The study examines the source of a cold temperature perturbation observed in the Wind Imaging Interferometer (WINDII) daytime daily zonal mean temperatures at 87 km height during March/April 1993, also seen to some extent in 1995 and 1997 over the latitude range 25°S to 25°N. The perturbation appears centered at 10°S as a departure of 10K to 35K below the semiannual climatological mean, which peaks at equinox. After accounting for tidal contributions to the observations, by adopting tidal parameters determined from the Microwave Limb Sounder (MLS) temperature observations the cold temperature signature still remained in the residual data. The perturbation is a global phenomenon immediately following March equinox and lasts over a period of 2 weeks. The period is marked by the presence of mesospheric temperature inversions over the altitude range of 70–90 km. Harmonic analysis reveals the presence of planetary waves of wave number 1 with periods of 3 days, 4–5 days and 7 days. A cross section of the temperature monthly zonal mean data and month of the year over the period from January 1992 to December 1995 revealed a distinct MSAO, modulated by a QBO with strong cold anomalies in March/April 1993, 1995 and 1997 in excellent agreement with the MSAO easterly phase of correlative MF radar wind observations at Tirunelveli (8.7°N, 77.8°E). It is shown that the observed temperature anomalies can be produced by the residual circulation associated with the wave driving of the MSAO itself.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The equatorial and tropical middle atmosphere is host to two of the most prominent long-term planetary-scale perturbations: the semiannual oscillation (SAO) and the quasi-biennial oscillation (QBO). The existence of the SAO in the tropical stratosphere was first observed by Reed  from examining the seasonal variability of tropical temperatures. Further studies [Reed, 1966] showed that the SAO is also present in the zonal wind and extends throughout the upper stratosphere and the lower mesosphere with a peak near the stratopause (∼50 km). Above 65 km, rocket and meteor radar observations showed that the temperature and wind SAO extend throughout the mesosphere [Groves, 1972; Cole and Kantor, 1975]. The SAO was found to consist of two separate signatures with amplitude maxima at the stratopause and mesopause, respectively and a minimum around 65 km [Hirota, 1978; Hamilton, 1982]. The phase of the zonal wind showed very little variation at this altitude and the SAO at the stratopause was out-of-phase with the SAO at the mesopause.
 Satellite temperature and wind observations revealed that the stratospheric SAO has a seasonal asymmetry with the first cycle (beginning with an easterly phase in the Northern Hemisphere) being stronger than the second cycle (beginning with the easterly phase in the Southern Hemisphere) [e.g., Hitchman and Leovy, 1986; Delisi and Dunkerton, 1988; Garcia and Clancy, 1990]; the difference was attributed to the stronger extratropical wave forcing in the Northern Hemisphere compared to that south of the equator.
 A great part of what is known about the mesospheric SAO comes from satellite and radar observations [e.g., Lieberman et al., 1993; Garcia et al., 1997, and references therein]. Analyzing wind observations from 1992 to 1995 by the High-Resolution Doppler Imager (HRDI) experiment on the Upper Atmosphere Research Satellite (UARS), Burrage et al.  were first to point out the apparent modulation of the easterly phase of the mesospheric semiannual oscillation (MSAO) by the QBO [for more on QBO see Baldwin et al., 2001]. It was suggested that small-scale gravity waves might play a role in the generation of the MSAO. A number of modeling studies have been directed into investigating the nature of this modulation and the role of small-scale gravity and planetary waves in that [Sassi and Garcia, 1997; Garcia and Sassi, 1999; Hagan et al., 1999; Mayr et al., 1999].
 The current presentation extends the work of Shepherd et al. [2004b] which examined the global variability of the mean temperature field in the upper mesosphere employing temperature observations by the Wind Imaging Interferometer (WINDII) [Shepherd et al., 1993] on the Upper Atmosphere Research Satellite (UARS) [Reber et al., 1993] for the period from November 1991 to April 1997. The harmonic analysis of daily zonal mean temperatures at latitudes from 50°S to 70°N revealed a strong SAO at equatorial and tropical latitudes, a strong annual oscillation at middle and high latitudes and a weaker QBO (with a period of 790–830 days) confined predominantly to the equatorial and tropical latitudes. The MSAO at altitudes above about 80 km was marked by temperature maxima during the equinox period and minima during the solstices, a pattern which is out of phase with the SAO observed below that height [Shepherd et al., 2004b, Figures 11–13]. The current study examines the global planetary-scale and interannual variability seen in the WINDII temperature observations particularly during the equinox periods at equatorial and tropical latitudes. Section 2 provides a brief description of the WINDII temperature data set, followed in section 3 by an examination of the variability of the upper mesosphere temperature field at those latitudes. The results obtained are discussed in the context of the current state of knowledge of the MSAO and are summarized in section 4.
2. WINDII Temperature Data
 The WINDII temperature data are derived from the solar Rayleigh scattering radiances observed at the limb by the atomic oxygen green line background filter at a wavelength of 553 nm. The retrieval procedure and the data validation have been discussed elsewhere [Shepherd et al., 2001, 2004a]. The WINDII temperatures from November 1991 to April 1997 provide global coverage from 50°S to 72°N over an altitude range from 65 to 115 km. Orbital constraints and instrument viewing geometry limit WINDII global coverage from 42° in one hemisphere to 72° in the other and the reverse as a yaw of the satellite takes place approximately every 36 days. As a result the latitude range 42°S–42°N is viewed all the time independently of the yaw. The satellite orbit precesses at a rate of 5°/day, which is equivalent to a change in local time by 20 min/day at given latitude, because of the 57° orbital inclination. Thus ∼36 days of observations are required to provide a full daytime local time coverage.
 For the purpose of the present study the temperature data in the height range of 65–95 km are binned in 10° latitude ranges with an overlapping interval of 5° to provide daytime daily zonal mean values from 40°S to 40°N. Each daily zonal mean value corresponds to the mean local time at which the satellite orbit crossed the 10° latitude bin and thus contains the tidal contribution at that particular local time. Although zonal averaging exposes the tidal component by averaging over the planetary-scale variations, it does not isolate the diurnal component from the background values.
3.1. Variability and Tidal Contribution at Equatorial and Tropical Latitudes
 The structure of the temperature field at 87 km at middle and high northern latitudes during the equinox transition [Shepherd et al., 1999] was examined and discussed by Shepherd et al. [2002, 2004a]. However, not much is known about the response of the equatorial and tropical mesospheric temperature field to the equinox transition.
Shepherd et al. [2004b] have shown that the mesospheric temperature field, as observed by WINDII over a period of seven years experiences a significant variability at the equator, both from day to day and from year to year. Examples of this variability are shown in Figure 1 for the latitude range ±15° at 87 km. The composite mean shows a semiannual variability with minima during solstices and maxima during the equinoxes, a pattern consistent with the MSAO as described by Garcia and Clancy . Further, the annual variability of the WINDII climatological mean temperature during the March/April equinox period appears larger than that in the September/October equinox period, while the observations expressed in terms of day of the year, show a distinct March/April temperature perturbation below the climatologic average determined by Shepherd et al. [2004b] (Figure 1, solid red line). It is most prominent in 1993, 1995 and 1997 (Figure 1) with a deviation of ∼30–35K with respect to the mean; the least mean square (LMS) parameters of the annual mean are listed in Table 1 [see also Shepherd et al., 2004b]. As can be seen from Table 1 and is discussed in greater detail by Shepherd et al. [2004b] it is difficult to determine the phase of the annual oscillation at 10°S–20°N on the basis of the available WINDII data while the semiannual oscillation is quite prominent.
Table 1. LMS Fit Parameters to the WINDII Daily Zonal Mean Temperature Climatology at 87 km Height [Shepherd et al., 2004b]
Annual Amplitude, K
Annual Phase, days
Semiannual Amplitude, K
Semiannual Phase, days
194.3 ± 6.8
4.2 ± 0.9
204.8 ± 5.6
2.4 ± 0.9
123.5 ± 18.0
192.8 ± 7.1
2.7 ± 1.0
212.0 ± 7.2
3.6 ± 1.0
117.7 ± 2.6
191.6 ± 7.8
0.5 ± 0.1
130.6 ± 10.9
4.0 ± 1.1
103.5 ± 3.5
192.3 ± 8.1
1.3 ± 0.5
341.4 ± 59.2
3.7 ± 1.1
102.3 ± 4.3
193.1 ± 8.0
1.9 ± 0.2
310.6 ± 102.8
4.4 ± 1.0
98.4 ± 4.1
194.9 ± 8.3
2.2 ± 0.1
313.8 ± 80.7
4.1 ± 0.9
92.1 ± 5.8
196.3 ± 8.5
2.2 ± 0.4
297.5 ± 203.8
4.9 ± 0.9
91.7 ± 4.9
197.3 ± 8.5
3.6 ± 0.2
307.6 ± 64.1
4.4 ± 1.0
97.6 ± 4.4
198.3 ± 9.2
3.8 ± 0.1
318.9 ± 43.4
4.0 ± 1.3
110.7 ± 1.9
198.8 ± 10.6
3.1 ± 0.1
313.4 ± 79.0
4.6 ± 1.6
111.6 ± 1.3
199.5 ± 8.5
4.2 ± 0.6
347.5 ± 17.0
2.6 ± 1.2
123.5 ± 21.5
 At both sides of the equator during the first half of the semiannual cycle the day-to-day and year-to-year variability are stronger as the main contributions come from the perturbation seen in the observations in 1993 (blue circles) and the available data from 1995 and 1997 (red and purple circles, respectively). This leads to the somewhat warmer second cycle centered at September equinox.
 The latitudinal development of the perturbation during the March/April equinox 1993 (days 60–120) over the range from 25°N to 25°S is illustrated in Figures 2a and 2b, where the left column is for the Northern Hemisphere, and the right one for the Southern Hemisphere.
Figure 2a gives the temperature perturbation observed in 1993 around March/April equinox from 10°S to 10°N. In the Northern Hemisphere (Figure 2a, top and left column) the daily zonal mean temperature (solid circles), decreases by as much as 25K (from 195K to 170K) at the equator within 3 days (days 77–79) (18–20 March) and remains low, ∼165–170K over about 8 days following March equinox, 21 March (day 80), after which it gradually returns to high values in the second half of April (after day 100, 10 April). The WINDII yaw took place on 19 March (day 78), so no observations were considered from that day. There are no data between days 93 and 98 (3 and 8 April) as well, at these northern latitudes. As the latitude increases away from the equator the amplitude of this signature decreases, becoming weaker as if it “fills in” until it reaches about 25°N, where it is of the order of 10K (Figure 2b, left column). Between 25°N and 35°N (not shown here) the perturbation gradually disappears as the temperature becomes comparable with the climatologic mean before another perturbation in the form of temperature enhancement above the annual mean, starts developing at middle and high northern latitudes as was observed and discussed by Shepherd et al. [2002, 2004a].
 In the Southern Hemisphere (Figures 2a and 2b, right column) the temperature perturbation is very well developed from 5°S to about 20°S, over a period of 30 days from day 71 (12 March) to day 102 (12 April), with a minimum on days 87–88 (28–29 March) and an amplitude of ∼35K. The latitudinal development of the temperature perturbation shows that it is not symmetrical with respect to the equator, but is centered at ∼10°S. The fact that the amplitude of the perturbation observed decreases with latitude away from the equator suggests that the feature might be at least partially associated with effects of the thermal diurnal tide.
 At this initial presentation of the WINDII data as daytime daily zonal mean temperatures the diurnal tidal contribution has not been accounted for. Because the measurements for a given latitude bin on a specific day are made at the same local time the zonal mean, migrating diurnal and semidiurnal tides all contribute to the zonal average. Accounting for the migrating diurnal tide contribution at equatorial latitudes during equinox is important since as theory predicts these are the spatial and temporal conditions for the largest tidal amplitudes in temperature [e.g., Forbes, 1995]. The problem of the observed signatures being biased by the thermal tides always exists as the data cover only half of the diurnal cycle. Shepherd and Fricke-Begemann , employing WINDII daytime- and potassium-lidar nighttime temperature data, showed that this shortcoming of the data set can be alleviated if collocated and correlative temperature data are combined, from which the tidal information can be retrieved. Shepherd et al. [2004a] extended the approach to the comparison of WINDII and OH rotational temperatures to determine the diurnal and semidiurnal tidal parameters at midlatitudes (50°N) and 87 km height, before being removed from the data set to study the longitudinal variability of the temperature field at those latitudes. This approach cannot be applied here as there are very few ground-based temperature observations available in the 15°S–15°N latitude region and they either do not reach the altitude of 87–90 km or do not cover the 1992–1997 observation period. Fortunately, accounting for the diurnal and semidiurnal tidal contributions became possible thanks to the very recent results of J. F. Forbes and D. L. Wu (Solar tides as revealed by measurements of mesospheric temperature by the MLS experiment on UARS, submitted to Journal of the Atmospheric Sciences, 2005, hereinafter referred to as Forbes and Wu, submitted manuscript, 2005) employing temperature data from the Microwave Limb Sounder (MLS) experiment on the UARS. Thus the MLS tidal parameters for March/April at 70 to 86 km height and at 25°S–25°N have been adopted in the current study.
 The tidal contribution to the daily zonal mean temperatures for the respective local times, according to the MLS observations was calculated and subtracted from the WINDII data. The residual daily zonal mean temperatures thus corrected at 87 km are also shown in Figures 2a and 2b, as squares. The correction changed the magnitude of the perturbation at the different latitude ranges by further decreasing the amplitude of the signature observed at ±10°. Even after the tidal correction the perturbation remains quite distinct and persistent. The local time between day 77 and day 81, when the abrupt decrease in the daily zonal mean values occurs, changes from 14h to 10h. As was mentioned the WINDII yaw takes place on day 79, 19 March 1993. If we assume that the diurnal tide is the only source of the perturbation observed than it implies that the amplitude should change by 25K over 4 hours of local time between days 77 and 81. To our knowledge there is no experimental or theoretical evidence that the migrating tide can cause such a change. This leads to the suggestion that there is some additional source contributing to the perturbations observed.
3.2. Equatorial Temperature Maps
 The development of the March equinox perturbation observed in the ±40° latitude range can also be seen in the daily global latitude/longitude maps at 87 km, shown in Figure 3 (not corrected for the tide), for the period of 12–28 March 1993. The time sequence progresses down and across the plot. A similar presentation was first introduced by Shepherd et al.  at 85 km height (their Figure 5) using an earlier version of the WINDII temperature data. In general the maps at 87 km height show a more variable global field with a series of crest/trough structures centered at about 10°S. By 17 March a single enhancement has developed at about 0° longitude centered at the equator with a slight tilt from west to east (from the Southern Hemisphere to the Northern Hemisphere) extending to about ±30° latitude around the equator. With the WINDII yaw taking place on 19 March, the next available day of temperature observations is 21 March (day 81), March equinox, when the temperatures of the entire equatorial and tropical region (±20°) appear very low, at ∼160K, and are comparable with midlatitude summer mesopause values. On the following days the equatorial region remains considerably colder than the MSIS (Mass-Spectrometer-Incoherent-Scatter: MSIS model, e.g., Hedin ) predicted climatology for these latitudes (not shown here) and even indicates a few local minima with temperature below 160K (Figure 3). To examine further the pattern of the March equinox perturbation the MLS-tide-corrected WINDII temperatures were also mapped in day-of-year/longitude fashion and some of the results at the equator, 5°N and 10°N are shown in Figure 4. At 5°N a four-maxima signature can be seen at the beginning of March, around days 65–70 (5–10 March) at about 50°E, 110°E, 210°E (150°W) and 320°E (40°W), with the maximum at 210°E (150°W) being the strongest (Figure 4, middle). At 10°N (Figure 4, bottom) the maximum at 120°E has almost disappeared, while the maximum at 50°E is the strongest while the other two maxima are still present but reduced in magnitude.
 Following March equinox (day 81, 21 March onward) there is a decrease in mesospheric temperature at 87 km all across the globe on both sides of the equator, between day 85 (25 March) and day 90–95 (31 March to 5 April). At the equator, for example, a series of local minima have formed around days 81–87 (21–27 March), replacing the maxima seen prior to equinox and the observed temperature decrease is extended both globally and temporarily well into April.
3.3. Mesospheric Temperature Inversions and Planetary Waves
 In studying the global variability of the mesospheric temperature field Shepherd et al. [2004b] report a phase reversal of the semiannual variability during equinox at equatorial and tropical latitudes between 75 km and 87 km. The vertical temperature profiles from 70 km to 90 km at ±15° latitude range for March/April 1993 (when the perturbations around March equinox are the greatest) showed a considerable range of temperature variability from day-to-day in the magnitude and shape of the observed profiles. Profiles prior to 17 March indicated a mesopause situated above 90 km consistent with the winter state of the mesopause, while the profiles from 1 April on exhibit a minimum around 85–87 km, similar to the mesopause summer state [e.g., She and von Zahn, 1998]. Between 17 and 31 March 1993 the temperature profiles show a distinct mesospheric temperature inversion (MTI) between 77 and 87 km. The MTI minimum and maximum are separated by ∼7 km, while the temperature difference between them is ∼25–30K, both consistent with lidar and falling sphere observations [e.g., Leblanc and Hauchecorne, 1997; Lübken et al., 1994; Meriwether and Gardner, 2000, and references therein]. All profiles intersected around 80–82 km suggesting that at this altitude the temperature field remains more or less isothermal over the period of consideration independently of the different local times of the individual observations or the source of the temperature inversion observed. The temperature inversion layer observed between 77 km and 84 km contributed to the variability seen at 75 km and 87 km in the annual temperature cycle at equatorial latitudes and remained present even after the MLS-tidal correction, as can be seen in Figure 5. The presence of a MTI seen from 15°S to 15°N (not shown here) indicates the large horizontal extent of this phenomenon on both sides of the equator.
 The planetary wave structure (corrected for tidal contributions according to the MLS temperature observations, Forbes and Wu (submitted manuscript, 2005)), seen in Figures 4 and 5 are further analyzed by employing Lomb-Scargle periodograms [Schultz and Stattegger, 1997]. At the equator (5°S–5°N) harmonic spectral analysis, revealed the presence of wave 1 (wave number 1) with periods of 4.5–5 days and 7 days with a confidence level of 95% (Figure 6a). Another perturbation with a period of 2.5 days is also present. Similar signatures are observed also at 5°N (0°–10°N) (Figure 6b) where a strong planetary wave 1 with a period of 7 days is seen; a 5-day wave is also identified, with spectral power comparable to that seen at the equator. By 10°N (Figure 6c) the wave 1 with period of 3 and 4.5 days have become the dominant planetary wave while the 7-day wave spectral power is well below the 95% significance level (Figure 6c).
 South of the equator, at 5°S and 10°S, where the center of the cold temperature anomaly is located, the 7-day wave 1 continues to be the dominant planetary wave, while the 4–5 day wave gradually weakens (Figure 6d and 6e). In general, the magnitude of the 7-day wave was found to be the strongest at 5°N but decreases with latitude away from the equator in the Northern Hemisphere, while it remains almost constant from 5°N to 15°S; the 4.5–5 day wave appears to be a persistent feature at the 5°S–15°N latitude range.
 In spite of the correction for tidal contribution the periodograms indicate the presence of short period perturbations with periods of ∼1 day or less, suggesting that tidal contribution still has not been fully accounted for. This result is not surprising in view of the study of Salby and Callaghan [1997, Figure 12b], which showed that the asynoptic sampling by a single satellite platform with orbital parameters like those of UARS can introduce an error of 10–20% in the determination of the diurnal tide parameters at equatorial latitudes. However, these tidal signatures are well below the 95% confidence level. In all spectral calculations linear trends have been subtracted. The harmonic analysis employed gives only the period of the oscillations, but cannot determine the direction of their propagation.
3.4. MSAO/MQBO Effects
Shepherd et al. [2004b] also identified the presence of a QBO in the mesospheric temperatures from 70 to 90 km with a period of 790–830 days. The dynamical aspects of the QBO are best illustrated in a time/height cross section of the monthly zonal mean temperature field. These monthly zonal mean temperatures from January 1992 to December 1995 at the equator are shown in Figure 7. The gap in the data indicates the lack of observations for the period of March to December 1994. As there are no data from October 1993 the mean of the September and November 1993 monthly averages has been assigned to replace the missing October observations. A similar problem exists for April and May 1995, when there are no temperature data. In the latter case the missing profiles have been replaced with the mean of the April and May monthly averages from 1992 and 1993. The interannual variability of the monthly mean zonal temperature field is marked by alternating bands of warm and cold temperature anomalies, which progress downward with a temperature gradient of 2.5 km/month. At altitudes below ∼80 km the semiannual oscillation in the WINDII temperatures, first discussed by Shepherd et al. [2004b] is revealed with maxima at the solstice and minima at the equinox periods. Around 80–82 km the phase reverses and above that height the pattern of warm temperature anomalies is replaced by cold temperature anomalies forming a downward progression characteristic of the MSAO [Garcia and Clancy, 1990; Garcia et al., 1997; Garcia and Sassi, 1999].
 A particularly strong cold anomaly is observed during March equinox 1993, which extends through the entire upper mesosphere from about 72 km to at least 90 km height with a minimum at ∼79 km. A similar cold anomaly is observed during March equinox in 1995, although it is somewhat weaker than that in 1993 and is confined to altitudes around 80 km. A second broader cold anomaly is observed during the 1995 summer months extending to 90 km. Because of the way the data for April/May 1995 were obtained it is not possible to determine whether these two cold anomalies are connected or are independent signatures. The strong cold March anomalies from March 1993 to June–July 1995 have a period of 27–28 months consistent with the QBO signature with a period of 790–830 days reported by Shepherd et al. [2004b].
 The monthly averaging of the WINDII temperature data (without the MLS correction for tidal contributions) has reduced aliasing by the diurnal tide depending on the phase of the latter with respect to the range of the local times observed during the averaging period. However, the possibility of tidal contamination must be borne in mind when examining the data above 85 km [e.g., Garcia et al., 1997]. Semiannual variability of tidal amplitude has been observed in the wind tides [Burrage et al., 1996; Lieberman and Hays, 1994] although this has not been reported with respect to temperature.
 To understand better the source of the temperature MSAO observed we need to examine the wind field of the background atmosphere. Recent results on the semiannual oscillation of zonal mean wind in the upper mesosphere by Sridharan et al.  provide an excellent opportunity for comparison with the WINDII temperatures. Figures 8a and 8b show a comparison of the WINDII zonal mean temperature profiles for the period of January 1992 to December 1995 at 10°N and the monthly zonal mean winds over Tirunelveli (8.7°N, 77.8°E) from January 1993 to December 1995. At 10°N (Figure 8a) a pattern similar to that at the equator (shown in Figure 7) is observed with enhanced cold anomalies during spring equinox 1993 and 1995 and a cold summer/fall anomaly in 1995. There is excellent agreement between the MSAO pattern seen in the two monthly zonal fields, in particular the strong easterly wind shear anomaly and the strong cold temperature anomaly both seen above 80 km in March/April 1993 and 1995. There is also very good correspondence between the westerly phase of the monthly zonal mean winds and the warm temperature anomalies below ∼80 km. Both data sets show somewhat weaker easterly shear/cold anomalies in 1995 compared to 1993. The QBO seen in the two data sets (about the wind QBO, see also Rajaram and Gurubaran ) is coupled with the MSAO which leads to the pattern observed. There is also a very good correspondence between the WINDII temperature MSAO and the HRDI monthly mean zonal winds at the equator [e.g., Garcia et al., 1997, Figure 9a; Lieberman and Riggin, 1997, Figure 1], with a much stronger response of the easterly zonal mean winds to the MSAO/MQBO coupling during equinox than for the westerly winds.
 Observations of the wind and temperature MSAO [Hirota, 1978, 1979; Hamilton, 1982; Garcia and Clancy, 1990; Lieberman et al., 1993; Garcia et al., 1997] have shown that the easterly phases of the wind MSAO lie almost directly above the westerly extrema of the stratospheric wind SAO. Dunkerton  modeled this behavior and attributed it to vertically propagating gravity waves and ultrafast Kelvin waves. The modeling results showed that the time-mean easterlies exist in the upper mesosphere, while the absolute westerlies at these altitudes are weak or nonexistent.
 The WINDII monthly zonal mean temperatures showed the normal MSAO pattern described by Garcia et al.  with maxima at solstice below 80 km and at equinox above this height. The MSAO amplitude at December solstice (March equinox) was larger than that at June solstice (September equinox), respectively indicating a strong first cycle (beginning with the warm temperature anomaly in January at ∼70 km) and weaker second cycle of the MSAO. In addition to the MSAO, enhanced cold anomalies in the first cycle in 1993 and 1995 in March–April above 80 km were also observed while there was little change in the warm anomalies below ∼80 km, a pattern suggesting modulation by the mesospheric QBO.
Garcia et al.  observed that in the stratospheric SAO the westerlies (which will have a maximum) descend south of the equator causing the asymmetry in the stratospheric SAO and suggested that the descent is affected by small-scale gravity waves (horizontal wavelength ∼100 km). The asymmetry in the temperature cold anomaly, centered at 10°S, illustrated by the plots in Figures 2a, 2b, and the global temperature maps in Figure 3, and its association with easterly wind shear, illustrated by the WINDII/MF radar comparison, imply an asymmetry in the wind MSAO consistent with the observations of Garcia et al.  and the indirect effect of small-gravity waves on the temperatures observed. However, this possibility has not been directly examined in the current study.
 Although there is evidence that both the SAO and QBO are caused by interactions between the mean flow and vertically propagating planetary waves, the nature of these waves and their excitation mechanisms remains uncertain. Modeling studies by Sassi and Garcia  and Garcia and Sassi  with regard to the stratospheric SAO suggested that planetary-scale (wave number 1–3) Kelvin waves are not sufficient to drive the stratospheric easterly phase as suggested by Hitchman and Leovy ; however, little is known about the waves producing the MSAO.
 Similar to the MSAO, the mesospheric QBO has also been observed to be out of phase with its stratospheric signature [e.g., Baldwin et al., 2001, Plate 6]. Simulations with the GCM with parameterized gravity waves suggest that the mesospheric QBO is driven by selective filtering of small-scale gravity waves by the underlying wind through which they propagate [Burrage et al., 1996; Mayr et al., 1999]. Hagan et al.  also showed that the mesospheric QBO signature is characterized by pronounced peaks centered at ∼85 km during April, which is consistent with results reported by Garcia and Sassi  reinforcing the evidence of MSAO/QBO coupling both in monthly zonal mean winds and temperature data.
 Another signature observed in the WINDII temperatures during the period of the cold temperature anomaly was the presence of MTI, as was shown in Figure 5. Recent modeling results of mesospheric temperature inversion involving the Whole Atmospheric Community Climate Model (WACCM) by Sassi et al.  showed that episodes of wintertime planetary wave amplification and gravity wave breaking can produce strong MTI with amplitudes of 20–25K and a vertical extent of 10 km. According to the model the inversion is a direct result of the rapid dissipation of planetary waves in the mesospheric surf zone. The MTI were found to occur in phase with the cold temperature anomaly in the upper mesosphere, which constitutes the base of the inversion layer with an amplitude approaching 10K. The WACCM results indicate that although gravity waves did not play a direct role in the MTI produced they were a dominating force in the mesospheric momentum budget and played an essential indirect role in setting up a critical line allowing quasi-stationary Rossby waves to propagate into the upper mesosphere surf zone, where they dissipate causing the MTI. A scenario relevant to the results obtained in this study and based on the model results by Sassi et al.  may be that the observed temperature inversions are associated with the propagation of small-scale gravity waves at March equinox, during the transition from winter to summer, as indicated by the MF radar wind data. These gravity waves weaken the wind field further thus creating conditions for planetary waves to propagate from below and reach the upper mesosphere, where they dissipate. However, in spite of the evidence that traveling planetary waves were present during the cooling event in March/April 1993 these waves cannot explain the apparent cyclic nature of the heating and cooling anomalies shown in Figures 7 and 8 and their downward phase progression with time.
 Another plausible explanation of the warm and cold anomalies could be related to the thermal anomalies produced by the residual circulation associated with the wave driven vacillation of the equatorial zonal winds [Andrews et al., 1987]. As can be seen in Figures 7 and 8, negative (positive) temperature anomalies are associated with the westward (eastward) vertical shears of the MSAO, which is what would be expected of the effect of a residual circulation on temperature. An estimate of the temperature anomaly due to the zonal mean vertical shear can be made employing the relationship δT ∼ L2HβR−1uz [Andrews et al., 1987, equation 8.2.1], based on the thermal wind equation on an equatorial β plane. Here L is the width of the meridional scale of the semiannual oscillation (∼15° or 1500 km).
 For this estimate we use the HRDI monthly averaged zonal mean zonal wind data at the equator for the period of August 1992 to March 1995 [e.g., Lieberman and Riggin, 1997, Figure 1]. As the HRDI sampling of the 60–110 km is confined to the daytime hours, with the exception of 95 km, where nighttime emissions are also sensed, the migrating tides cannot be filtered from the daily zonally averaged wind over the vertical domain below ∼90 km [Lieberman, 1998]. Thus the HRDI zonal mean zonal wind sampling is similar to that of the WINDII zonal mean temperatures. According to the HRDI wind observations at the end of March 1993 the westward monthly averaged zonal mean zonal winds have a maximum of ∼95 ms−1 at about 84 km height. Below that altitude the vertical wind shear is ∼10 ms−1 km−1. Using these estimates for L and uz gives an anomaly δT ∼ 13K. Unfortunately, there is a gap in the monthly zonal mean winds at Tirunelveli in late March 1993 and we cannot access directly the magnitude of the zonal wind shear. However, the contours suggest zonal wind of the order of ∼90 ms−1 at altitudes of ∼86–87 km, which also leads to a zonal wind shear of ∼10 ms−1 km−1 and consequently to the same δT as at the equator. The same peak height is seen also in March 1995. While the vertical shears of the zonal wind uz are practically the same, the peak altitude at Tirunelveli is higher than that seen in the satellite observations by about 3 km. The estimated δT is about a factor of 2 less than the amplitude of the perturbation seen at the equator in Figures 2a and 2b before correcting for tidal contribution. However, keeping in mind that the estimate of zonal wind vertical shear is quiet crude and adopting a larger-scale value for the meridional scale in the upper mesosphere [Hitchman and Leovy, 1986], of the order of 1700 km will bring the estimated temperature anomaly in line with the observations.
 The harmonic analysis performed identified the presence of planetary wave 1 with periods of 3 days, 5 days and 7 days which dominate in the WINDII mesospheric temperatures at 87 km and at ±10° latitude around the equator, although the direction of their propagation could not be determined. Planetary waves, propagating eastward with wave numbers 1, 2 and 3 and a period of 3 days, consistent with ultrafast Kelvin waves at the equator were observed in the HRDI [Lieberman and Riggin, 1997] and radar [Vincent and Lesicar, 1991; Vincent, 1993; Riggin et al., 1997; Kovalam et al., 1999; Tsuda et al., 2002] wind measurements. The presence of 3–4 day oscillations at equatorial latitudes maximizing at the equinox season were observed also in zonal winds (3–3.8 days) [Yoshida et al., 1999] and mesospheric airglow emissions and temperature (3.5 days) [Takahashi et al., 2002].
 MF radar wind observations at equatorial latitudes also revealed the presence of a 6.5-day wave [e.g., Vincent and Lesicar, 1991; Vincent, 1993; Kovalam et al., 1999; Kishore et al., 2004] whose amplitude also maximizes during equinox, but appears out of phase with the strong MSAO/QBO cycle seen in March 1993, 1995 and 1997. Cross-spectral and harmonic analyses have shown that the 3.5-day and 6.5-day waves propagate in opposite directions in the zonal direction suggesting that they are of a different type and have different origins. Planetary waves with a period of 6.5-day observed by Kovalam et al.  and Kishore et al.  appeared to have characteristics more consistent with an unstable mode of a westward propagating wave number 1 resulting in response to in situ momentum forcing at high latitudes in the Southern Hemisphere. Additional support that this wave is an unstable perturbation has also been given by Meyer and Forbes  and Lieberman et al. . Evidence of the 5-day and 6.5-day waves were also identified in the HRDI wind and temperature data [Wu et al., 1994; Talaat et al., 2001] and were found to appear frequently during equinoxes at the tropics closely resembling the structure of the gravest symmetric wave number 1 internal Rossby wave (1,1) mode.
5. Summary and Conclusions
 In conclusion, the investigation of the source of the large temperature cold anomaly observed in WINDII daily zonal mean temperatures at equatorial and tropical latitudes during the March/April equinox in 1993, 1995 and 1997 has led to the following findings:
 1. The perturbation appeared as a temperature decrease of 25–30K following March equinox and lasted about 2 weeks over the latitude range from 25°S to 25°N.
 2. The perturbation was a global phenomenon asymmetrical with respect to the equator and centered at 10°S.
 3. Corrections for tidal contribution based on the MLS temperature observations at 86 km were in excellent agreement with the WINDII observations and led to a further increase in the magnitude of the temperature anomaly observed.
 4. Mesospheric temperature inversions and planetary wave number 1 with periods of 3 days, 5 days and 7 days were identified during the March/April period of 1993. The 7-day planetary wave is the dominant planetary-scale perturbation seen between 15°S and 5°N.
 5. The temperature monthly zonal mean data revealed a distinct MSAO, modulated by a QBO with strong cold anomalies in March/April 1993, 1995 and 1997 in excellent agreement with the MSAO easterly phase of correlative ground-based wind observations.
 6. It was shown that the observed temperature anomalies can be produced by the residual circulation associated with the wave driving of the MSAO itself.
 The results presented in this study add to the description of the temperature MSAO and its global pattern. The full daytime local time coverage, the consideration of the tidal effect, the simultaneous temperature and wind observations, and the higher vertical resolution (2 km) of the WINDII temperature data compared to those from the Solar Mesospheric Explorer (SME) satellite have allowed us to extend further those observations to present the global onset of the MSAO in the mesospheric temperature field, as observed by WINDII.
 The authors thank J. M. Forbes and D. L. Wu for providing the MLS tidal data and Xiaoli Zhang for her assistance in making these data available. This work has been supported through research grants from the Natural Sciences and Engineering Research Council of Canada and the Canadian Foundation for Climate and Atmospheric Science. One of the authors (M.S.) would also like to acknowledge the support of the Japan Society for the Promotion of Science, and both M.S. and G.S. are grateful for support provided by the Research Institute for Sustainable Humanosphere of Kyoto University, Japan. The authors thank R. A. Vincent and the two anonymous referees for their helpful comments and constructive criticisms.