G. G. Shepherd, R.G. Roble, 2 S.-P. Zhang, 3 C. McLandress, 3 R.H. Wiens

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. A7, PAGES 14,741-14,751, JULY 1, 1998 Tidal influence on midlatitude airglow: Comparison of satellite and ground-based observations with TIME-GCM predictions G. G. Shepherd, R.G. Roble, 2 S.-P. Zhang, 3 C. McLandress, 3 R.H. Wiens Abstract. WINDII, the Wind imaging interferometer on the Upper Atmosphere Research Satellite measures winds and emission rates from selected excited metastable species. Measurements of the 558-nm emission from atomic oxygen provide both the wind from the Doppler shift, and the atomic oxygen concentration from the emission rate. Thus the tides and their influence on atomic oxygen are measured with the same instrument. Ground-based airglow measurements provide vertical integrals of the same quantities and coordinated observations were obtained between WINDII and ground-based instruments at Bear Lake (42.5øN, 212øE) for O( 1 S) 558 nm winds and O2(b lz) (0,1) band emission rate and temperature. The TIME-GCM model has recently incorporated airglow photochemistry, so that direct comparisons may be made with airglow observations, without inverting those observations to atomic oxygen distributions. In this study, the influence of tides on airglow emission at midlatitude is studied through the comparison of the above data sets with the TIME-GCM model, extending earlier studies conducted for the equatorial region. At the vernal equinox the upward propagating diurnal tide is found to be the dominant influence on airglow diurnal variation. At solstice the diurnal tide does not penetrate to as high an altitude, so that the domi- nant influence is then the in situ semidiurnal tide. This conclusion is consistent with both WINDII observations and TIME-GCM predictions, whose data sets agree extremely well with one another. The ground-based results agree well in the local time variation pattern, but the amplitudes observed are larger than for WINDII or the TIME-GCM by roughly a factor of 2. This difference illustrates very clearly the differences between a tidal pattern observed at a single site for a few nights, and a global pattern that is first zonally averaged, and then combined in local time over about 1 month, as is done with the WINDII data. The agreement of these average data with the TIME-GCM model strongly suggests that they accurately representhe behavior of the zonally averaged atmosphere. 1. Introduction WINDII, the Wind imaging interferometer on the Upper Atmosphere Research Satellite (UARS), was launched on September 12, 1991, and has now operated for more than 6 years in orbit. The instrument measures winds, temperatures, and emission rates from a variety of airglow emissions [Shepherd et al., 1993a] that cover the altitude range 80 to 300 km. Winds obtained from the atomic oxygen O(1S) emission at 558 nm have been validated [Gault et al., 1996a] and as part of this validation, ground-based instruments from the Centre for Research in Earth and Space Technology (CRESTech) were operated at Bear Lake during the winters of and The thermosphere-ionosphere-mesosphere-electrodynamics general circulation model (TIME-GCM) [Roble and Ridley, 1994] is now able to predict airglow emission rates, as well as winds, temperatures, and atmosphericharacteristics. This has made possible combined studies involving O(1S) satellite observations, ground-based measurements, and model predictions. In this paper we describe such a study of the tidal influence on airglow at midlatitudes. The observation of tides from a ground station appears straightforward, as a single day of observations is sufficiento describe the local time variation over 24 hours. The problem is that such observationshow considerable variability from day to day, so that it is difficult to extend such results to describe the global scale features of the tide. A satellite such as UARS has the advantage of complete longitudinal coverage, solving to some extent the problem of local variability. However, a new problem appears in that the UARS satellite precesses at a rate of 20 min/d, requiring about 36 days to cover 24 hours of local time, and the analysis of 36 days of data to obtain the local time variation of the wind field. One can well ask how well the tidal variation obtained in this way represents the global tidal field. The effects of aliasing have been discussed by McLandress et al. [1996b]; 1Centre for Research in Earth and Space Science, York University, the conclusion is that the zonal averaging of the satellite data Toronto, Canada. essentially removes the variability observed at a single site, so 2High Altitude Observatory, National Center for Atmospheric Research Boulder, Colorado. that the average data do well represent the tidal wind field. Thus 3Centre for Research in Earth and Space Technology, Toronto, Canada. they are directly comparable to the predictions from the TIME- GCM. The good agreement obtained in the current study between WINDII and the TIME-GCM model very strongly supports this Copyfight 1998 by the American Geophysical Union, conclusion. Comparisons of these averaged results with a ground Paper number 98JA station, such as Bear Lake, then provide a measure of local /98/98JA variability. 14,741

2 14,742 SHEPHERD ET AL.' TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW The global tidal structure of the atmosphere as seen by WINDII has been described by McLandress et al. [1994, 1996a] and is illustrated in Figure l a where we show the meridional wind as a function of latitude, for a local time of 12 hour [from McLandress et al., 1996b]. Here the WINDII data have been extended to lower altitude by combining them with data from the UARS high-resolution Doppler imager (HRDI), which measures winds from the 0 2 atmospheric band emission [Hays et al., 1993]. The unshaded areas are contours of northward wind, and the shaded areas correspond to southward wind. This gives an overall view of how the tidal effects vary with latitude. Centered at 20ø-30 ø and 93 km for this particular local time, namely noon, there are cells of meridional wind in opposition across the equator, producing a strong convergence of flow there, and zero wind. One-half wavelength higher, at 104 km, the cells have flows in the opposite directions, so that the winds blow away from the equator on both sides, causing a divergence of flow there. At about 52 ø latitude there is another set of wind cells ,,,,, O 5O LAT I TUDE ( DEG ] Figure la. Meridional winds as a function of latitude as observed by WINDII and HRDI for a local time of 1200 LT at equinox [from McLandress et al., 1996b]. 10- z Figure lb. Meridional winds as a function of latitude as calculated by the TIME-GCM for a local time of 1200 LT at equinox. shifted in altitude with respect to those near the equator; one prominent cell is evident at 111 km. As described by McLandress et al [1994], the subtropical cells at 20ø-30 ø are dominated by the diurnal tide, while the high-latitude cells are manifestations of remarkable diurnal variation of atomic oxygen O(]S) 558 nm emission rate at the equator, which through comparison with the the semidiurnal tide. The latitude of interest for this study, 42% is Forbes [1982a, b] model they attributed to tidal forcing of the located in the transition region between diurnal and semidiurnal atomic oxygen distribution arising from the strong dynamical dominance. Thus one expects to see dominantly diurnal control at effects described above. A similar conclusion was reached earlier these latitudes for these equinox data as described by by Burrage et al. [1994]. The influence on emission rates was McLandress et al. [1996a] but with some contribution from the very large; the peak value of volume emission rate observed by semidiurnal tide as well. The corresponding TIME-GCM WINDII was modulated by a factor of 3, and the altitude change prediction, calculated for equinox conditions is shown in Figure in the volume emission rate distribution was as much as 5 km. lb for comparison, and the agreement is extremely good, though the tidal input to the model has been adjusted as described below. The strongest effects occurred within about 5 ø of the equator. One consequence of the large tides observed is the influence Roble and Shepherd [1997] used the WINDII tidal winds of on species concentrations, and Shepherd et al. [1995] describe a McLandress et al. [ 1996a] to adjust the diurnal tidal input at the bottom of the TIME-GCM in order to obtain global agreement with the wind observations near 100 km. It was found that the tidal input had to be sufficiently strong to overcome the gravity wave dissipation of the tide in the upper mesosphere in order to produce the meridional wind amplitudes observed by WINDII for the spring equinox period. These adjusted tidal inputs were referred to as "strong tides." With these strong tides as input to the TIME-GCM, Roble and Shepherd [1997] compared the predicted green emission line behavior with that observed by WINDII at the equator and found very close agreement. Tidal forcing enhances the emission rate of the evening airglow layer and causes it to descend in altitude, which is consistent with the general relationship between airglow emission rate and altitude as described by Ward et al. [ 1994] and by Yee et al. [1997]. The emission rate then drastically weakens at midnight, but begins to be reestablished again in the morning hours. The TIME-GCM shows that atomic oxygen concentrations from higher altitudes are pushed down to lower altitudes, by about 5 km in the evening hours, until near midnight the emission is nearly extinguished owing to quenching and the reduced supply of atomic oxygen. After midnight, the emission begins to recover, perhaps due to downward winds, and a fresh supply of oxygen from above. The most significant conclusion is that this effect is observed only when the upward propagating diurnal tide reaches the altitude of the airglow emission. We follow a similar approach here in the investigation of tidal influence on airglow at midlatitude. To reiterate, the TIME-GCM model is adjusted as described by Roble and Shepherd [1997] to obtain agreement between the global model and observed wind fields for equinox, and again for to-" 10-

3 SHEPHERD ET AL.: TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW 14,743 11o lo6 v lo2 i- 98 ß i ß ß ß i ß ß ß i ß ß - 0,,.,\ lo -10 '.. ',,',, - 0',,,, \ v LOCAL SOLAR TIHE (H) LOCAL SOLAR TIHE (H) Figure 2a. Meridional winds observed by WINDII for the period Figure 2c. Meridional winds at 45øN observed by WINDII for March/April 1992/1993, for a latitude of 40øN, as a function of December /93 and January 1993/1994 as a thnction of local time at solstice. local time at equinox. The white vertical band is a result of missing data. Donahu et al. [1973] noted a strong semiannual variation in the green line, with the largest mean emission rates in April and solstice. The emission rates are obtained from the model with no October. Cogger et al. [1981] obtained a large data set with the further adjustment, so that the model and observed emission rates ISIS-II satellite and confirmed the large amplitude of the green are independent of one another. line semiannual variation at midlatitudes. Garcia and Solomon [1985] used model results to show that the semiannual behavior 2. Previous Measurements of Midlatitude of breaking gravity waves could explain this variability in emission rate. Petitdidier and Teitelbaum [1977, 1979] provided Airglow Variability the first definitive interpretation of midlatitude oxygen airglow The measurement of airglow variability began with its earliest observations, as it was a feature which caughthe attention of the first investigators [Rayleigh, 1931]. A survey of papers on emission in terms of tides, while Akmaev and Shved [ 1980] were the first to apply dynamical models to the prediction of airglow emission rate perturbations. airglow variability published between 1970 and 1987 has been presented by Forsyth and Wraight [1987], from which it is apparenthat the most common type of diurnal variation was a maximum near midnight. Brenton and Silverman [1970] noted that this was true only at mid latitudes, and that at low latitudes a 3. Wind Comparisons In Figure 2a we show the WINDII observations relevant to the latitude of Bear Lake, namely the local time variation of the midnight minimum was more common. This midnight equatorial meridional wind for a latitude of 40øN at equinox, using data minimum was confirmed with WINDII observations, as from March/April In Figure 2b we present the described in detail by Shepherd et al. [1995] and noted earlier. equivalent data for the TIME-GCM for equinox. The patterns are Midlatitude oxygen airglow had already attracted attention similar in magnitude but slightly different in phase, indicating because of its strong annual variation. Using OGO satellite data, 10-4, 102 ' 'z X,,,, 94 "" O UT (HOURS) (MTIMES ,00 TO ,001 UT (HOURS) (MTIMES ,00 TO ,00), m, I I I I 20. O0 O. O0 4. O0 8. O0 LOCAL T[HE (HRS) (LON= ),,, I I I I 20. O0 O. O0 4. O0 8. O Figure 2b. TIME-GCM meridional wind predictions for a latitude of 42øN as a function of local time at equinox. Figure 2d. TIME-GCM meridional wind predictions for a latitude of 42øN as a function of local time at solstice.

4 14,744 SHEPHERD ET AL.' TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW LST (c OUrS) O o UT (hours) Figure 3a. Bear Lake vector winds in m s -1 observed by GWAMDI for the night of April 7, that the model adjustment is reasonably compatible with the WINDII wind values. There is a phase shift in the observed winds above 98 km that appears to be associated with a shift from a diurnal pattern below 98 km to one that is semidiurnal above. The same effect is seen differently in the TIME-GCM results, as a convergence of the wind contours with increasing altitude that is also consistent with a change from a more diurnal pattern at the bottom to a semidiurnal pattern above. In Figures 2c and 2d respectively we show the corresponding WINDII observations for December/January and the TIME- GCM predictions with solstice tides. Here the agreement between WINDII and the TIME-GCM is very good in that both show a clear semidiurnal pattern, with a slight phase shift between the observations and the predictions. In contrasto the highly averaged satellite data we show Bear Lake ground-based vector wind measurements from a single night, April 7, 1992, in Figure 3a. These were obtained by a Doppler Michelson interferometer called GWAMDI (ground wide angle Michelson Doppler interferometer), which is a ground-based version of an instrument originally designed for space shuttle flight [Shepherd et al., 1985] and was the precursor of an advanced Doppler Michelson instrument called ERWIN (E region winds), which has been fully described elsewhere [Gault et al., 1996b]. The ground-based results for winds obtained from the O(1S) 558 nm emission show a moderate (about 40 m s -1) southward wind, early in the evening, falling to low values after 0800 UT (0100 LT), and then Oust barely) reversing to northward at 1000 UT. The TIME-GCM prediction for a groundbased atomic oxygen green line vector wind, computed by adding elementary Doppler line profiles from each airglow layer, each UT (hours) Figure 3c. Bear Lake vector winds in m s -1 observed by GWAMDI for the night of January 27, with the temperature, wind, and emission rate appropriate to that layer, is shown in Figure 3b for equinox conditions. Here we also see a southward wind, reaching its maximum amplitude at UT ( LT), consistent with the Bear Lake observations, and thereafter reducing in magnitude to zero at about 0800 UT (0100 LT), but turning much more strongly northward at 0900 UT. Thus the patterns agree for part of the night, though not for all, and the predicted maximum wind amplitude is about 25 m s -1 a factor of 2 smaller than the observed value for this particular night. Bear Lake winds for a single night at solstice are shown in Figure 3c for January 27, The meridional component is now much smaller than the zonal component with values that are near-zero between UT ( LT), and again at 1000 UT (0300 LT), with a southward component of m s -1 between these times. The corresponding TIME-GCM prediction is shown in Figure 3d, in which a pronounced semidiurnal pattern is evident with a wind amplitude of about 25 m s -1. The predictions are not consistent with the winds measured on this particular day, but they are consistent with the zonally averaged wind pattern shown in Figure 2c. In harmony with the results of Figures 2a and 2b, the vector winds of Figures 3a and 3b for equinox show a more diurnal pattern with a slight modulation by the semidiurnal tide whereas for solstice the winds have a pronounced semidiurnal pattern. 50 ' ' I ' ' I ' ' I ' ' I ' ' I ' ' I ' ' I ' ' ' ' I ' ' I ' ' I ' ' I ' ' I ' ' I ß ' I ' ' , ,, 3 I,, 6 I,, 9 I,, 12 I,, 15 I,, 18 I I I ; 1 I 0 UT (HOURS) (MTIMES ,00 TO ,00),,, I,,, I,,, I,,, I,,, I,, i I 20. O0 O. O0 4. O0 8. O0 12. O0 16. O0 Figure 3b. TIME-GCM prediction of Bear Lake vector winds for equinox ,, I,, I,, I,, I,, I,, I,, I,, UT (HOURS) (MTIMES ,00 TO ,00) I,,,I,,,I,,,I,,,I,,,I,,,I LOCAL TIME (HRS) (LON= 0.00) Figure 3d. TIME-GCM prediction of Bear Lake vector winds for solstice.

5 SHEPHERD ET AL.' TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW 14, WINDII/TIME-GCM Emission Rate Comparisons at Equinox In the preceding comparisons we have shown excellent agreement between the TIME-GCM model predictions and lo2 averaged winds observed by WINDII. The WINDII averaging process involves two distinct steps; the first is that for each day the winds are zonally averaged over that day, and the second that a full month of data is binned in order to cover the full range of to-a 9o local times. This means that different local time samples come from different days of the month, allowing the possibility of aliasing between monthly and diurnal variations. The good agreement between WINDII data and the TIME-GCM indicates that with this type of averaging, the tidal behavior over I month 80, I,. I,, I, i is sufficiently well behaved to agree with the model predictions. For the ground-based results obtained on a single day there is UT (HOURS) (MTIME$ ,00 TO ,00) agreement in the pattern of the variation, but the magnitudes do LOCAL TIME (HRSI (LON= not agree well, indicating the difference between zonally averaged data combined over 1 month, and that obtained for a Figure 4b. Volume emission rate of the atomic oxygen green single longitude and a single day. Overall we may conclude that line emission (photon cm '3 s -1 as predicted by the TIME-GCM as the observed and predicte dynamics are compatible. We now proceed to compare other variables, which from an a function of altitude and local time for a latitude of 42øN at equinox. observational point of view are entirely independent of the winds. We begin by examining the behavior of the emission rate over Bear Lake as seen by WINDII for March/April The to-morning variation is smaller than that observed. HRDI procedure for the analysis of emission rate data has been measurements for a fall equinox period of 3 days duration, of the described by Shepherd et al. [1995]. As for the wind measurements, the emission rate data have been processed with emission rates for O(1S), 02 atmospheric and OH were V4.98 of the production processing software. In Figure 4a we present contours of volume emission rate from the atomic oxygen with TIME-GCM model predictions by Yee et al. [1997], who also found good agreement. 558 nm line over an altitude range from 80 to 110 km, as a function of local time from 1900 to 0500 hours. At the beginning of the evening the emission rate is very low, 24 photon cm -3 s -1 at 1900 LT (the numbers on the contours are in these units), but the emission gradually rises throughouthe night, reaching a maximum at 3.5 LT, of 130 photon cm '3 s -1. Using the TIME-GCM the O(1 S) volumemission rates were calculated using the formulation of Murtagh et al. [1990], and these are shown in Figure 4b. The predicted volume emission rate pattern agrees well with the WINDII data of Figure 4a, and the morning value of 130 photon cm '3 s 'l agrees very well with the TIME-GCM value of 160, a remarkable agreement for an absolute measurement compared with an absolute calculation. However, the evening value for WINDII is much smaller than for the TIME-GCM, so that the magnitude of the predicted evening- 11o loo 95 9O LOCAL SOLAR TIME (hr) Figure 4a. Volume emission rate of the atomic oxygen green line emission (photon cm -3 s 'l) as observed by WINDII as a function the same factor. Since a factor of 3 for the 0 2 atmospheric band of altitude and local time for March/April 1993 at 40øN. corresponds to a factor of 5 for O(1S) when the atomic oxygen 11o lo8 lo6 lo4 10-* compared 5. Emission Rate and Temperature Comparisons at Equinox We now compare these spacecraft observations with observations made at Bear Lake and previously reported upon by Wiens et al. [1995]. The most precise emission rate results obtained for the nocturnal oxygen airglow were acquired from the 02 atmospheric band, by MORTI, the mesopause oxygen rotational temperature imager [Wiens et al., 1991 ]. Since both the O(1S) 558 nm and and 02 atmospheric band emissions originate from the recombination of atomic oxygen, the two emissions are related, but the chemistry needs to be taken into account. The first s tep in the production of the atomic oxygen green line and the atmosph,eric band is common, namely O+O+M O 2+M, and for O(1S) the subsequent step involves 02* + O, while for the 02 atmospheric band it involves 02* Approximately then, the O(1S) emission rate depends on the cube of the atomic oxygen density while for 0 2 atmospheric it depends on the square. The MORTI results for the 02 atmospheric (0,1) band emission rate as obtained by Wiens et al. [1995] are shown in Figure 5a, for data averaged from the nights of March 19-24, There is a cleor minimum at 2000 LT (shown as -4), at about 140 R. From this minimum the emission rate rises throughouthe night to a value of 490 R. This is similar to the gradual increase in the WINDII emission rate data of Figure 4a, except that the fractional increase for MORTI is somewhat less, a factor of 3.5, compared to a factor of 5 for WINDII: the MORTI values are integrated emission rates, while the WINDII values are volume emission rates, but when the WINDII values were vertically integrated, the increase was by

6 ,746 SHEPHERD ET AL.: TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW dependence is taken into account, the observed value of 3 represents excellent agreement between WINDII and MORTI. The temperature derived from the rotational temperature of the 0 2 atmospheric band is shown in Figure 5b. It follows the pattern of the emission rate very closely, with a peak-to-valley difference of 36 K, the change occurring in a time of about 8 hours. The corresponding TIME-GCM predictions are shown in Figure 6. As for the green line emission, the 0 2 atmospheric (0,0) volume emission rates were calculated using the formulation of Murtagh et al. [1990]; they are then integrated vertically to correspond to what would be seen by a vertically viewing ground-based instrument. Because the (0,0) band is absorbed in the lower atmosphere, the observations are for the (0,1) band so the absolute emission rates of the two cannot be equated. However, the emission rates are proportional to one another, so that the variations may be compared. These predicted integrated emission rates are shown in Figure 6a where we see a low value of emission rate at 2000 LT, as is true for the WINDII 558 nm volume emission rate at 95 km as shown in Figure 4a, and for the MORTI 0 2 atmospheric (0,1) emission rate as shown in Figure 5a. For both the TIME-GCM prediction and the MORTI observations the emission rate monotonically increases throughout the night, but the fractional increase is more for MORTI than for the TIME-GCM. Both the observations and prediction display a shoulder, in the form of a secondary maximum at near 0000 LT. As is discussed in more detail later, this shoulder is a manifestation of the semidiurnal tide, showing that it is weakly present at midlatitudes equinox. The TIME-GCM prediction of the predicted line-of-sight Doppler temperature obtained from a ground-based instrument observing the 0 2 Atmosphere band is shown as a function of local time in Figure 6b. As for the predicted wind, elementary o o 5OO loo LST (( ours) (o) z o ' I ' ' I ' ' I ' ' I ' I,, I,, I,, I, 3 6 g 12 UT (HOURS) (MTIMES oog 01,00 TO oog 14,00),, I,,, I,,, I,, ' I ' ' I ' ' I ' ' I ' UT (HOURS) (MTIMES ,00 TO ,00) m, I,,, I, m, I, Figure 6. TIME-GCM prediction of MORTI ground-based measurements for the 0 2 atmospheric (0,0) band as a function of local time for Bear Lake: (a) EO200 stands for the integrated emission rate of the 0 2 (0,0) band, and (b) DOP-TN for the emission and temperature weighted Doppler neutral temperature for the 0 2 atmospheric band o ß..... ß ø UT (hours) Figure 5. Results obtained by the MORTI instrument at Bear Lake from the (0,1) 0 2 atmospheric band for the nights of March 19-24, 1993, as a function of local time for (a) the vertically integrated emission rate and (b) the vertically averaged rotational temperature. The results are averaged over the nights indicated; the solid circles are individual data points, and the solid curve is a smoothed fit to these points. 14 Doppler line profiles for each airglow layer are added together, each with the temperature, wind, and emission rate appropriate to that layer. This is directly comparable to the ground-based MORTI results presented in Figure 5b. The patterns are very similar. The 0000 LT shoulder seen for the emission rate is more prominent for the temperature; this is true for both the TIMEo GCM and MORTI. The peak temperature is similar for both model and MORTI at around 195 K, but the minima are significantly different. The temperature change predicted by the model is about 17 K, peak to peak, compared with 35 K for MORTI. Thus the model predictions for temperature are not in as good agreement (in amplitude) with MORTI as are the winds and emission rates observed by both WINDII and MORTI. It should be noted that the model calduiates Doppler temperature from atomic oxygen emi'ssion, while MORTI measures the rotational temperature from the 0 2 atmospheric band. According to present

7 SHEPHERD ET AL.: TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW 14,747 understanding these should both correspond to kinetic temperatures and therefore agree, but to our knowledge this belief has never been tested. 11o lo4 6. Midlatitude Emission Rate Comparisons at Solstice We can examine the balance between diurnal and semidiurnal tides at midlatitudes by moving from equinox to solstice. The WINDII volume emission rates for solstice as a function of local time are shown in Figure 7a, for December 1992/January The emission rate pattern has drastically changed from that at equinox, with a maximum value of 140 photon cm '3 s '1 at 0.5 LT, and minima of about 90 at 1900 and 0500 LT. The TIME- GCM prediction of OI 558 nm emission rate shown in Figure 7b is closely consistent with this pattern, with a minimum value of 70 photon cm '3 s '1 and a maximum volumemission rate of 90 photon cm '3 s '1 both of which compare well with the observed values. Unlike the equinox, the predicted variation of a factor of 1.3 is not much different from the observed value of 1.5. Burrage et al. [1994] also studied the local time variation of oxygen airglow emission rate, using as a parameter the integrated emission rate for a constantangent point altitude of 94 km. The results, shown in their Figure 4, reveal less difference in the local time emission rate pattern between equinox and solstice than was found here, but an effect was evident. The inconsistency may arise from interannual variability, as the time periods selected are not exactly the same. The MORTI data for solstice are shown in Figure 8a as taken from Wiens et al. [ 1995]; they display the same deep minimum at 21.5 LT as observed by WINDII and a broad double peak extending from 0200 to 0500 LT. The temperature pattern shown in Figure 8b is coherent with this, but shifted earlier in phase by about 0.5 hours. The corresponding TIME-GCM prediction of 0 2 atmospheric band emission is shown in Figure 9a, and the temperature in Figure 9b. These show the deep minimum at 21.5 LT and a broad (but not double) peak at 0300 LT. The overall pattern resembles an 8 hour wave- the MORTI data were fitted by Wiens et al. [ 1995] as the superposition of a semidiurnal tide and a 6 hour wave. The predicted peak to peak temperature perturbation is 9 K, compared with 50 K observed for MORTI. The predicted emission rate perturbation is about 30% compared to the observed value of 80%. ' r UT (HOURS) (HTIHœ$ ,00 TO ,00! LOCAL TIHE (HRSI (LON= ) Figure 7b. The corresponding TIME-GCM predictions. The excellent agreement between WINDII and the TIME- GCM, compared with the ground-based observations which show good agreement in pattern, but differences of a factor of 2 or more in magnitude, elucidate the differences between tidal perturbations observed at one location and their zonal average. The averaging process employed with the WINDII data effectively removes planetary scale features, and also averages out any longitudinal variations in tidal amplitude. Apparently, this averaging process conforms well to the atmosphere as predicted with the TIME-GCM model. Observations at one ground-based site can be vastly different from those at another; planetary scale features can cause emission rate variations by a factor of 4 over all longitudes observed ibr the same latitude [Shepherd et al., 1993b], which is much more than the discrepancy observed here. Zhang et al. [1998] studied zenithal emission rates observed with WINDII at low latitudes and found that longitudinal variations are somewhat larger than tidal variations, and that maximum emission rates occur when tidal maxima are superposed upon planetary or gravity waves of long horizontal wavelength. It seems likely that similar effects occur at midlatitudes. 7. Discussion 11o O LOCAL SOLAR TIME (hr) Figure 7a. Volumemission rate of the atomic oxygen green line emission as observed by WINDII as a function of altitude and local time for December 1992/1993 and January 1993/1994 at 40øN. The patterns of equinoctial emission rate variation as a function of local time seen at the equator by Shepherd et al [1995] and at midlatitudes as reported here show some interesting similarities, and some differences. Both show a diurnal variation in the magnitude of the volume emission rate at 95 km by about a factor of 2. The two patterns are similar with respecto local time, but the minimum appears at 2000 LT for the midlatitude data and at 2400 LT for the equatorial data; that is, at midlatitudes the tidal phase is shifted forward by 4 hours. The equatorial data show a dramatic change in altitude of the emission layer, by about 5 km. The mid-latitude data show a smaller change, perhaps one-half as much. These observational results are consistent with those shown by Burrage et al [1994] for the 02 atmospheric band emission. The TIME-GCM predictions for the O( S) 558 nm emission rate at equinox agree extremely well with that observed by

8 14,748 SHEPHERD ET AL.: TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW OO 250 ß E LST ( ours) " UT (hours) 14 (o) (b) Figure 10a shows the TIME-GCM prediction for 96 km at equinox, which is clearly diurnal at all latitudes, and has a phase which is independent of latitude. The maximum amplitude is near 20 ø latitude, and this weakens towards midlatitudes. The corresponding WINDII data for 96 km are shown in Figure 1 la and the pattern is very similar except that the weakening towards midlatitudes progresses at a lesser rate. The TIME-GCM prediction for solstice is shown in Figure 10b; the pattern is drastically different from that at equinox. The contours are organized in a diagonal way, indicating a systematic shift in phase with latitude, and are clearly semidiurnal. The corresponding pattern for WINDII as shown in Figure 1 lb is remarkably similar, showing most of the same features. However the overall pattern is more complex, showing a clearly semidiurnal tide at midlatitudes in the northern hemisphere, a more diurnal tide at tropical latitudes, and something less clear in the southern hemisphere, in part owing to missing data Figm'e 8. Results obtained by the MORTI instrument at Bear Lake from the 0 2 atmospheric band for the nights of Jan , 1993, as a function of local time for (a) the vertically integrated emission rate and (b) the vertically averaged rotational lemperature. The solid circles are averages over the nights indicated, and the solid curve is a smoothed fit to the points WINDII, both at the equator as shown by Roble and Shepherd [1997] and at midlatitudes as presented here. The good agreement between the model results and the observations indicates that the emission rate changes are the result of largescale mixing driven by the tides, a conclusion that is in agreement with the analysis of Ward [ 1998]. The equatorial tide appears to involve larger vertical motions than occur at midlatitude, for both the WINDII data and the model. The model agreement with the ground-based MORTI observations of O2(blZ) emission rate and temperature is also extremely good for the pattern of local time variation, but the observed amplitudes are greater by a factor of 2 for emission rate, and a factor of 3 for temperature. The good agreement between the WINDII-observed emission rates and those predicted by the TIME-GCM, when the latter is adjusted to agree with the WINDII wind fields, confirms the validity of the WINDII (and HRDI) wind fields. Put differently, such large wind amplitudes are required to produce the dramatic emission rate variations that are observed. At solstice, the diurnal emission rate pattern changes to one that is primarily semidiurnal, for WINDII, for MORTI and for the TIME-GCM. This is in agreement with the conclusions of Wiens et al. [ 1995] drawn from the ground-basedata only. The overall conclusion supports that deduced in the equatorial study by Roble and Shepherd [ 1997] in that at solstice the diurnal tide does not reach to sufficient altitude to perturb the O2(blZ) and O(1S) airglow emissions. Burraget al. [1995] studied the seasonal variation of the semidiurnal tides. They found the semidiurnal tide to be much less evident in the March/April period compared to summer and winter solstice, and even when compared with the September equinox. This is consistent with what we have found here. This conclusion is illustrated in Figures 10 and 11 where the behavior of the meridional wind component of the tides as seen in a latitude versus local time plot for a fixed altitude is shown UT (HOURS) (MTIMES ,00 TO ,00) I,,, I,,, I,, UT (HOURS) (MTIMES ,00 TO ,00),, I,,, I,, I, 20. O0 O. O0 4. O0 Figure 9. TIME-GCM pre, dictions of the 02 atmospheric (0,0) band for Bear Lake at solstice for (a) EO200, the 0 2 (0,0) vertically integrated emission rate and (b) DOP-TN, the vertically averaged neutral Doppler temperature, weighted by emission rate and temperature.

9 SHEPHERD ET AL.: TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW 14,749 (a) represents an average over the interannual variability of several years. The observations of WINDII and MORTI agree best, as expected, when taken for the same year. 4O 2O -4O -6O -8O UT (HOURS) (MTIMES ,00 TO ,00) I,,, I,,, I I, m, I,,, I,,, I O. O0 4. O0 8. O0 12. O0 1 O. O0 20. O0 O. O0 LOCAL TIHE (HRSI (LON= (b) 80 k ' "x' ',. x ' [._!0_._,..Xx, '. "o.' 8. Conclusions 1. The diurnal tide is manifested in the meridional wind as a cell with its maximum amplitude near 20 ø latitude and its boundaries at the equator and near 42 ø, at midlatitudes. As a consequence, there is a large diurnal variation in emission rate at the equator as reported by Shepherd et al. [1995)] where the opposing winds across the equator produce very large effects. At equinox, WINDII also observes large diurnal tidal perturbations at Bear Lake, 42øN, which is located at the poleward boundary of the meridional wind cell. The primary difference between the two is that the minimum occurs 4 hours earlier in local time at midlatitudes than at the equator. The altitude perturbation of the emission layer at midlatitudes is less than at the equator for equinox, about 2.5 km instead of 5 km. 2. At solstice, themission rate pattern changes drastically for '-, o eo,.-a,, o,. WINDII, to a clearly se diumal pattern. ß., ß ß ß ß ß ß, ß.. ', ; '.-' i --10, C T-,-,? I,,, I,,, I I I,, I,,, I O O0 8. O0 12. O0 16. O0 20. O0 O. O0 Fibre 10. ME-GCM prediction of mefidional wind at 96 km titude as a function of latitude d longitude for (a) equinox LOCAL SOLAR TIHE {H) d (b) solstice. Although the seasonal variability of the diurnal tide is now b 6o well established through observation, its cause remains unknown. The observed seasonal behavior has been reproduced with the 40 Canadian Middle Atmosphere Model (CMAM) by McLandress [1997], who showed that it could not be accounted for by to 2o seasonal variability in the associated thermal heating, but could "' not specifically identify the mechanism. to 0 Because ordy a limited amount of data are shown here, it is important to point outhat the WINDII data for the vernal -2O equinox of 1993, do not represent the equinoxes of all years. When the WINDII data were averaged for the vernal equinoxes -40 of 1992 and 1993, much better agreement with the TIME-GCM model was obtained than is shown here, but the agreement with MORTI (for 1993 only) was not as good. The WINDII data shown here for 1993 only do not agreed as well with the TIME- GCM but agree much better with MORTI. This is in part because 1993 seems to have some unique dynamical characteristics, the discussion of which is beyond the scope of this paper, but more generally that the TIME-GCM model likely ' LOCAL SOLAR TIHE (H) Figure 11. WINDII observations of meridional wind at 96 km es a function of latitude and longitude for (a) March/April 1992/ 1993 and (b) December 1992/1993 and January 1993/94

10 14,750 SHEPHERD ET AL.: TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW 3. The TIME-GCM predictions agree extremely well with the WINDII observations of wind and emission rate at midlatitudes. This agreement holds for both equinox and solstice; the model results show that at solstice the diurnal tide does not reach sufficient altitude to perturb the airglow layer, so that the variation pattern arises almost solely from the in situ semidiurnal tide. 4. The TIME-GCM predictions agree reasonably well in pattern with the MORTI observations of O2(blZ) emission rate and rotational temperature, but the observed amplitudes are a factor of 2 larger for the emission rate, and a factor of 3 larger for the temperature. 5. The good agreement between the satellite and model observationshow that the zonal and temporal averaging of the satellite data conform well to the model formulation. The Acknowledgments, The WINDII project is jointly supported by the Canadian Space Agency and the Centre National d'etudes Spatiales, France. The authors are grateful for their support, along with NASA, in making this work possible. Support for scientific analysis of the data is provided by the Natural Sciences Engineering and Research Council of Canada. The Centre for Research in Earth and Space Technology (CRESTech) is a designated Centre of Excellence supported by the Technology Fund of the Province of Ontario. We thank the referees for their helpful suggestions. The Editor thanks Richard Walterscheid evaluating this paper. for his assistance in Garcia, R.R., and S. Solomon, The effect of breaking gravity waves on the dynamics and chemical composition of the mesosphere and lower thermosphere, J. Geophys. Res., 90, , Gault, W.A., et al., Validation of O(Is) wind measurements by WINDII: The WIND imaging interferometer on UARS, J. Geophys. Res., 101, , 1996a. Gault, W.A., S. Brown, A. Moise, D. Liang, G. Sellar, G.G. Shepherd, and J. Wimperis, Erwin: An E-region wind interferometer, Appl. Opt., 35, , 1996b. Hays, P.B., V.J. Abreu, M.E. Dobbs, D.A. Gell, H.J. Grassl, and W.R. Skinner, The high resolution Doppler imager on the Upper Atmosphere Research Satellite, J. Geophys. Res., 98, , Roble, R.G., and E.C. Ridley, A thermosphere-ionosphere-mesosphere References electrodynamics general circulation model (time-gcm), Geophys. Res. Lett., 21, , Akmaev, R.A., and G.M. Shved, Modeling of the composition of the lower thermosphere taking account of the dynamics with applications Roble, R.G. and G.G. Shepherd, An analysis of wind imaging to tidal variations of the [Oil 5577 airglow, J. Atrnos. Terr. Phys., 42, interferometer observations of O(1S) equatorial emission rates using , the thermosphere-ionosphere-mesosphere-electrodynamics general Brenton, J.G., and S.M. Silverman, A study of the diurnal variations of circulation model, J. Geophys. Res., 102, , the 5577 A [OI] airglow emission at selected IGY stations, Planet. Shepherd, G.G., W.A. Gault, D.W. Miller, Z. Pasturczyk, S.F. Johnston, Space Sci., 8, , P.R. Kosteniuk, J.W. Haslett, J.W. Kendall, and J.R. Wimperis, Burrage, M.D., N. Arvin, W.R. Skinner, and P.B. Hays, Observations of WAMDII: Wide-angle Michelson Doppler imaging interferometer for the 0 2 atmospheric band nightglow by the high resolution Doppler Spacelab, Appl. Opt., 24, , 1985 Shepherd, G.G. et al., WINDII: The Wind imaging interferometer on the imager, J. Geophys. Res., 99, , Upper Atmosphere Research Satellite, J. Geophys. Res., 98, Burrage, M.D., D.L. Wu, W.R. Skinner, D.A. Ortland, and P.B. Hays, 10750, 1993a. Latitude and seasonal dependence of the semidiurnal tide observed by the high resolution Doppler imager, J. Geophys. Res., 100, Shepherd, G.G. et al., Longitudinal structure in atomic oxygen 11321, concentrations observed with WINDII on UARS, Geophys. Res. Lett., 20, , 1993b. Cogger, L.L, R.D. Elphinstone, and J.S. Murphree, Temporal and latitudinal 5577 A airglow variations, Can. J. Phys., 59, , Shepherd, G.G., C. McLandress, and B.H. Solheim, Tidal influence on O( S) airglow emission rate distributions at the geographic equator Donahue, T.M., B. Guenther, and R.J. Thomas, Distribution of atomic observed by WINDII, Geophys. Res. Lett., 22, , oxygen in the upper atmosphere deduced from Ogo 6 airglow Ward, W.E. Tidal mechanisms of dynamical influence on oxygen observations, J. Geophys. Res., 78, , recombination airglow in the mesosphere and lower thermosphere, Forbes, J. M., Atmospheric tides, 1, Model description and results for the Adv. Space Res., 21, No. 6, solar diurnal component, J. Geophys. Res., 87, , 1982a. Ward, W.E., Y.J. Rochon, C. McLandress, D.Y. Wang, J.R. Criswick, Forbes, J.M., Atmospheric tides, 2, The solar and lunar semi-diurnal B.H. Solheim, and G.G. Shepherd, Correlations between the components, J. Geophys. Res., 87, , 1982b. mesospheric O(IS) emission peak intensity and height, and Forsyth, R.J., and P.C. Wraight, A survey of research on nightglow temperature 98 km using WINDII data., Adv. Space Res., 14, 57-60, variability, Planet. Space Sci., 35, , McLandress, C., Seasonal variability of the diurnal tide: Results from the Canadian middle atmosphere general circulation model, J. Geophys. Res., 102, , McLandress, C., Y. Rochon, G.G. Shepherd, B.H. Solheim, G. Thuillier, and F. Vial, The meridional wind component of the thermospheric agreement of both with the diurnal pattern of the ground-based measurements, but with large differences in amplitude, give a tide observed by WINDII on UARS, Geophys. Res. Lett., 21, good indication of the zonal variability that has been averaged 2420, out in the satellite data, and which does not appear in the TIME- McLandress, C., G.G. Shepherd, and B.H. Solheim, Satellite observations GCM predictions. This serves as a good reminder of the of thermospheric fides: Results from the Wind imaging interferometer complexity of the real atmosphere. on UARS, J. Geophys. Res., 101, , 1996a. McLandress, C., G.G. Shepherd, B.H. Solheim, M.D. Burrage, P.B. Hays, and W.R. Skinner, Combined mesosphere/thermosphere winds using WINDII and HRDI data from the Upper Atmosphere Research Satellite, J. Geophys. Res., 101, , 1996b. Murtagh, D.P., G. Witt, J. Stegman, I.C. McDade, E.J. Llewellyn, F. Harris, and R.G.H. Greer, An assessment of proposed O(IS) and O2(bl ) nightglow excitation parameters, Planet. Space Sci., 38, 43-53, Petitdidier, M., and H. Teitelbaum, Lower thermospheremissions and tides, Planet Space Sci., 25, , Petitdidier, M., and H. Teitelbaum, O( S) excitation mechanism and atmospheric tides, Planet. Space Sci., 27, , Rayleigh, Lord, On a night sky of exceptional brightness, and on the distinction between the polar aurora and night sky, Proc. Roy. Soc. London Ser. A, 131, , 1931.

11 SHEPHERD ET AL.: TIDAL INFLUENCE ON MIDLATITUDE AIRGLOW 14,751 Wiens, R.H., S.P. Zhang, R.N. Peterson, and G.G. Shepherd, MORTI: A mesopause oxygen rotational temperature imager, Planet. Space Sci., 39, , Wiens, R.H., S.P. Zhang, R.N. Peterson, and G.G. Shepherd, Tides in emission rate and temperature from the 0 2 nightglow over Bear Lake Observatory, Geophys. Res. Lett., 22, , Yee, J.-H., R.G. Robie, W.R. Skinner, M.D. Burrage, and P.B. Hays, Global simulations and observations of O(IS), O2(Iy,) and OH mesospheric nightglow emissions, 0r. Geophys. Res., 102, , Zhang, S.P., R.H. Wiens, B.H. Solheim and G.G. Shepherd, Nightglow zenith emission rate variations in O(Is) at low latitudes from WINDII observations, 0r. Geophys. Res., 103, , C. McLandress and S.-P. Zhang, Centre for Research in Earth and Space Technology, 4850 Keele Street, Toronto, Ontario, Canada M3J 3KI. ( mcland@windii.yorku.ca: sheng@windii.yorku.ca) R. G. Robie, High Altitude Observatory, National Center for Atmospheric Research, Boulder, Colorado, ( roble@hao.ucar.edu) G. G. Shepherd and R. H. Wiens, Centre for Research in Earth and Space Science, York University, 4700 Keele Street, Toronto, Ontario, Canada M3J 1 P3 windii.yorku.ca (Received November 20, 1997; revised February 23, 1998; accepted March 5, 1998.)