Inversion of O₂1.27 μm nightglow emissions: A climatological analysis using satellite Limb-Viewed spectra and Harmonic analysis method

1.   Introduction

As an important tracer for the atmospheric photochemical and dynamical processes in the middle and upper atmosphere, the airglow is widely used to study the mechanisms of these processes due to the joint influence of atmospheric oscillations and external solar radiation (Xu JY et al., 2012; Yang XJ et al., 2020; Wang WJ et al., 2021). Oxygen molecules have seven electronic energy levels, which means the transition between different energy levels can produce airglow in different bands. Among these bands, the 1.27 μm airglow is the product of spontaneous radiation of excited oxygen molecules returning to the fundamental state (${\text{O}}_{{2}}({\text{a}}^{{1}}{{\Delta }}_{\text{g}}\to{\text{X}}^{{3}}{\text{Σ}} _{\text{g}})$). The 1.27 μm airglow is one of the commonly used sources to investigate the physical characteristics of the middle and upper atmosphere.

In the 1970s and 1980s, rocket-based measurements were the main detection method of O₂1.27 μm nightglow. For example, the Energy Transfer in the Oxygen Nightglow (ETON) program explored the emission profile of O₂1.27 μm nightglow to analyze the photochemical and radiative mechanisms in the nightglow formation processes (McDade et al., 1987; Lopez-Moreno et al., 1988). The S-310-16 rocket project, combined with a ground-based grating monochrometer, initially clarified the transmission process of the O₂1.27 μm nightglow from the generation layer to the ground (Yamamoto et al., 1992). However, early rocket measurements had some limitations. For example, the instantaneity and discontinuity of the observations made them unsuitable for any global, long-duration, understanding of the distribution characteristics of nightglow emission. Satellite remote sensing technology can overcome these limitations; spaceborne instruments can provide data demonstrating the airglow’s vertical variation by limb-scan observation, greatly aiding study of the three-dimensional spatial structure and temporal variations of the middle and upper atmosphere (Wang YP et al., 2016; Li HC, 2023). The Solar Mesosphere Explorer (SME) satellite launched in 1984 was able to measure the oxygen airglow continuously for more than 12 hours per day. Based on the long-term observations of the SME mission, Howell et al. (1990) deduced the corresponding photochemical reaction mechanism. Gao H et al. (2011) revealed the global distribution characteristics of OH airglow and O₂ 1.27 μm nightglow from 2002 to 2010 based on airglow data detected by the SABER payload. Zarboo et al. (2018) retrieved the dayglow and nightglow VERs of the ${\text{O}}_{\text{2}}({\text{b}}^{\text{1}}{\text{Σ}}_{\text{g}}^{{+}})$ band and ${\text{O}}_{\text{2}}({\text{a}}^{\text{1}}{\Delta}_{\text{g}})$ band from airglow limb observations made at an altitude of 70−150 km by the SCIAMACHY payload, using the least square method. Li AQ et al. (2020) retrieved the O₂1.27 μm dayglow VERs based on observations from the Optical Spectrograph and Infrared Imaging System (OSIRIS), and discussed correlation of VERs with the number density of daytime ozone.

Research on the O₂1.27 μm airglow has gradually expanded to other fields. The wavelength band of this airglow is close to the the ${\text{CO}}_{\text{2}}$ absorption band; thus the uncertainty was small when it was used as an index of the atmospheric path length. The atmosphere’s greenhouse gas content can also be monitored by the radiative intensity of O₂1.27 μm airglow (Sun K et al., 2018; Bertaux et al., 2020). Since the O₂1.27 μm airglow has relatively strong emission and weak self-absorption in the Mesosphere and Lower Thermosphere (MLT) region, it is an ideal precursor for detecting atmospheric wind and temperature fields (He WW et al., 2019). For example, Sun K et al. (2022) used Bayesian inversion to retrieve temperature profiles by minimizing the difference between simulated and observed limb radiance spectra, using a priori regularization.

As mentioned above, systematic research on the long-term variation and spatial distribution of O₂1.27 μm nightglow has been relatively lacking. The O₂1.27 μm nightglow dataset from the SCIAMACHY payload has the advantages of long continuous observation time and high resolution at night, allowing detailed analysis of the periodic variation and spatiotemporal characteristics of nightglow. In this study, based on monthly average spectra from the SCIAMACHY payload, VERs of O₂1.27 μm nightglow from 2002 to 2012 are retrieved using a linear global optimization algorithm, and the accuracy of the inversion algorithm is evaluated. Furthermore, the spatial structure and temporal variation of the VERs of O₂1.27 μm nightglow in the MLT region are analyzed, and the periodic variation of oscillation frequency is summarized through harmonic analysis.

This paper is organized as follows: Section 2 introduces the instruments, datasets, and analysis methods. Section 3 describes the O₂1.27 μm nightglow VERs results obtained from linear inversion, including characteristics of the nightglow’s spatiotemporal distribution and periodic variation, and briefly discusses a possible physical mechanism. Section 4 summarizes our findings.

2.   Data and Methods

2.1   Data

As a passive detection instrument, SCIAMACHY is part of the atmospheric chemistry payload on the Envisat satellite launched by the European Space Agency (Bovensmann et al., 1999). Envisat was in service from March 2002 to April 2012 and operated on a polar sun-synchronous orbit, with 14 tracks per day and 35 days global coverage. The instrument covered the tangent altitude range of 75−150 km through limb observation, with a vertical spacing of 3.3 km (Sun K et al., 2022). It is worth mentioning that nighttime limb measurements were all taken at 22:00 Local Solar Time (LST) (von Savigny, 2015). SCIAMACHY was tasked with observing all atmospheric radiation in the wavelength range of 280−2380 nm (Bovensmann et al., 1999). Figure 1 illustrates the geographical coverage of SCIAMACHY nighttime limb measurements as a function of time and latitude. As a whole, the maximum latitude at which data could be collected in the northern hemisphere was 77.5°N; in the southern hemisphere, the maximum latitude was 47.5°S. In this study, Level 1 data in channel 6 (1200−1360 nm) are used, which are monthly zonal mean O₂1.27 μm nightglow spectra, in 5° latitude bins.

Figure  1.  Latitude coverage of SCIAMACHY limb nighttime observations.

DownLoad: Full-Size Img PowerPoint

2.2   Retrieval Method

First, spectral data from SCIAMACHY are preprocessed. Two defective pixels at 1262 nm and 1282 nm in the ${\text{O}}_{\text{2}}({\text{a}}^{\text{1}}{\Delta}_{\text{g}})$ band are excluded, and the average spectral intensity above 110 km is then considered as a background value, as follows: According to previous studies, the main altitude range of the O₂1.27 μm nightglow distribution is 80−110 km (McDade et al., 1987; Lopez-Moreno et al., 1988). Information collected by the instrument above 110 km is thus mainly from radiation in the background atmosphere rather than the nightglow emissions, so the data beyond this altitude is regarded as background values. To eliminate the atmospheric background effect, this average value is subtracted from all valid data.

Figure 2 displays the distribution of spectral intensity at different heights after background value filtering. The wavelength sampling interval in Figure 2 is approximately 0.76 nm. The spectral data, processed through the aforementioned steps, are now ready for subsequent linear global optimization inversion.

Figure  2.  Nightglow spectral intensity around O₂1.27 μm at various altitudes, observed at ${30} - {\text{35°N}}$ in April 2008.

DownLoad: Full-Size Img PowerPoint

In this study, the maximum posterior (MAP) method is used to do a linear global optimization inversion to obtain the VERs (Rodgers, 2000). The fundamental equations of the algorithm are outlined below:

where ${\boldsymbol{x}}_{{a}}$ is the mean value of the prior information, ${\boldsymbol{S}}_{{a}}$ is the covariance of the prior information, $\boldsymbol{y}$ is the spectral radiance measured by SCIAMACHY, ${\boldsymbol{S}}_{{e}}$ is the error covariance matrix of the spectral radiance, and $\boldsymbol{K}$ is expressed as the weight function representing the linear relationship between the VER and the spectral radiance. The inverted VER is denoted as $\widehat{\boldsymbol{x}}$.

Note that in the maximum posterior method, the ${\boldsymbol{x}}_{{a}}$ is a zero matrix, which means the prior information is set to zero. ${\boldsymbol{S}}_{{a}}$ is a matrix with a constant value of 1.1 × 10⁻⁵ photons s⁻¹ cm⁻³ for diagonal elements, representing altitudes between 70 km and 110 km which then exponentially decays to zero for grid points above and below these levels (Li AQ et al., 2021). Considering that the self-absorption effect of oxygen becomes more significant below 60 km (Sun K et al., 2022), and that the target altitude of this study is above 70 km, we ignore the influence of ${\text{O}}_{\text{2}}$ self-absorption. The loss in the radiative transmission path is not taken into account, meaning that the matrix $\boldsymbol{K}$ denotes essentially the optical path length of each layer scanned by the spaceborne instrument.

In addition, the quality of the inversion is evaluated by the average kernel Matrix (AVK Matrix), abbreviated as $\boldsymbol{A}$, calculated with the following equation:

This matrix represents the extent of deviation from the actual state with respect to vertical resolution. As shown in Figure 3a, within the range of 80−100 km, the measurement contribution (sum over each row of the averaging kernel) is approximately one, and the averaging kernels peak at the corresponding tangent altitudes. Figure 3b illustrates that the vertical resolution is 3.3 km, which closely aligns with the sampling distance measured by SCIAMACHY.

Figure  3.  The averaging kernel and the vertical resolution of the retrieval at ${0} - {\text{5°N}}$ in April 2008.

DownLoad: Full-Size Img PowerPoint

3.   Results and Discussion

3.1   Inversion Results

An independent emission profile can be derived by integrating the inverted VER of each individual spectral line along the entire band. An example is illustrated in Figure 4, displaying the altitudinal variation of the ${\text{O}}_{\text{2}}^{\text{1}}{\Delta }\;{\text{band}}$ nightglow VERs in the wavelength range of 1260−1271 nm. The peak of the nightglow band-VER is $9.5\times 10^{4}$ photons s⁻¹ cm⁻³, occurring at an altitude of approximately 90 km.

Figure  4.  The inverted VERs of nightglow emissions in different wavelengths at ${20} - \text{25°N}$ in April 2010. The dark red line represents the integrated VERs of each individual limb spectrum.

DownLoad: Full-Size Img PowerPoint

The linear inversion results of SCIAMACHY and monthly zonal mean SABER VERs are shown in Figures 5a and b. Compared with Figures 5a and b, the spatial structure of the inversion results resembles the data detected by SABER, showing similar peak height and peak value above the equatorial region. The VER peak of nightglow emission is ${20} - 25\times 10^{{4}}\;{\rm{photons}}\;{\text{s}}^{{-1}}\,{\text{cm}}^{{-3}}$ and the peak height is 86−88 km. However, as shown in Figure 5c, there is bias between the two datasets, with the calculated average deviation being approximately $-1.3\times {{10}}^{{4}}\;{\rm{photons}}\;{\text{s}}^{{-1}}\,{\text{cm}}^{{-3}}$. The largest deviation, approximately 40%, appears at 90 km. The large bias between the two datasets, possibly due to the different detection principles of the two instruments. It is worth mentioning again that the two datasets are monthly zonal mean data. Figure 6 shows the profiles of oxygen nightglow VERs from the two different datasets as a function of latitude and altitude. The SCIAMACHY dataset at 40°S−10°N is systematically smaller than the SABER dataset, with a maximum deviation of 25%; at 10−40°N, however, the two datasets demonstrate relatively consistent results.

Figure  5.  (a) Distribution of inverted VERs based on SCIAMACHY observations in April 2007. (b) Distribution of VERs in April 2007 after filtering from SABER measurements. (c) Distribution of difference between the inverted VERs and filtered data from SABER. The black and orange error bars represent the uncertainty of the inversion results and of the SABER filtered dataset, respectively.

DownLoad: Full-Size Img PowerPoint

Figure  6.  Comparison of VERs between two datasets at different latitudes and altitudes. The orange lines represent the SABER filtered dataset; the black lines represent the inversion results from the SCIAMACHY dataset. Panels (a)−(d) give results for the southern latitude region; panels (e)−(h) present results for the northern latitude region.

DownLoad: Full-Size Img PowerPoint

3.2   Spatial and Temporal Distribution

Figure 7 shows the contours of O₂1.27 μm nightglow VERs in 2007, as inverted using the linear algorithm as a function of latitude and altitude. Due to changes in the satellite's orbit, the latitude coverage measured by SCIAMACHY was limited. In the mesopause region above the equator, the O₂1.27 μm nightglow emission reveals evident annual and semi-annual variations, with similar annual variations appearing at mid-latitudes in the northern hemisphere. The data from March reveal two distinct emission peaks in the mesopause region, with one peak at 88 km above the equator and a weaker one at 91 km above the latitude of 60°N. The maximum annual VER was recorded in April with a value of around 2 × 10⁵ photons ${\text{s}}^{{-1}}\,{\text{cm}}^{{-3}}$; the peak altitude was around 88 km. In September, two peaks appeared, with the weaker one at 88 km above the equator and the stronger one at 90 km above the latitude of 45°N. In March, by contrast, the stronger peak was above the equator. The bimodal structure in October closely mirrors that observed in September. In November and December, the intensity of VERs over the 45°N region decreased significantly, and the gap with the equator region gradually widened.

Figure  7.  Spatial distribution of inverted VERs of O₂1.27 μm nightglow in 2007 after the interpolation.

DownLoad: Full-Size Img PowerPoint

Figure 8 illustrates the time-altitude distribution of O₂1.27 μm nightglow VERs inverted by the linear algorithm from 2002 to 2012. The altitude range is 79−96 km, and the inverted VERs outside this interval are very small, at least an order of magnitude smaller than the peak. Comparing the intensity of the O₂1.27 μm nightglow VERs from 2002 to 2012, it appears that the VERs show a relatively strong dependence on solar activity, although the dataset does not cover a complete 11-year cycle. In 2002−2003, with maximum solar activity, O₂1.27 μm nightglow reached a corresponding peak in emission. Subsequently, as solar activity decreased and reached its minimum at the end of 2008 and beginning of 2009, the intensity of O₂1.27 μm nightglow experienced a trough. With the beginning of a new solar cycle, the increasing solar activities led to a gradual increase in the VERs of O₂1.27 μm nightglow. As shown in Figure 8, the intensity of nightglow regained its peak in April 2012. Figure 8 also shows that the responses to solar radiation vary with latitude. The average emission intensity near the equator region is significantly stronger than that at 20°N, which is generally consistent with the results of Gao H et al. (2016); they analyzed the 13-year average oxygen nightglow intensity to study the response to solar radiation and pointed out a large peak around the equator, two small peaks around 35°S/N, and two valleys around 20°S/N.

Figure  8.  Inverted VERs of O₂1.27 μm nightglow in the time-altitude frame. From top to bottom, the latitudes are ${5} - \text{10°N}$, ${10} - \text{15°N}$, and ${15} - \text{20°N}$, respectively. The white areas in the figure denote data gaps.

DownLoad: Full-Size Img PowerPoint

As shown in Figure 8, over the past 11 years the intensity of the oxygen nightglow VERs was stronger in spring and summer than in autumn and winter. Two peaks and one valley occur each year — the valley in July, the first peak in April, and the second peak in October, albeit with lower intensity than the April peak. This pattern is consistent with the conclusions of Gao H et al. (2016), that seasonal variations dominate the time series of oxygen nightglow intensity at all latitudes.

3.3   Oscillation Period

To quantitatively analyze the temporal variation of inverted O₂1.27 μm nightglow VERs, we used harmonic analysis to extract semi-annual, annual and 11-year solar cycle variations. All the VERs of each latitude point in the altitude range of 79−106 km per month were added and summed to obtain a dataset suitable for harmonic analysis. The polynomial fitting algorithm was then applied to determine the short-term variation and long-term tendency of the oscillation frequency. The multiple linear regression analysis used the following equation:

where, F_10.7 is the F_10.7 cm solar flux proxy in units of ${\text{10}}^{{-22}}\;{\text{Wm}}^{{-2}}\;{\text{Hz}}^{{-1}}$ (Tapping, 2013), and t represents months. ${A}_{F10.7}$ is the amplitude of the F_10.7 cm solar flux. ${A}_{\mathrm{S}\mathrm{A}\mathrm{O}}$, ${A}_{\mathrm{A}\mathrm{O}}$, ${P}_{\mathrm{S}\mathrm{A}\mathrm{O}}$ and ${P}_{\mathrm{A}\mathrm{O}}$ are the amplitudes and phases of the Semi-Annual Oscillation (SAO) and Annual Oscillation (AO), respectively. ${A}_{\mathrm{Q}\mathrm{B}\mathrm{O}}$ and ${P}_{\mathrm{Q}\mathrm{B}\mathrm{O}}$ are the amplitudes and phases, respectively, of the Quasi-Biennial Oscillation (QBO) corresponding to 24 months. The fitting parameters include the amplitudes and phases of the SAO, AO and QBO, supplemented by a factor (${A}_{F10.7}$) scaling the F_10.7 cm solar flux proxy, and a constant term. The “shift” is the time lag of the solar cycle impact while the “offset” is the mean value of the time series (Zhu YJ, 2016).

Figure 9 displays the harmonic analysis results of inverted O₂1.27 μm nightglow VERs at ${10} - {{15}}{\text{°}}\text{N}$. The SAO maximizes in April and October, and minimizes in June and January. It shows an amplitude of about 9.4%. In addition, the AO has a maximum near the summer and a minimum near the winter and its amplitude is about 9.0%. Annual mean differences of VERs between the maximum (2002) and minimum (2008) of the 23rd solar cycle reveal variations of 19.8%.

Figure  9.  Harmonic analysis of O₂1.27 μm nightglow VERs at ${10} - \text{15°N}$. In panel (a), the black dots represent the inverted data and the black line is the fitted curve. Panel (b) presents the offset and residual between the fitted values and the inverted data. Panels (c)−(f) present the fitting results of SAO, AO, QBO and solar components, respectively.

DownLoad: Full-Size Img PowerPoint

The semi-annual and annual variations in the inverted O₂1.27 μm nightglow VERs are consistent with those reported by Gao H et al. (2016). However, a comprehensive understanding of the underlying physical mechanisms requires further detailed investigation. Similar oscillation characteristics are observed in ${\text{O}}_{\text{2}}\text{A (0-1)}$ band airglow, but the distribution of airglow intensity differs with latitude and altitude (López-González et al., 2004; Perminov and Pertsev, 2010). Perminov and Pertsev (2010) attributed the semi-annual variation in the ${\text{O}}_{\text{2}}\text{A (0-1)}$ band to change in atomic oxygen density driven by atmospheric circulation. Besides, a study reported on global observations of the (${\text{O}}^{\text{1}}\text{S}$) equatorial airglow by WINDII shows a strong local time control on the vertical distribution of VERs from the atomic oxygen green line airglow and concludes that it originates in tidal dynamics (Shepherd et al., 1995). Liu GP et al. (2008) indicated that vertical advection associated with the tides and the large-scale circulation plays a major role in airglow seasonal variations. The semi-annual variation, which is dominant around the equator, can be attributed mainly to the influence of diurnal tide (Gao H et al., 2016). The semi-annual variation in O₂1.27 μm nightglow might also be explored in conjunction with photochemical reaction components and atmospheric circulation transport processes.

As a key reactant in the production of O₂1.27 μm nightglow, atomic oxygen plays a crucial role in influencing photochemical reaction rates and, consequently, VERs intensity. The movement of atomic oxygen with atmospheric circulation impacts the intensity of O₂1.27 μm nightglow emissions. Ward (1999) simulated how tides affect atmospheric density, temperature, and atomic oxygen mixing ratio, which in turn regulate the behavior of nightglow emissions. He confirmed that nightglow emissions are enhanced with the increase of tidal amplitude, which was particularly obvious at low latitudes. Hays et al. (2003) highlighted the downward transport of atomic oxygen from the thermosphere by tides, enhancing nightglow emissions in the MLT region, while the weakened region of nightglow was associated with ascending tide motion. The modulating effect of the tides is strong on the atomic oxygen in the mesopause because of its long lifetime (Gao H et al., 2011). On the other hand, the collision of excited OH molecules with oxygen molecules or oxygen atoms can also produce ${\text{O}}_{\text{2}}$ nightglow, so there may be a certain correlation between ${\text{O}}_{\text{2}}$ nightglow and OH nightglow (Lednyts’kyy and von Savigny, 2020). Based on satellite observations, Marsh et al. (2006) indicated that the OH nightglow intensity and its temporal variation were strongly modulated by tides. Therefore, similar to the OH nightglow, the altitude-latitude distribution of ${\text{O}}_{\text{2}}$ nightglow emission is influenced by tides. Gao H et al. (2011) found that both the highest radiative intensity and the lowest peak height of O₂1.27 μm nightglow occurred over tropical areas during the spring and autumn equinoxes, with greater intensity observed during the spring equinox. Xu et al. report that the amplitude of diurnal tidal is larger during the March equinox than during the September equinox (Xu JY et al., 2009).

Our research (see Figure 9) confirms these equinoctial extremes, which we suggest may indeed be attributed to tidal influences. The downward transport of atomic oxygen increases under the influence of tides. The three-body reaction O + O + M→O₂* + M is the main source of the oxygen 1.27 μm emissions. These factors affect the intensity of ${\text{O}}_{\text{2}}$ nightglow emission during the spring equinox. Ward (1999) also pointed out that the modulation of airglow emission by tides is related to the phase and wavelength of tides. In addition, Zhu YJ (2016) identified a close relationship between the generation mechanism of oxygen airglow and solar ultraviolet radiation, affecting atomic oxygen concentration. The photochemical factors involved in airglow emissions, including formation reaction rates and quenching rates with the surrounding atmospheric components, are temperature-dependent (Liu WJ, 2016). Another study showed that the temperature at the mesopause region significantly responds to the electron precipitation during periods of intense solar activity (Zou ZC et al., 2020). Previous observations and model studies have proven that, when geomagnetic events are frequent, energetic particles stored in the magnetotail are accelerated Earthward to impact the dynamics and chemical composition of the middle and upper atmosphere (Sinnhuber et al., 2012; Yi W et al., 2018; Zou ZC et al., 2018). Solar activity’s influence on MLT temperature (Zou ZC et al., 2020) and density (Yi W et al., 2018) can lead to changes in airglow emission intensity.

4.   Conclusions

Based on limb spectra data from 2002 to 2012 collected by the SCIAMACHY payload, Volume Emission Rates (VERs) of O₂1.27 μm nightglow were obtained using a linear global optimization algorithm. The results are summarized as follows:

(1) A comparison between the O₂1.27 μm nightglow VERs data measured by the SABER payload and the linear inversion results of this study demonstrates consistency between the two datasets. The deviation between them is negligible, an average of $-1.3\times {\text{10}}^{\text{4}}\;\text{photons}\;{\text{s}}^{{-1}}\,{\text{cm}}^{{-3}}$, indicating that the linear inversion results accurately reflect the observed spatial and temporal distribution characteristics.

(2) The inverted VERs of O₂1.27 μm nightglow reveal a peak at 88 km above the equator in spring, particularly in April. Two other peaks occur between August and October, with the peak intensity over the tropical region at 88 km weaker than the peak over 45°N at 90 km.

(3) The inverted VERs of O₂1.27 μm nightglow show a dependency on the 11-year solar cycle. That is, nightglow intensity strengthens with increasing solar activity and vice versa, displaying obvious semi-annual and annual variations.