Journal of Geophysical Research: Space Physics

Long-term variation in the thermosphere: TIMED/GUVI observations



[1] Long-term changes in the thermospheric composition (O/N2 column density ratio) and exospheric temperature (Texo) have been examined. The data are obtained from TIMED/GUVI disk and limb measurements over half of a solar cycle from 2002 to 2007. The data indicate a positive correlation between O/N2 and solar EUV flux (Qeuv), despite large variability in O/N2 due to geomagnetic activity and season and local time effects. A function fitting provides quantitative specification of the O/N2 dependence on Qeuv. The magnitude of changes in O/N2 due to solar EUV variation between 2002 and 2007 is 40% of the mean (0.5) O/N2 over the same period and is of the same order to the changes in O/N2 due to nonsolar EUV effects (geomagnetic activity, season, and local time). Similar results are also obtained for the Texo dependence on Qeuv. The O/N2 dependence on Qeuv is due to thermal expansion or contraction that alters the reference height of the fixed N2 column density (1017 cm−2). This differs from the O/N2 depletion associated with high Texo during storm time where upward wind plays an important role.

1. Introduction

[2] Thermosphere changes continuously because of energy and momentum inputs from the solar radiation [Richards et al., 1981; Burns et al., 2004; Forbes et al., 2006; Solomon et al., 2010], particle and Joule heating [Fuller-Rowell et al., 1994; Rodger et al., 2001; Baker et al., 2004; Zhang et al., 2005; Zhang et al., 2010a], inter hemisphere transport or seasonal effects [Waldteufel, 1970], and waves (e.g., tides [Hagan and Forbes, 2002; Häusler et al., 2010; Zhang et al., 2010b]). Condition and convection in the thermosphere strongly affects the ionosphere [Rishbeth, 1998]. Intense magnetic storms are the major driver for disturbances in thermosphere over a time scale of hours to days [Mayer et al., 1978; Prölss, 1980, 1997; Fuller-Rowell et al., 1996; Zhang et al., 2003, 2004]. Solar rotation leads to thermosphere changes in time scale of 27 days [Fleming et al., 1995] and its harmonics (e.g., 9 days [Crowley et al., 2008; Lei et al., 2008; Zhang et al., 2010a]), season variations, solar cycle dependence and secular changes [Bremer et al., 1997; Keating et al., 2000].

[3] The term “long-term variation” has different definitions in various papers. For example, seasonal variations over a period of 2 years were considered a long term by some investigators [Alcaydé et al., 1974; Lieberman, 1997]. Others defined “long-term” as a few years [Hernandez, 1982; Breig et al., 1985], one solar cycle (11 years) [Fleming et al., 1995] and a few decades [Bremer et al., 1997; Keating et al., 2000; Laštovička, 2005; Marcos et al., 2005]. Here we refer the long term to a period of one half of a solar cycle between 2002 and 2007.

[4] It is well known that the thermospheric density, temperature and composition strongly depend on solar EUV flux, solar wind (geomagnetic activity), season and local time. Based on simultaneously measured thermospheric N2 densities and solar EUV fluxes obtained by the AE - E satellite, Hedin [1984] found that short-wavelength (coronal) EUV emissions correlate better with thermospheric density than the F10.7 cm flux, although the reduction in density residuals is small. Observations of nightside kinetic thermospheric temperature from Doppler width of the [OI] (630 nm) emission at 39.87°N, 105.05°W indicates that the temperature depends on four parameters: solar activity, solar declination, geomagnetic activity and a semiannual variation [Hernandez, 1982]. Stamper et al. [1999] found that the near-Earth interplanetary space makes a significant contribution to the long-term increase in geomagnetic activities. Specifically, the largest contribution to the drift in geomagnetic activity over solar cycles 20-22 originate from rises in the average interplanetary magnetic field (IMF) strength, solar wind concentration, and speed. There are many reports that magnetic storms cause significant decrease in thermosphere O/N2 ratio and ionosphere density [Prölss, 1997; Fuller-Rowell et al., 1996; Zhang et al., 2003, 2004]. Based on radar measurement at midlatitude (45°N) between 1969 and 1970, Alcaydé et al. [1974] found that seasonal variation of 1.5 in winter to summer ratio of O density, and 4 for N2 summer to winter density ratio at 200 km altitude. However, they did not find significant correlation between F10.7 flux and the normalized O density at the altitude.

[5] In addition to the sources of the thermospheric variations discussed above, increasing green house effect may lead to secular change in thermosphere. Bremer [1992] examined ionosonde measurements at 54.6°N, 13.4°E between 1957 and 1990 and found hints of a decrease of the peak height of the F2 layer. This quantitatively agrees with the prediction of Rishbeth [1990], who expected a lowering of the E and F2 layer caused by a global cooling of the stratosphere, mesosphere, and thermosphere because of the increased greenhouse effect. Keating et al. [2000] shows the evidence of a decline averaging 9.8% ± 2.5% in thermospheric density over 20 years pointing toward a long-term cooling of the upper atmosphere. Besides the global cooling, there are other possible causes of secular long-term changes in the ionosphere [Rishbeth, 1998]. For example, changes in tidal forcing in the middle atmosphere resulting from ozone depletion [Ross and Walterscheid, 1991] might lead to dynamical changes in the thermosphere.

[6] The above secular trends need continuous observations over tens of years or longer period to be identified. On other hand, the thermospheric variations are dominated by geomagnetic storms, local time and seasonal effects over a shorter-time period (days to months). Here we focus on solar cycle (solar EUV flux) dependence of O/N2 and Texo based on multiyear GUVI observations. The enhanced (reduced) solar EUV heating of the thermosphere leads to increase (decrease) in the thermospheric temperature and Texo. This causes thermal expansion and contraction of the thermosphere. Such change affects both O and N2 density profiles in similar ways. One may expect that the O/N2 ratio remain unchanged. Here we show that O/N2 does change with solar EUV as the reference height of a fixed N2 column density varies so as the O column density.

2. Data

[7] Disk and limb far ultraviolet (FUV) radiance data from TIMED/GUVI were used to estimate the auroral hemispheric power, dayside thermospheric neutral density and temperature profiles. GUVI (Global Ultraviolet Imager) on the NASA TIMED satellite (launched on 7 December 2007 into a 630 km circular orbit with an inclination of 74°) provides disk and limb images of major airglow and auroral emission features. They include Lyman α, 121.6 nm), OI (130.4 nm), OI (135.6 nm) lines and N2 Lyman-Birge-Hopfield (LBH) bands. The LBH bands are denoted LBHS, for LBH “short” and LBHL for LBH “long” and cover the approximate ranges of 140 to 150 nm and 165 to 180 nm, respectively [Paxton et al., 1999; Paxton and Meng, 1999; Christensen et al., 2003; Paxton et al., 2004]. The thermospheric O to N2 column density ratio (O/N2) are obtained from the disk 135.6 nm and LBHS dayglow data [Zhang et al., 2004]. The GUVI dayglow limb data can be used to find the thermospheric temperature and density [Meier et al., 2005]. GUVI is a very stable instrument. A number of stellar calibrations between 2002 and 2007 indicate that there is no detectable degradation or long-term trend in GUVI sensitivity over the 6 year period. More information can be found on the GUVI website (

[8] In addition to the GUVI data, the solar EUV flux between wavelengths 26 and 34 nm are also used to find the dependence of the O/N2 and thermosphere temperature on the solar EUV flux. The Charge, Element, and Isotope Analysis System/Solar Extreme Ultraviolet Monitor (CELIAS/SEM) transmission grating spectrometer onboard SOHO satellite has been continuously providing the full disk solar extreme ultraviolet and soft X-ray irradiance with high quality since 16 December 1995 [Hovestadt et al., 1995; Judge et al., 1998, 1999, 2002]. Though SEM is a highly stable instrument, modest degradation has been detected based on rocket measurements and calibrations over years. The degradation has been modeled and taken into account to the final SEM data products [Judge et al., 2002]. This provides a reliable and accurate solar EUV flux over many years. The solar EUV flux (26–34 nm) data [Judge et al., 1998] are obtained from Selection of the 26–34 nm data is because that the data include the solar He II 30.4 nm emission which is the brightest EUV (e.g., 10–100 nm) emission [Hinteregger et al., 1981; Woods et al., 1998]. The He II 30.4 nm solar radiation is also the dominant source of ionizing and heating radiation in the thermosphere [Roble, 1995].

[9] Figure 1 shows daily zonal mean of GUVI disk O/N2 data over a period of 6 years between 2002 and 2007. This period covers the decline phase of the 23rd solar cycle. The V or inverted V shaped structure indicates the season variations each year. The black areas in middle- to high-latitude region are the regions without O/N2 data. A few black vertical lines also indicate data gap. The vertical structures with blue or dark blue colors are due to reduced O/N2 resulting from geomagnetic activities. Regardless the vertical structures, the global O/N2 tend to decrease from 2002 to 2007 (this can be easily seen Figure 2). While the selected color bar in Figure 1 may make it difficult to identify the O/N2 trend, the GUVI O/N2 in Figure 1 is further averaged in latitude to provide daily global mean O/N2 that is plotted in Figure 2 (black dots). Despite a large scatter (owing to season, local time and geomagnetic effects), the GUVI daily mean O/N2 shows a rapid decrease between 2002 and 2004. The O/N2 reduction is weakened after 2004. The over plotted solar EUV flux (red line in Figure 2) also shows similar feature: rapid decrease between 2002 and 2004 and much slower decrease after 2004. This suggests that the solar EUV flux has noticeable effect on the global mean O/N2 or background O/N2 ratio.

Figure 1.

GUVI disk O/N2 daily zonal mean from 2002 to 2007.

Figure 2.

GUVI disk daily mean O/N2 (black dots) and solar EUV flux (26–34 nm) from SOHO/SEM instrument (red line) between 2002 and 2007.

[10] The dependence of global mean O/N2 on the solar EUV flux can be better seen in a scatterplot (Figure 3). Figure 3 replots the data in Figure 2 (daily mean O/N2 and their associated solar EUV flux). As it is expected, the plot in Figure 3 has a large scatter. Nevertheless, there is a clear trend: high O/N2 tends associated with high solar EUV flux. To determine the trend, a nonlinear function fitting was performed to the data set and it leads to following fitting function (see red line in Figure 3):

equation image

where Qeuv is the solar EUV flux (26–34 nm) from SOHO/SEM in unit of 1010 photons/cm2/s. The blue lines in Figure 3 indicate the standard deviation (0.113).

Figure 3.

A scatterplot of GUVI disk O/N2 and solar EUV flux (same data set in Figure 2). The red line is the fitted curve between the O/N2 and EUV flux. The blue lines show the standard deviation.

[11] The fitting function confirms that the underline O/N2 is positively correlated with the solar EUV flux. Equation (1) indicates that when the solar EUV flux varies from 1.0 (solar minimum condition) to 3.0 (solar maximum condition), the net O/N2 changes from 0.42 to 0.62. The difference (0.2) is 40% of the mean (0.5) of all O/N2. On the other hand, under a given solar EUV flux (Figure 3), the magnitude of variation in O/N2 is about 2 × 0.113 = 0.226. Such variation is due to season, geomagnetic and local time effects. Thus, the solar EUV or solar cycle effect on O/N2 is as big as the effect of all other factors mentioned above. This indicates that the solar EUV flux plays a very important role on the long-term changes in O/N2.

[12] It is well known that changes in solar EUV flux lead to variations in the thermosphere temperature including the exosphere temperature (Texo). In addition to the disk imaging data, GUVI also provides limb data. The dayside limb data are used to retrieve neutral density and temperature profiles of O, N2 and O2 [Meier et al., 2005]. The Texo based on GUVI limb data are plotted in Figure 4 (black dots) with solar EUV flux (red line). The large fluctuation in Texo is also due to season, local time and geomagnetic activity. However, it is clear that Texo shows a decreasing trend with decreasing solar EUV flux between 2002 and 2007. Such a trend is similar to the O/N2 trend in Figure 2. The Texo dependence on solar EUV flux is more clearly seen in Figure 5, a scatterplot of Texo versus solar EUV flux. A fitted function can be listed as:

equation image

This fitted function indicates that the background Texo (excluding the nonsolar EUV flux effect) increased from ∼700° to 1350°K when the Qeuv changed from 1 to 3. The net difference (1350–700 = 650) is about 66% of the mean Texo (980°K) over the period between 2002 and 2007. It is also much larger than the magnitude (2 × 120 = 240°K) of Texo variation due to nonsolar EUV effect.

Figure 4.

Time history of exospheric temperature (Texo, black dots) and the solar EUV flux (26–34 nm from SOHO/SEM, red line) between 2002 and 2007.

Figure 5.

A scatterplot (black dots) of Texo and solar EUV flux. The red dotted line is a fitting curve.

[13] Enhanced solar EUV heats the thermosphere and leads to thermal expansion of the thermosphere. This changes the density profiles of both O and N2. One may expect that the changes in O/N2 ratio will be small. However, as O density is higher (lower) than the N2 density at high (low) altitudes, respectively. The O/N2 ratio referenced at a fixed N2_column = 1017 cm−2 may still have a big change because of change in the reference height. The effect of the reference height on O/N2 ratio can be addressed by the MSIS model. To do this, we run MSIS86 model [Hedin, 1987] under a fixed UT (12:00), latitude (41), longitude (0), extremely quiet geomagnetic activity (Ap = 1) and different F10.7 flux (50 to 320). Figure 6 shows the dependence of O/N2 (black line) and reference height (red line) versus F10.7 flux. This plot demonstrates that O/N2 variation is mainly due to the changes in the reference height. When the reference height is moved to higher altitude, the relative column abundance of O is higher than N2. When the reference height moves to a low altitude, the O/N2 reduces. It is important to note that the O density profile variation due to changes in thermospheric temperature also contributes to the O/N2 variation.

Figure 6.

O/N2 (black line) and reference height (at N2_column = 1017 cm−2, red line) dependence on F10.7 cm flux from MSIS86 model.

3. Discussion

[14] Large variations in O/N2 over a short-time period (days) and mediate time scale (months) are mainly dominated by geomagnetic activity and season effect, respectively. The northern summer O/N2 between latitude of 40° and 50° N in 2002 is around 0.2–0.4 which is about half or less the values (∼1.0) in northern winter (December 2002) or southern winter (June 2002) between latitude −50° and −40° (Figure 1). This O/N2 season effect pattern in 2002 is repeated in the following years. Such season effects are also modulated by local time change in TIMED orbit (every 60 days). Over a short time scale (days), the O/N2 reduction is clearly seen at almost all latitudes from time to time. Nevertheless, the 6 year period (2002–2007) is long enough to show the general decrease in O/N2 due to the decrease in solar EUV flux. Though it can't be easily identified, the O/N2 also varies with solar rotation (see Qeuv in Figure 2). Despite that the daily mean O/N2 or the mean of all O/N2 between 2002 and 2007 are affected by the seasonal, local time and magnetic activities, the net change (0.2) in O/N2 (Figure 3 and equation (1)) when the Qeuv increased from 1.0 to 3.0 is significant and nonnegligible. Figures 2 and 3 also indicate that at the low Qeuv (e.g., around 1.0), the O/N2 tends to bottom and at high Qeuv (e.g., >2.5), O/N2 tends to increase rapidly with Qeuv, indicating nonlinear response of O/N2 to changes in Qeuv. Such a feature is consistent with the idea that O/N2 changes associated with Qeuv are mainly due to the reference height variation (Figure 6) and the O density profile change. At low Qeuv or low Texo (low scale height of the neutrals), further decrease in Qeuv will have little effect on the reference height as small change in height will cause a large change in neutral density. However, the Texo tends to saturate at high Qeuv (Figure 5).

[15] It is interesting to note that O/N2 reduces during geomagnetic storms due to enhanced vertical and horizontal wind [Meier et al., 2005]. High Texo is associated with reduced O/N2. On other hand, high Texo due to enhanced Qeuv increases O/N2. Such dramatic different thermospheric responses can be understood by the differences in size of spatial areas (local versus global), time scales of the aurora and Qeuv heating, and dynamics (nonequilibrium or equilibrium) of the thermosphere. The Qeuv heating changes slowly (except for solar flares) and covers whole globe. This leads to global thermosphere expansion. Therefore, redistribution of the thermosphere neutrals between different areas is limited. On the other hand, during geomagnetic storms, localized heating in high-latitude auroral regions is enhanced over a short time scale (minutes to hours). Such a process leads to strong vertical and horizontal wind. Unlike N2 density profile that increases monotonically with decreasing altitude, the O density profile peaks around 100 km and decreased with decreasing altitude below 100 km [Roble et al., 1987]. The strong vertical wind reduces the O density in the 100–200 km altitude region where it contributes most to the O column density. For N2, the strong upward wind actually increases the N2 density in the same altitude region due to huge supply from the N2 at altitude below 100 km. Such enhanced vertical and horizontal wind redistributes O and leads to O depletion in some areas and O enhancement in other areas.

[16] Finally, the cooling of the thermosphere due to the green house effect may also contribute to the thermospheric variations seen by GUVI. But because of the relatively short-time period, such effect is probably too small to be identified.

4. Summary

[17] The TIMED/GUVI thermospheric data products (O/N2 ratio and exospheric temperature) have been used to extract their dependence on solar EUV flux between 200 and 2007 (half of a solar cycle). Results show that over the 6 year period, O/N2 and Texo is positively correlated with solar EUV flux despite their large short-term (days to months) variations. The short- and mediate-term variations are caused by geomagnetic activity, seasonal, local time, and monthly solar EUV variation (solar rotation, especially at solar maximum) effects. A function fitting provides quantitative specification of the O/N2 dependence on Qeuv. The change (0.2) in O/N2 due to solar EUV variation between 2002 and 2007 is about 40% of the mean O/N2 (0.5) over the same period and is of the same order to the changes in O/N2 due to nonsolar EUV effects (geomagnetic activity, season, and local time). Similar results are also obtained for the Texo dependence on Qeuv. The net difference in Texo due to changing Qeuv between 2002 and 2007 is about 66% of the mean Texo over the same period. The O/N2 dependence on Qeuv is due to thermal expansion or contraction that alters the reference height of the fixed N2 column density (1017 cm−2) as well as the O density profile. This differs from the O/N2 depletion associated with high Texo during storm time where strong upward wind reduces O density and increases N2 density in the thermosphere.


[18] This work is supported by NASA grant NNX07AC57G.

[19] Robert Lysak thanks Alan Burns and another reviewer for their assistance in evaluating this paper.