Geophysical Research Letters

Thermospheric observations of equatorial wavenumber 4 density perturbations from WINDII data



[1] Previous studies have established that the diurnal eastward-propagating non-migrating tide of wavenumber 3 (DE3) is produced by latent heat release in deep tropospheric convection and therefore is longitude dependent. Longitudinal variations of F-region ionospheric electron density show a zonal wavenumber 4 pattern, expected for satellite observations of DE3 for a fixed local time with respect to a rotating Earth. Satellite observations of winds show wavenumber 4 reaching 180 km, indicating neutral coupling of the troposphere with the thermosphere. In the present work, WINDII measurements of O(1S) green line emission at 250 km excited by photoelectron impact on atomic oxygen are interpreted as neutral density observations; showing wavenumber 4 to be a common density perturbation, with a fractional amplitude variation of about 10%. This result confirms direct neutral coupling of troposphere and thermosphere but also opens the way to other studies of neutral density variations at 250 km.

1. Introduction and Background

[2] A topic of great current interest in atmospheric dynamics is the developing recognition of the influence of the diurnal eastward-propagating non-migrating tide of wavenumber 3 (DE3) on the thermosphere. Since DE3 is produced in the troposphere by latent heat release in deep tropospheric convection [Hagan and Forbes, 2002] thermospheric observations can potentially provide undeniable evidence of coupling between the troposphere and thermosphere. Identification of the thermospheric imprint began with analysis of the night airglow O+ 135.6 nm emission from IMAGE/FUV by Sagawa et al. [2005]. This emission is produced by the radiative recombination of O+ with electrons, and so is closely proportional to the square of the electron density. They identified an equatorial perturbation of wavenumber 4, which is how DE3 is seen from a spacecraft at fixed local time with respect to a rotating Earth. These observations were extended by Immel et al. [2006] and England et al. [2006] in proposing that the DE3 interaction with the E-region atmosphere produced a modulation of the dayside equatorial fountain, an ionospheric effect. Forbes et al. [2009] used density perturbations from the CHAMP and GRACE satellites (which occur over a range of altitudes) to infer exospheric temperature variations of diurnal and semidiurnal tides that were concluded to be excited in the troposphere, strongly influenced by the global land-sea distribution. This was evidence that the neutral thermosphere was being influenced directly. Talaat and Lieberman [2010] used the full suite of WINDII data to show that a prominent zonal wavenumber 4 appeared in daytime zonal winds, extending up to 180 km, a further strong indication of direct coupling between the troposphere and thermosphere.

[3] This suggested to the present author that thermospheric density perturbations might be discernable in the WINDII airglow emission rate data. The O(1S) atomic oxygen green line at 557.7 nm is observable during both day and night; at night it is confined to a thin strong layer near 100 km arising from the recombination of atomic oxygen and a weak F-region layer that occurs from the recombination of O2+ ions. The O(1S) dayglow emission observed routinely by WINDII extends throughout much of the thermosphere, from about 90 km to nearly 300. The 14 vertical profiles of volume emission rate for a single day, January 27, 1996, in the latitude range −0.2° to −2° are shown in Figure 1. The interpretation of this emission in terms of atmospheric constituents is complex, but is well modeled and well understood [Singh et al., 1996]. Near 100 km the nighttime peak is enhanced by hνLyman-β + O2, raising the altitude of the emission rate peak a little above 100 km [Maharaj-Sharma and Shepherd, 2004]. There is a second peak in the middle thermosphere near 130 km that is dominantly due to N2(A3Σu+) + O with some contribution from e + O. Here the N2(A3Σu+) is excited by photoelectron impact and the “e” in e + O are also photoelectrons, arising from photoionization of various species by solar radiation. Untangling these contributions to the volume emission rate profile in terms of density perturbations is a formidable challenge so the approach chosen here, which is conveniently consistent with the goal of higher-altitude thermospheric investigation is to focus on the highest possible altitude of measurement, near 300 km. At this altitude the molecular reactions make negligible contributions and in fact the dominant contribution is from e + O → e + O(1S). The observed emission rate is thus a measure of the atomic oxygen density at this altitude. Observations of wavenumber 4 in the neutral density would further reinforce the interpretation of direct coupling through DE3 between the troposphere and thermosphere.

Figure 1.

WINDII daytime volume emission rate profiles for January 27, 1996 for an equatorial latitude band between −0.2° and −2.0°. There are fourteen profiles each corresponding to one UARS orbit.

[4] Figure 1 indicates significant variations with longitude, as the fourteen profiles cover the full range of longitudes for a single day. Since the local time and latitude are fixed for Figure 1 the solar contributions to the excitation are the same for all profiles. Zhang and Shepherd [2005] showed that the characteristic features of the two striking peaks in Figure 1, the peak emission rates and the altitude of the peak, could be empirically fit with simple equations involving only two variables; the solar flux and the solar zenith angle. However, this does not take account of the marked variations in peak volume emission rate for both the 100 km and 130 km peaks which must be the result of dynamical perturbations. These involve dynamical components additional to the one sought here and will not be considered further.

[5] Near 200 km there is a sharp spike in the profile, which arises from the WINDII baffle. This meter-long baffle was extremely successful in shielding the instrument from the bright cloud tops only 1.5° in elevation below the bottom of the altitude range of observation, making possible these daytime observations, along with the daytime winds that have been so extensively investigated [Zhang et al., 2007]. However there is one design flaw, probably a reflection, that produces this sharp peak and the smaller one not far above it, near 220 km. These regions must be avoided and the top-most altitudes are also not utilized because they involve the starting point of the inversion. Consequently the altitude of 250 km, very near the minimum in the emission rate profile, has been chosen for this study.

[6] It is shown that zonal wavenumber 4 is not only evident, but is dominant at this altitude, with a secondary contribution from wavenumber 1. It can be said to be ubiquitous, albeit with day-to-day variations. At this altitude it occurs in a relatively narrow region around the equator, which is consistent with other observations [England et al., 2006; Talaat and Lieberman, 2010]. These well-defined characteristics appear to exist only between the altitudes of 245 and 255 km; outside this range other wave components become dominant. This investigation provides one more way of studying the DE3 non-migrating tide, additionally to those that have already been made and further study of this emission promises to shed further light on this fascinating troposphere-thermosphere interaction.

2. WINDII Data and Analysis

[7] The WINDII (WIND Imaging Interferometer) instrument was launched on NASA's Upper Atmosphere Research Satellite (UARS) on September 12, 1991 (twenty years ago!) and provided measurements of winds from airglow emissions along with the volume emission rates themselves as described by Shepherd et al. [1993a]. The selected WINDII emissions included the O(1S) emission already discussed, along with lines in the hydroxyl (8,3) band and O2 Atmospheric (0,0) bands. However, the CCD imaging detector was large enough to allow the field-of-view to extend to 300 km and for that reason, a filter was included for observation of the O(1D) atomic oxygen red line at 630.0 nm. To conform to the UARS primary objectives it was normally employed only one day per week. It was also true that to obtain valid winds and emission rates at 100 km that the O(1S) emission be inverted from the top of the emission downwards, requiring a maximum altitude of about 300 km. The same arguments outlined for the choice of 250 km as a study altitude apply to the O(1D) airglow and results for one day of data from that emission are presented here.

3. Results

[8] The volume emission rate of the O(1S) dayglow at 250 km is shown in Figure 2 for September 28, 1993, day number 271 of 1993, UARS day 748, a near-equinox day. The latitude range allowed was from −20 to 20 degrees, which captured 155 profiles; these have been smoothed by a rectangular window 9 data points in width and interpolated to intervals of 8° longitude for fitting and plotting. A zonal wavenumber 4 perturbation is clearly dominant with the amplitude of successive peaks varying only slightly from one to the next. The solid line fit includes a constant term yielding 12.4 photons cm−3 s−1, a small linear term, and terms for wavenumbers 1 and 4. The amplitude of wavenumber 4 is small in an absolute sense, 0.9 photons cm−3 s−1 but this is a very significant 7.2% of the mean value. The amplitude of wavenumber 1 is 0.7 photons cm−3 s−1, only a little less than that of wavenumber 4. The standard deviation of the WINDII emission rates for these data is about 3 photons cm−3 s−1 for individual measurements, larger than these values, but the consistency in pattern over 155 individual data points (for this day) makes the amplitudes well defined.

Figure 2.

WINDII O(1S) volume emission rates (squares) as a function of longitude for day number 271, September 28, 1993 for the latitude band from 20° S to 20° N at an altitude of 250 km. The smoothed and interpolated data (see text) have been fitted with a constant value, a linear value and wavenumbers 4 and 1 (solid line).

[9] Figure 3 shows the pattern for a winter day, January 27, 1996, UARS day 1599, the same data as shown in Figure 1 except that here they are plotted versus longitude for a fixed altitude of 250 km. Wavenumber 4 is again clearly evident and is well fitted with a mean value of 12.3 photons cm−3 s−1 and an amplitude that is 1.8 photons cm−3 s−1, 14% of the mean value. The wavenumber 1 component has a “huge” amplitude of 3.1 photons cm−3 s−1, 25% of the mean value. The wavenumber 1 feature is not well understood even though it is seen in all WINDII O(1S) nighttime data [Shepherd et al., 1993b]. Altogether it is remarkable that such clear signals are present in this weak airglow emission at 250 km altitude.

Figure 3.

WINDII volume emission rates as in Figure 2, but for January 27, 1996.

[10] For a preliminary assessment of the day-to-day variability in wave amplitudes, days from December 31, 1995 to February 1, 1996 were examined, of which 19 days yielded amplitude fits. In viewing these 33 days wavenumber 4 is an essentially ubiquitous feature. A wavenumber 4 pattern is normally evident, but sometimes the amplitude of one peak is so much larger than the others that a satisfactory fit is not possible; an indication of variability during the course of a day, but also the possible identification of an intense localized convection event. That aspect cannot be considered here and these days are simply not included. For the remaining days shown adequate fits were obtained as shown in Figure 4 and these still show a high degree of variability over one month. There is an event of enhanced wavenumber 4 amplitude centred on January 5, 1996 that lasts for six days; the local time is indicated and this occurred during the morning hours. Higher values are also seen at the end of January, with superimposed variations – this likely reflects the fact that these values occur around noon, with smaller solar zenith angles.

Figure 4.

Wavenumber 4 amplitudes for January, 1996 (solid line, plus symbols), along with the corresponding local times in hours (squares).

[11] The WINDII O(1D) measurements have already been mentioned. At 250 km the dominant excitation process for this emission in the dayglow is again photoelectron impact on O. It is therefore to be expected that the longitudinal variations of O(1S) and O(1D) are similar. The longitudinal variation for September 15, 1993, UARS day 753 is shown in Figure 5; this is just a few days before the O(1S) measurements shown in Figure 2. The emission rates are larger by a factor of about 15, corresponding to the differences in excitation cross-sections, and a clear wavenumber 4 is present with an amplitude of 6.5 photons cm−3 s−1; this fractional variation of 3.7% lies between that shown in Figures 2 and 3 for O(1S). This confirms that the excitation processes for the O(1D) emission are very similar. Overall, the variation patterns for the two emissions are similar, as expected.

Figure 5.

WINDII O(1D) volume emission rates otherwise as in Figure 2, but for September 15, 1996.

4. Discussion and Conclusions

[12] The present results for selected days near winter solstice and the autumnal equinox offer convincing evidence for atmospheric density perturbations of wavenumber 4 with amplitudes that vary around 10% of the mean value. This extends the original observations of electron density perturbations in providing evidence of direct coupling between the troposphere and the neutral thermosphere. The result supports the finding by Talaat and Lieberman [2010] that wavenumber 4 is present in WINDII daytime zonal winds extending up to 180 km. One day of data for the O(1D) emission was shown, with results similar to those for O(1S), except for the factor of fifteen larger emission rates. WINDII maps of O(1D) emission near the equator were shown by Thuillier et al. [2002]; these were not inconsistent with a wavenumber 4 component, but were much clearer for OGO 4 results shown in the same paper.

[13] For this study the latitudinal range between 20° S and 20° N was employed. Reducing the range to say 0° to 20° changed the patterns very little so the results appear not to be very sensitive to latitudinal differences from the equator. However, the wavenumber 4 pattern drops rapidly outside the latitude bound of ±20°. The limited data shown here exhibit no difference between winter and equinox, unlike the results of England et al. [2006] which found the wavenumber 4 to favour equinoxes, and the postsunset ionosphere prior to midnight (the results shown here were for morning hours, or around noon). Differences between the neutral and ionized atmosphere are to be expected.

[14] The results shown here are preliminary, in that they are based on a limited volume of data, and a limited investigation. There is much more to be explored in further studies, including the latitudinal, seasonal, solar flux and local time variations as well as the potential presence of localized tropospheric events. The patterns at lower altitudes can also be investigated, although the interpretation is more complex because of the additional emission excitation processes.

[15] In conclusion, the non-migrating tide DE3 appears to couple to the thermosphere at 250 km, producing a roughly 10% variation in neutral density for wavenumber 4. Further, the O(1S) and O(1D) dayglow emissions in the vicinity of 250 km may prove to be a valuable way of studying density perturbations at these altitudes more generally. This can allow studies of perturbations at these altitudes from a variety of dynamical sources.


[16] The WINDII project was sponsored jointly by the Canadian Space Agency and the Centre National d'Etudes Spatiales of France, in collaboration with NASA. Science analysis is supported by the Natural Sciences and Engineering Research Council of Canada. The author is grateful to Richard Link, Brian Solheim and Marianna Shepherd for their helpful suggestions.

[17] The Editor thanks one anonymous reviewer for their assistance in evaluating this paper.