N 1071W

Fort Collins Lidar 411N 1041W

Platteville MF 401N 1041W

Kauai MF 221N 1601W

Trivandrum Meteor 91N 771E

Kototabang Meteor 01 1001E

Cariri Meteor 71S 371W

Learmonth Meteor 221S 1141E

Rarotonga Meteor 221S 1601W

Cachoeira Paulista Meteor 231S 451W

Santa Maria Meteor 301S 541W

Rothera Meteor 681S 681W

Fig. 1. Geographic location of ground-based instruments used for this study. Meteor radar (MR) sites denoted by asterisks, medium frequency (MF) radar sites denoted by

diamonds, and lidar denoted by triangle.

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–30 21

3. Model descriptions

Computer models are of great use in interpreting observationsof tidal structure and variability, as well as for conductingnumerical experiments in a controlled environment. Modelingefforts to date can be roughly classified as either empirical orphysics-based models, the latter of which includes both generalcirculation models (GCMs) and comparatively simpler mechan-istic models. For this study, we examine tidal results from fivedifferent models, which are described in the following.

3.1. General circulation models: extended CMAM, WACCM3 & TIME-

GCM

Two fully nonlinear, time-dependent GCMs are utilized in thisstudy: version 3 of the NCAR Whole Atmosphere CommunityClimate Model (WACCM3); the Extended Canadian Middle Atmo-sphere Model (Extended CMAM). Both models are fully nonlinear,and capable of integrating the full set of primitive equations toresolve the global variation of the tides as a function of time.There are however, several distinct differences between themodels.

Extended CMAM (Fomichev et al., 2002; Du et al., 2007) extendsfrom the surface to roughly 210 km. The governing equations aresolved in spectral space, with dependent variables expandedhorizontally in spherical harmonics through wavenumber 32(T32). This corresponds to a horizontal grid of approximately6.0�6.01, while the vertical resolution varies from 150 m near thesurface to 2 km near the tropopause, and 3 km in regions furtherabove. Tidal oscillations are self-consistently generated from shortand long wave radiation absorption, large-scale condensation, andconvective heating. Nonorographic gravity waves are parameterizedusing the Doppler spread parameterization of Hines (1997a,b).

WACCM3 (Garcia et al., 2007) extends from the surface to roughly150 km altitude. The dynamical core of WACCM3 is derived from thatof NCAR’s Community Atmosphere Model, version 3 (CAM3), whichuses a finite volume method for evaluating the primitive equations inphysical space. The horizontal resolution of the model as run for thisstudy was 1.9�2.51, with 66 vertical levels corresponding to aresolution of approximately 1.1 km in the troposphere, 1.75 km at

the stratopause, and 3.5 km above 65 km. The tides are forced self-consistently through long and shortwave heating, while a modifiedLindzen scheme detailed in Garcia et al. (2007) is used to parameter-ize the effects of gravity waves.

Unlike the Extended CMAM and WACCM3, the NCAR Thermo-sphere Ionosphere Mesosphere Electrodynamics General Circula-tion Model (TIME-GCM) (Roble and Ridley, 1994; Hagan andRoble, 2001; Yamashita et al., 2010), is a time-dependent modelwith a lower boundary around 30 km, and extends upward toapproximately 500 km in the thermosphere. The horizontal reso-lution is similar to that of WACCM3 at 2.5�2.51, with a verticalresolution of approximately 1.5 km in the MLT region. Waves andother phenomena propagating upward from below the 30 kmlower boundary must be implemented using GSWM model fields,reanalysis data, or specified explicitly by the user. Though themodel does not extend to the tidal forcing regions in the tropo-sphere, the use of a lower boundary in the stratosphere allows formore flexibility in terms of the numerical experiments that can beperformed, as certain lower atmospheric forcing may be custo-mized or eliminated altogether.

The diurnal tide and planetary waves in this particular run ofTIME-GCM are forced at the lower boundary solely using ECMWF

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–3022

reanalysis fields at the same 2.51 resolution as the TIME-GCMhorizontal grid, corresponding to the first month of the first tidalcampaign (September 2005), with the intention of approximatingactual atmospheric conditions and tidal forcing during that time.This is in contrast to past TIME-GCM studies, such as Hagan andRoble (2001), which used GSWM seasonal values for tidal forcingat the lower boundary, or coupled the TIME-GCM lower boundaryto another general circulation model (CCM3) (Liu and Roble,2002). While quasi-realistic forcing of planetary waves at theTIME-GCM lower boundary has also been implemented in thepast with NCEP Reanalysis fields (Liu and Roble, 2005), thismethod cannot be used for tidal forcing, due to the once perday resolution of the NCEP fields. The four times daily timeresolution of the ECMWF fields makes it an attractive candidatefor replacing GSWM seasonal fields for the purposes of forcing thediurnal tide in TIME-GCM to approximate a specific period oftime, though the time resolution is still insufficient to accuratelyexcite the semidiurnal tide. As such, TIME-GCM is the only modelexamined in this study that is forced by conditions particular tothe September 2005 time period, with the other models repre-senting climatological results.

3.2. Linear mechanistic model: GSWM

We also examine the 2002 version of the linear mechanisticGlobal Scale Wave Model (GSWM-02) (Hagan et al., 1999; Haganand Forbes, 2002). Unlike the aforementioned general circulationmodels, which numerically integrate the full set of perturbationequations forward in time and space, the GSWM solves asimplified set of linear equations under the assumption ofsteady-state conditions and sinusoidal behavior in longitude andtime, with prescribed time independent background windsderived from High Resolution Doppler Imager (HRDI) observa-tions, as well as prescribed background temperatures, forcing, anddissipation (in the form of Rayleigh friction and eddy diffusivitycoefficients). This produces wave amplitudes and phases as afunction of latitude and altitude for each specified frequency andwavenumber pair under the prescribed forcing and backgroundconditions, and is useful in understanding the effects of changesin these parameters on tidal structure. However, GSWM cannotcapture tidal components or variability generated by nonlinearinteractions due to the linearized nature of the equations solved.For this study, we use the diurnal tidal results of GSWM-02,which include both the migrating (s¼1) and 12 nonmigratingdiurnal tidal components from zonal wavenumbers s¼6 to �6.The horizontal resolution of 3.0 �5.01 and the vertical resolutionof the model are less than 1 km in the MLT region.

3.3. Empirical model: GEWM

The Global Empirical Wind Model (GEWM) (Portnyagin andSolovjova, 2000; Portnyagin et al., 2004; Portnyagin, 2006) is atwo dimensional model of the migrating diurnal tide, semidiurnaltide, and mean horizontal wind derived from a network of 46ground-based MF and MR radar stations, as well as observationsfrom the HRDI and WINDII (wind imaging interferometer) instru-ments aboard the UARS spacecraft. The ground-based observa-tions span a broad latitude range from 801N through 901S, whileHRDI measurements equatorward of 401, and WINDII measure-ments equatorward of 451 were utilized. Systematic biasesbetween the horizontal winds measured by the various instru-ments were identified with respect to the MR measurements,which were assumed to be the most accurate. Correction factorswere then applied to the data to minimize the aforementionedbiases (Portnyagin et al., 2004). The resulting empirical model

covers 80–100 km altitude, with a latitude resolution of 51. Weutilize the results of GEWM-II in the following study, whichprovides monthly amplitudes and phases for the migratingdiurnal tide as a function of latitude and altitude (Portnyagin,2006), derived from uncorrected MR, HRDI, and WINDII data, andwith MF data scaled upward above 88 km altitude.

4. Processing methodology

At each observational station, and the nearest model gridpointin WACCM3, Extended CMAM, and TIME-GCM, hourly averagedzonal and meridional winds from each altitude level over theentire two month campaign were fit to a basis function containingthe diurnal, semidiurnal, and terdiurnal tides using a linear least-squares fitting routine:

FðtÞ ¼X3

s ¼ 1

As cosðsOtþfsÞ ð1Þ

This provides tidal amplitudes (As) and phases (fs) as a functionof height for each individual ground-station and at the nearestmodel gridpoint. Errors were computed from the fit covariances.

Output from GSWM already provides diurnal tide amplitudesand phases at all model latitude, longitude, and altitude grid-points, while GEWM provides the amplitudes and phases of themigrating diurnal component with respect to latitude andaltitude.

5. Results

5.1. Latitude variation

Fig. 2 shows the model and observed diurnal tidal amplitudesas a function of latitude at 90 km for the zonal and meridionalwind fields during September/October 2005. The diurnal tidalamplitudes for all of the models except GEWM were computed ateach latitude and longitude grid point, then averaged zonally foreach model latitude grid point. As no phase information wasincluded in the average, this method will not eliminate the effectsof tidal components with nonzero zonal wavenumbers, whichwould otherwise average out to zero when integrated over 3601of longitude. The contribution of both migrating and nonmigrat-ing components to the overall diurnal tide is also manifested inthe form of differences in tidal amplitudes from observations atsimilar latitudes, but different longitudes. This longitudinal varia-bility can be seen in the observations near 201S. The two stationsnear 401N (Platteville MF radar and Fort Collins lidar) are locatedrelatively close to one another. Differences between these twosites are attributable to differing observational lengths, with thelidar being active for roughly seven days, while the MF radar wasactive for much of the two month campaign. When the MF radardata was sampled only during the times when the lidar wasactive, the resulting tidal results demonstrated good agreementbetween the two instruments.

With the exception of TIME-GCM (dot-dashed line), all of themodels resolve the expected bimodal structure with tropicalpeaks commonly associated with the diurnal (1,1) wind expan-sion functions in both horizontal wind fields, consistent with theobservational data. However, the models differ significantly fromone another on the magnitude of the tidal amplitudes in the lowlatitudes: GSWM and Extended CMAM display the largest ampli-tudes with peaks exceeding 50 m/s, WACCM3 and GEWM areintermediate with peak amplitudes in the vicinity of 15–30 m/s,while TIME-GCM displays the smallest peak amplitudes, around10–15 m/s. Additionally, the latitude at which the tropical peaks

Fig. 2. Model and observed diurnal tidal amplitudes as a function of latitude at 90 km during September/October 2005. Observational data denoted by same symbols as in

Fig. 1 with error bars. Model amplitudes computed at each model gridpoint and zonally averaged for each latitude circle, and denoted by lines as indicated.

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–30 23

occur also varies, with the peaks in WACCM3 generally occurringat lower latitudes compared to the other models, especially in theNorthern Hemisphere.

Using generalized Hough modes, Ortland (2006) found that anarrowing of the migrating diurnal tidal peaks could be attributed tothe presence of an eastward equatorial zonal mean zonal wind jet.This may be a factor in the narrower horizontal wind peaks found inWACCM3, though further comparison of the background wind fieldsof the various models is needed to confirm this. It is also worthnoting that Ortland (2006) also found that increased dissipationresulted in a broadening of the tidal peaks, though again, thisrequires analysis of the different gravity wave parameterizationsutilized in the GCMs, which is beyond the scope of this study.

The diurnal tidal amplitudes and phases in the zonal andmeridional wind fields as a function of height are shown in Fig. 3for observations and model gridpoints nearest to the Kauai MF radarstation (221N, 2001E), which is located near the Northern Hemi-sphere tropical peak of the diurnal (1,1) Hough mode. The differencein tidal amplitude between the models is again significant through-out the entire domain, with GSWM and Extended CMAM showingthe largest amplitudes and TIME-GCM displaying the smallestamplitudes. The amplitudes in WACCM3 and GEWM are closest tothe observed amplitudes throughout most of the observationaldomain. In the case of GEWM this is not unexpected as GEWMwas derived in part from ground-based observational data. It hasbeen noted that the magnitude of MLT winds observed by thesatellite-borne HRDI instrument were generally larger than thosemeasured by MF radar (Burrage et al., 1996; Khattatov et al., 1996).Given that the magnitude of GSWM Rayleigh friction coefficientswere tuned based upon diagnostics using HRDI diurnal windclimatologies (Hagan et al., 1999), this may explain the largerGSWM amplitudes compared to GEWM and radar at this location.

Causes for some of the aforementioned differences betweenthe models and observations may be revealed through

examination of the tidal phase structure. The slopes of the tidalphases reveal the vertical wavelengths of the diurnal tide resolvedby the various models and observations. Past studies have shownthat tidal vertical wavelengths are affected by eddy diffusion andmomentum deposition in the MLT region from gravity wavebreaking. Gravity wave momentum forcing can increase ordecrease tidal amplitudes, depending upon the phase of theforcing relative to the tide, while also shortening the tidal verticalwavelength by locally advancing the tidal phases. On the otherhand, gravity wave induced eddy diffusion has been found toincrease tidal vertical wavelengths, while providing tidal dissipa-tion (Ortland and Alexander, 2006). Numerical experiments byOrtland and Alexander (2006) have found that the overall effectsof these two gravity wave mechanisms on the tide are highlydependent upon the gravity wave parameterization and sourcespectrum utilized. Additionally, the tidal vertical wavelengths arealso affected by the background zonal mean zonal winds, thoughany increase or decrease is highly dependent upon the structureof the mean winds (Ortland, 2006). Finally, the overall diurnaltidal vertical wavelength at a particular location is also dependentupon the superposition of the migrating and nonmigrating tidalcomponents that comprise the total diurnal tide, as well as thesuperposition of the different meridional modes of each tidalcomponent. Therefore, differences in tidal vertical wavelengthmight also be the result of different migrating and nonmigratingtidal amplitudes due to differences in tidal forcing.

In the observational domain, the diurnal tide in both ExtendedCMAM and GSWM shows vertical wavelengths that are shorterthan observations with amplitudes that exceed those in observa-tions. In the case of Extended CMAM, it has been noted thatthe amplitudes of the eastward propagating components ofthe diurnal tide are up to a factor of three stronger than thevalues observed by the TIMED Doppler Interferometer (TIDI),while westward propagating components are of comparable

Fig. 3. Observed zonal (a) and meridional (b) wind diurnal tidal amplitudes (left panels) and phases (right panels) as a function of height during September/October 2005

at Kauai (221N, 2001E) and from nearest model gridpoints. Observations shown as red diamonds with error bars. Models denoted by same linestyles as in Fig. 2. (For

interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–3024

magnitudes. It is therefore possible that parameterizations oflatent heat release and deep convection responsible for generatingthe eastward propagating diurnal tidal components are overesti-mated (J. Du, private communication, 2010). Another possibility forthe larger amplitudes and decreased vertical wavelengths in CMAMmay be the result of in-phase gravity wave momentum forcing,

commonly associated with the Hines parameterization employed inthat model (McLandress, 2002).

While similar in-phase forcing may also be possible in GSWMthough the prescribed eddy diffusivity coefficients utilized in thatmodel (Hagan and Forbes, 2002), earlier versions of GSWMshowed an increase in diurnal tidal amplitudes when eddy

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–30 25

diffusivity was omitted (Burrage et al., 1995), indicating that theeffect of eddy diffusivity in GSWM is primarily to suppress thetide. The positive real valued Rayleigh friction used to parameter-ize the effects of gravity wave drag in GSWM acts to reduce thetidal amplitudes with little to no effect on the vertical wavelength(Hagan et al., 1999; McLandress, 2002), and underestimated

Fig. 4. Same as Fig. 3, but for Platteville MF radar (red diamonds), and Fort Collins lidar

the references to color in this figure legend, the reader is referred to the web version

equinox Rayleigh friction coefficients could also potentially be afactor in the larger tidal amplitudes. For both models, it is alsopossible that the larger tidal amplitudes may be the result ofincreased tidal forcing in the lower atmosphere. As mentionedpreviously, the model zonal mean zonal winds may also play arole, though it should be noted that the prescribed zonal mean

(blue triangles). Model coordinates taken near 401N, 2551E. (For interpretation of

of this article.)

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–3026

zonal winds in GSWM were based on monthly HRDI-derivedclimatologies, and are therefore expected to be more realisticthan the self-consistently calculated values in Extended CMAM.

It is also worth noting that TIME-GCM shows very long verticalwavelengths with very small tidal amplitudes. This suggests thatthe ECMWF lower boundary in this TIME-GCM run was not strongenough to force the upward propagating diurnal (1,1) mode that

Fig. 5. Same as Fig. 3, but for Ande

should dominate the diurnal tide in this region, since thedissipation scheme is the same as that used for past runs withGSWM forced tides. This underestimation may be a result of the6 h time resolution of ECMWF not coinciding with the maximumheating signatures in the troposphere. The model vertical resolu-tion is also connected to the strength of the resolved diurnal tidedue to its relatively short vertical wavelength. However, this is

nes meteor radar (691N, 161E).

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–30 27

unlikely to explain the smaller than expected diurnal tidalamplitudes, given that the diurnal (1,1) mode is properly resolvedby TIME-GCM at identical resolutions using GSWM forcing.Consequently, it may be necessary to increase ECMWF diurnaltidal forcing in future runs of TIME-GCM.

Figs. 4 and 5 show the diurnal tidal amplitudes and phases atPlatteville/Fort Collins (both near 401N, 2551E) and Andenes (691N,161E), respectively representative of the tidal responses at mid andhigh latitudes. While the diurnal tidal amplitudes are generallysmaller than those at lower latitudes, there are still significantdifferences in the relative magnitudes of tidal amplitudes resolvedby the various models, especially at Platteville/Fort Collins.

The Platteville MF radar observations (Fig. 4a and b), as well asGEWM, WACCM3, and TIME-GCM all show amplitudes lowerthan 10 m/s with few changes in the vertical direction. The nearbyFort Collins sodium lidar shows amplitudes that are somewhatlarger than the MF radar, with the considerably better verticalresolution revealing a more complex vertical structure. Thedifferent observational time spans also contribute to differencesbetween the MF radar and lidar observations. The shorter sevenday period during which the lidar was operational is reflected inthe larger uncertainties and broader bandwidth for the diurnaltide fitted from the lidar data. As the radar was operational almostcontinuously during the two month campaign, the diurnal tidalfitted from the radar data has a narrower bandwidth, and is lessaffected by short term tidal variability.

The vertical wavelengths in the observations, WACCM3, TIME-GCM, and altitudes below roughly 85 km in GEWM are very long,indicative of evanescent vertical structure. In contrast, the ampli-tudes in the Extended CMAM and GSWM are much larger, and showsignificant vertical variation. The Extended CMAM resolves twoamplitude peaks in excess of 25 m/s at 88 and 105 km, while GSWMshows a single large peak at 105 km with an amplitude of 70 m/s.

Fig. 6. Zonal variability of zonal wind (top) and meridional wind (bottom) diurnal tide

Observational sites (from left to right) are Learmonth (231S, 1141E), Rarotonga (221S, 2

Additionally, the vertical wavelengths as determined by the phaseprogression of the tidal horizontal winds in Extended CMAM andGSWM are shorter than those in the observations and the othermodels (with the exception of GEWM above roughly 85 km),indicative of upward propagating modes. It is also noted that thereis generally good agreement in the tidal phases measured by the MFradar and the lidar below roughly 95 km, though the lidar verticalprofile again displays greater spatial variability.

Fig. 5a and b show observations and model results for Andenesmeteor radar. The diurnal tidal response from the models and thedata are much more consistent in the observational domaincompared to mid-latitudes, showing amplitudes at and below10 m/s below 95 km, and evanescent phases within approxi-mately 5 h of one another.

From this, it is apparent that the upward propagating diurnaltidal modes in the Extended CMAM and GSWM are larger in themid-latitudes than those in the other models and in the observa-tions, potentially due to increased lower atmospheric forcing. Incontrast, the TIME-GCM results from Kauai suggest that ECMWFforcing may be insufficient to excite the upward propagatingmodes even at low latitudes. This is consistent with previousanalysis of the latitudinal structure of the diurnal tide in Fig. 2.

5.2. Longitudinal variation

One limitation of ground-based observations in tidal studies istheir inability to distinguish the migrating tide from other nonmi-grating tidal components that are also present, resulting in observa-tions showing the superposition of all migrating and nonmigratingcomponents, which are zonally asymmetric (Ward et al., 2010). Thelongitudinal variability in the diurnal tidal structure that results canbe seen by comparing results from three meteor radar stations

amplitudes at 221S, 92.5 km altitude for three observational sites and four models.

001E), and Cachoeira Paulista (231S, 3151E). Model results shown are as indicated.

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–3028

located at similar latitudes near the Southern Hemisphere tropicalpeak at 221S: Learmonth (231S, 1141E), Rarotonga (221S, 2001E), andCachoeira Paulista (231S, 3151E).

Fig. 6 shows the zonal profiles of diurnal tide amplitude inobservations and model results at approximately 92.5 km altitudefrom the three aforementioned locations, as well as model resultsfrom WACCM3, Extended CMAM, TIME-GCM, and GSWM at thenearest latitude gridpoints. GEWM contains only the migratingdiurnal tide, and therefore does not account for longitudinalvariation in tidal amplitude. The observational data showssignificant differences in diurnal tidal amplitude between the threestations due to superposition of migrating and nonmigratingtides, with amplitudes largest at Cachoeira Paulista, and smallestat Rarotonga. However, the paucity of these ground-based

Fig. 7. Time variation of diurnal tide zonal winds at 90 km altitude in TIME-GCM

(solid line) and observational data (points). Sites shown from top to bottom are

Andenes (691N, 161E), Platteville (401N, 2551E), Kauai (221N, 2001E), Trivandrum

(91S, 771E), and Cachoeira Paulista (231S, 3151E).

observations precludes any definitive characterization of nonmi-grating tidal variations from this ground-based perspective.

While all four models also resolve significant longitudinalvariability in diurnal tidal amplitudes, the pattern of the ampli-tude variation in the models do not agree with observations. Thissuggests that the nonmigrating tides excited in the models differfrom those observed during the tidal campaign, though again, thiscomparison is limited by the sparse distribution of the ground-based observations. It is interesting to note, however, thatExtended CMAM, TIME-GCM, and GSWM display similar patternsof zonal variation. This may be due to similarities in loweratmospheric nonmigrating tidal forcing, such as that from latentheat, though this cannot be definitively stated without moredetailed examination of the model forcing parameters, particu-larly the nonmigrating components of latent heating. Addition-ally, some nonmigrating tidal components may be generatedthrough nonlinear interaction between the migrating tide andany stationary planetary waves that may be present (Oberheideet al., 2002). This may also be a factor in nonmigrating tidalgeneration in the three nonlinear models.

5.3. Time variation

As mentioned previously, one of the main rationales for forcingthe TIME-GCM lower boundary with ECMWF Reanalysis fields isfor the purpose of quasi-realistic tidal forcing. In particular, weare interested in determining how accurately TIME-GCM canreproduce the day-to-day variation of the diurnal tide using thismethod. Figs. 7 and 8 show the time variation of the diurnal tidehorizontal wind fields at 90 km in observational data and TIME-GCM at five selected sites spanning from 691N to 231S. Whilethere are occasional similarities in the time variation of thediurnal tide between the model and the observations, correlationis generally very low. Correlation coefficients between the modelresults and observations with zero lag time computed for all ofthe locations examined were less than 0.5 in nearly all cases, orwere unreliable due to long observational data gaps.

Additionally, it is apparent that the model amplitudes resolvedat sites near the low latitude peaks of the diurnal tide (Kauai andCachoeira Paulista) are much smaller than those observed. Thisagain indicates that the upward propagating diurnal (1,1) modethat is known to dominate the structure of the diurnal tide is notproperly resolved with ECMWF forcing in TIME-GCM. Addressingthe amplitude discrepancy will require comparison of the ECMWFtidal components at the lower boundary with similar GSWMseasonal values normally used to force the tides in TIME-GCM.

6. Conclusions

Diurnal tidal results from the Extended CMAM, WACCM3, TIME-GCM (with ECMWF forcing), GSWM-02, and GEWM have beencompared to observations from 14 ground-based instrument sitesparticipating in the first CAWSES Global Tidal Campaign in Septemberand October 2005. With the exception of TIME-GCM with ECMWFforcing, all the models examined, as well as the observational dataresolved a diurnal tide around 90 km altitude dominated by thebimodal structure indicative of the upward propagating diurnal (1,1)Hough mode at the low latitudes, transitioning to evanescent modesat mid to high latitudes. However, several differences between themodels were also identified:

(1)

Compared to the other models and observations, ExtendedCMAM and GSWM-02 showed much stronger tidal ampli-tudes and shorter vertical wavelengths near the low latitudepeaks of the upward propagating diurnal (1,1) mode. The

Fig. 8. Same as Fig. 7, but for meridional winds.

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–30 29

upward propagating modes also penetrated further pole-wards than the other models and the observations.

(2)

In contrast, low latitude tidal amplitudes in WACCM3, GEWM,and the observations were much weaker, with longer verticalwavelengths. WACCM3 also showed a much narrower latitu-dinal structure, with the tropical peaks in the horizontal windfields located further Equatorward compared to the othermodels.

(3)

TIME-GCM with ECMWF forcing did not resolve the diurnal(1,1) mode at all in the zonal wind fields, and only weakly inthe meridional wind fields. Combined with the very longvertical wavelengths resolved by TIME-GCM at low latitudes,this indicates that the ECMWF diurnal tidal forcing as imple-mented in this particular model run is insufficient. Addition-ally, the correlation between the time variation of the diurnaltide resolved in TIME-GCM and observations is still quite low,though some similar features are occasionally seen.

(4)

Longitudinal variation of the diurnal tidal structure at 221Sresolved by WACCM3, Extended CMAM, TIME-GCM, andGSWM-02 differs from that seen by the three radar siteslocated at similar latitudes, though some of the model resultsare similar to one another. The sparse distribution of theground-based observations is a significant limitation on anyfurther comparison of the longitudinal tidal variabilitybetween the models and the ground-based data.

We discuss first, the tidal response in GSWM and GEWM,which have fewer parameters and are therefore easier to analyze.The primary difference exhibited by these two models is in termsof tidal amplitude, with GSWM showing larger amplitudes andGEWM showing smaller amplitudes. This is not entirely surpris-ing as the Rayleigh friction magnitude in GSWM was specificallytuned to produce a tidal response of similar magnitude to thatobserved by HRDI, which is known to resolve stronger winds thanthe ground-based radar results that were used to create GEWM(Burrage et al., 1996; Khattatov et al., 1996; Hagan et al., 1999).

Further validation efforts, especially in the case of the threeGCMs examined, will require coordinated input from the model-ing community. In this comparison, we have identified thefollowing parameters that require further examination:

(1)

The differences in tidal vertical wavelength and amplitudepoint to a need for comparison of the eddy diffusivities, tidalforcing, and zonal mean zonal winds in the physics-basedmodels. In the cases of Extended CMAM, the larger tidalamplitudes and shorter vertical wavelengths may be relatedto the overestimation of eastward propagating diurnal tidalcomponents due to the latent heating and deep convectionparameterizations. Additionally, the Hines gravity wave para-meterization employed by CMAM is also known to producein-phase momentum forcing. In the case of GSWM, theshorter vertical wavelengths and larger amplitudes may bethe result of in-phase forcing from the prescribed eddydiffusivities, or stronger lower atmospheric tidal forcing.Lower values of Rayleigh friction in GSWM might also con-tribute to underdamping of the tidal amplitudes. In the case ofthe smaller tidal amplitudes in TIME-GCM, tidal amplitudes inthe ECMWF lower boundary should be compared with similarvalues from the other models.

(2)

Differences in longitudinal variation of the tide are indicativeof differences in the nonmigrating diurnal tides resolved inthe various models. This in turn suggests that further com-parison of the migrating and nonmigrating heating rates inthe models is necessary, particularly latent heating rates,which are believed to be important in generating the non-migrating tides. Analysis of the stationary planetary wavesexcited in the three nonlinear models will also be useful inascertaining the potential for nonmigrating tidal generationvia nonlinear interaction with the migrating tide.

References

Burrage, M.D., Hagan, M.E., Skinner, W.R., Wu, D.L., Hays, P.B., 1995. Long-termvariability in the solar diurnal tide observed by HRDI and simulated by theGSWM. Geophys. Res. Lett. 22, 2641–2644.

Burrage, M.D., Skinner, W.R., Gell, D.A., Hays, P.B., Marshall, A.R., Ortland, D.A.,Manson, A.H., Franke, S.J., Fritts, D.C., Hoffmann, P., McLandress, C., Nicie-jewski, R., Schmidlin, F.J., Shepherd, G.G., Singer, W., Tsuda, T., Vincent, R.A.,1996. Validation of the mesosphere and lower thermosphere wind from thehigh resolution Doppler imager on UARS. J. Geophys. Res. 101, 10365–10392.

Chang, L.C., Palo, S.E., Liu, H.-L., 2009. Short-term variation of the s ¼ 1 nonmigratingsemidiurnal tide during the 2002 stratospheric sudden warming. J. Geophys.Res. 114, D03109 doi:10.1029/2008JD010886.

L.C. Chang et al. / Journal of Atmospheric and Solar-Terrestrial Physics 78–79 (2012) 19–3030

Du, J., Ward, W.E., Oberheide, J., Nakamura, T., Tsuda, T., 2007. Semidiurnal tidesfrom the extended Canadian Middle Atmosphere Model CMAM and compar-isons with (TIMED) Doppler interferometer (TIDI) and meteor radar observa-tions. J. Atmos. Sol. Terr. Phys. 69 (17–18), 2159–2202 doi:10.1016/j.jastp.2007.07.014.

Fomichev, V.I., Ward, W.E., Beagley, S.R., McLandress, C., McConnell, J.C., McFar-lane, N.A., 2002. Shepherd T.G., 2002. Extended Canadian middle atmospheremodel: zonal-mean climatology and physical parameterizations. J. Geophys.Res. 107 (D10), 4087 doi:10.1029/2001JD000479.

Forbes, J.M., Bruinsma, S.L., Zhang, X., Oberheide, J., 2009. Surface-exospherecoupling due to thermal tides. Geophys. Res. Lett. 36, L15812 doi:10.1029/2009GL038748.

Garcia, R.R., Marsh, D.R., Kinnison, D.E., Boville, B.A., Sassi, F., 2007. Simulation ofsecular trends in the middle atmosphere 1950–2003. J. Geophys. Res. 112,D09301 doi:10.1029/2006JD007485.

Hagan, M.E., 1996. Comparative effects of migrating solar sources on tidalsignatures in the middle and upper atmosphere. J. Geophys. Res. 101 (D16),21213–21222.

Hagan, M.E., Forbes, J.M., 2002. Migrating and nonmigrating diurnal tides in themiddle and upper atmosphere excited by tropospheric latent heat release.J. Geophys. Res. 107 (D24), 4754 doi:10.1029/2001JD001236.

Hagan, M.E., Roble, R.G., 2001. Modeling diurnal tidal variability with the NationalCenter for Atmospheric Research thermosphere-ionosphere-mesosphere-elec-trodynamics general circulation model. J. Geophys. Res. 106, 24869–24882.

Hagan, M.E., Forbes, J.M., Vial, F., 1995. On modeling migrating solar tides.Geophys. Res. Lett. 22, 893–896.

Hagan, M.E., Burrage, M.D., Forbes, J.M., Hackney, J., Randel, W.J., Zhang, X., 1999.GSWM-98: results for migrating solar tides. J. Geophys. Res. 104 (A4),6813–6827.

Hagan, M.E., Maute, A., Roble, R.G., 2009. Tropospheric tidal effects on the middleand upper atmosphere. J. Geophys. Res. 114 doi:10.1029/2008JA013637.

Hays, P.B., Wu, T., 1994. HRDI Science Team, 1994. Observations of the diurnal tidefrom space. J. Atmos. Sci. 51, 3077–3093.

Hines, C.O., 1997a. Doppler-spread parameterization of gravity-wave momentumdeposition in the middle atmosphere. Part 1: Basic formulation. J. Atmos. Sol.Terr. Phys. 59, 371–386.

Hines, C.O., 1997b. Doppler-spread parameterization of gravity-wave momentumdeposition in the middle atmosphere. Part 2: Broad and quasi-monochromaticspectra, and implementation. J. Atmos. Sol. Terr. Phys. 59, 387–400.

Hocking, W.K., Thayaparan, T., 1997. Simultaneous and collocated observation ofwinds and tides by MF and meteor radars over London, Canada (431N, 811W)during 1994–1996. Radio Sci. 32, 833–865.

Immel, T.J., Sagawa, E., England, S.L., Henderson, S.B., Hagan, M.E., Mende, S.B.,Frey, H.U., Swenson, C.M., Paxton, L.J., 2006. Control of equatorial ionosphericmorphology by atmospheric tides. Geophys. Res. Lett. doi:10.1029/2006GL026161.

Khattatov, B.V., Geller, M.A., Yudin, V.A., Hays, P.B., Skinner, W.R., Burrage, M.D.,Franke, S.J., Fritts, D.C., Isler, J.R., Manson, A.H., Meek, C.E., McMurray, R.,Singer, W., Hoffmann, P., Vincent, R.A., 1996. Dynamics of the mesosphere andlower thermosphere as seen by MF radars and by the high-resolution Dopplerimager/UARS. J. Geophys. Res. 101, 10393–10404.

Liu, H.-L., Roble, R.G., 2002. A study of a self-generated stratospheric suddenwarming and its mesospheric-lower thermospheric impacts using the coupledTIME-GCM/CCM3. J. Geophys. Res. 107 doi:10.1029/2001JD001533..

Liu, H.-L., Roble, R.G., 2005. Dynamical coupling of the stratosphere and meso-sphere in the 2002 Southern Hemisphere major stratospheric sudden warm-ing. Geophys. Res. Lett. 32 doi:http://10.1029/2005GL022939.

McLandress, C., 2002. The seasonal variation of the propagating diurnal tide in themesosphere and lower thermosphere: Part I The role of gravity waves andplanetary waves. J. Atmos. Sci. 59, 893–906.

Oberheide, J., Forbes, J.M., 2008. Tidal propagation of deep tropical cloudsignatures into the thermosphere from TIMED observations. Geophys. Res.Lett. 35, L04816 doi:10.1029/2007GL032397.

Oberheide, J., Hagan, M.E., Roble, R.G., Offermann, D., 2002. Sources of nonmigrat-ing tides in the tropical middle atmosphere. J. Geophys. Res. 107 (D21), 4567doi:10.1029/2002JD002220.

Ortland, D.A., 2006. A study of the global structure of the migrating diurnal tideusing generalized hough modes. J. Atmos. Sci. 62, 2684–2702.

Ortland, D.A., Alexander, M.J., 2006. Gravity wave influence on the global structureof the diurnal tide in the mesosphere and lower thermosphere. J. Geophys. Res.22 doi:10.1029/2005JA011467.

Portnyagin, Yu., 2006. A review of mesospheric and lower thermosphere models.Adv. Space Res. 38, 2452–2460 doi:10.1016/j.asr.2006.04.030.

Portnyagin, Y.I., Solovjova, T.V., 2000. Global empirical wind model for theupper mesosphere and thermosphere. I Prevailing wind. Ann. Geophys. 18,300–315.

Portnyagin, Yu., Solovjova, T., Merzlyakov, E., Forbes, J., Palo, S., Ortland, D.,Hocking, W., MacDougall, J., Thayaparan, T., Manson, A., Meek, C., Hoffmann,P., Singer, W., Mitchell, N., Pancheva, D., Igarashi, K., Murayama, Y., Jacobi, C.,Kuerschner, D., Fahrutdinova, A., Korotyshkin, D., Clark, R., Taylor, M., Franke,S., Fritts, D., Tsuda, T., Nakamura, T., Gurubaran, S., Rajaram, R., Vincent, R.,Kovalam, S., Batista, P., Poole, G., Malinga, S., Fraser, G., Murphy, D., Riggin, D.,Aso, T., Tsutsumi, M., 2004. Mesosphere/lower thermosphere prevailing windmodel. Adv. Space Res. 34, 1755–1762 doi:10.1016/j.asr.2003.04.058.

Riggin, D.M., Meyer, C.K., Fritts, D.C., Jarvis, M.J., Murayama, Y., Singer, W., Vincent,R.A., Murphy, D.J., 2003. MF radar observations of seasonal variability ofsemidiurnal motions in the mesosphere at high northern and southernlatitudes. J. Atmos. Sol. Terr. Phys. 65, 483–493.

Roble, R.G., Ridley, E.C., 1994. A thermosphere–ionosphere–mesosphere-electro-dynamics general circulation model (TIME-GCM): Equinox solar cycle mini-mum simulations (30–500 km). Geophys. Res. Lett. 21, 417–420.

She, C.Y., Sherman, J., Yuan, T., Williams, B.P., Arnold, K., Kawahara, T.D., Li, T., Xu,L.F., Vance, J.D., Acott, P., Krueger, D.A., 2003. The first 80-hour continuouslidar campaign for simultaneous observation of mesopause region tempera-ture and wind. Geophys. Res. Lett. 30 doi:10.1029/2002GL016412.

She, C.Y., Chen, S., Williams, B.P., Hu, Z., Krueger, D.A., 2002. Tides in themesopause region over Fort Collins, Colorado (411N, 1051W) based on lidartemperature observations overing full diurnal cycles. J. Geophys. Res. 107doi:10.1029/2001JD001189.

Ward, W.E., Oberheide, J., Goncharenko, L.P., Nakamura, T., Hoffmann, P., Singer,W., Chang, L.C., Du, J., Wang, D.-Y., Batista, P., Clemesha, B., Manson, A.H.,Riggin, D.M., She, C.-Y., Tsuda, T., Yuan, T., 2010. On the consistency of model,ground-based, and satellite observations of tidal signatures: initial resultsfrom the CAWSES tidal campaigns. J. Geophys. Res. 115, D07107 doi:10.1029/2009JD012593.

Wu, Q., Killeen, T.L., Ortland, D.A., Solomon, S.C., Gablehouse, R.D., Johnson, R.M.,Skinner, W.R., Niciejewski, R.J., Franke, S.J., 2006. TIMED Doppler interferom-eter (TIDI) observations of migrating diurnal and semidiurnal tides. J. Atmos.Sol. Terr. Phys. 68 doi: 10.1016/j.jastp.2005.02.031.

Wu, Q., Ortland, D.A., Killeen, T.L., Roble, R.G., Hagan, M.E., Liu, H.-L., Solomon, S.C.,Xu, J., Skinner, W.R., Niciejewski, R.J., 2008. Global distribution and interannualvariations of mesospheric and lower thermospheric neutral wind diurnal tide:1 Migrating tide. J. Geophys. Res. 113 doi:10.1029/2007JA012542.

Yamashita, C., Liu, H.-L., Chu, X., 2010. Responses of mesosphere and lowerthermosphere temperatures to gravity wave forcing during stratosphericsudden warming. Geophys. Res. Lett. 37, L09803 doi:10.1029/2009GL042351.

Yuan, T., She, C.-Y., Krueger, D.A., Reising, S., Zhang, X., Forbes, J.M., 2010. Acollaborative study on temperature diurnal tide in the midlatitude mesopauseregion (41N, 105W) with Na lidar and TIMED/SABER observations. J. Atmos.Sol. Terr. Phys. 72, 541–549 doi:10.1016/j.jastp.2010.02.007.