Response of GOLD Retrieved Thermospheric Temperatures to Geomagnetic Activities of Varying Magnitudes

Global‐scale Observations of Limb and Disk (GOLD) disk measurements of far ultraviolet molecular nitrogen band emissions are used to retrieve temperatures ( Tdisk ), which are representative of lower thermospheric altitudes. The present investigation studies the response of lower thermospheric temperatures to geomagnetic activities of varying magnitudes. In this study, it has been observed that Tdisk increases over all latitudes in response to enhanced geomagnetic activity. The increase in temperature is proportional to the strength of the geomagnetic activity and is greater at higher latitudes. Temperature enhancements vary from 10s to 100s of Kelvins from low‐ to mid‐latitudes. Local time behavior shows that pre‐noon enhancements in temperatures, during relatively stronger geomagnetic activities, are greater compared to afternoon, which may be caused by the combined action of daytime dynamics and geomagnetic forcing. This study, thus, demonstrates the utility of GOLD Tdisk when investigating the effects of dynamical and external forcings in the thermosphere.

• Global-scale Observations of Limb and Disk (GOLD) thermospheric temperature increases globally in response to geomagnetic activity • The increase in temperature is proportional to the strength of the activity and is greater at higher latitudes • Temperature enhancement during active geomagnetic events is greater in the morning than that in the afternoon mechanism through which energy is dissipated in the lower thermosphere (Killeen et al., 1997;Mikhailov & Perrone, 2020). In short, a geomagnetic storm can alter the whole thermosphere-ionosphere (TI) system enormously, so synoptic observations of local time variability of the thermosphere during storm events are important for a complete understanding of space weather.
The neutral atmosphere at TI altitudes (about 100 km and up) is primarily investigated using in-situ (e.g., Forbes et al., 1996;H. Liu & Lühr, 2005;Spencer et al., 1981) and optical remote sensing techniques (Aksnes et al., 2006;Meier et al., 2015;Pallamraju et al., 2004Pallamraju et al., , 2013Pant & Sridharan, 1998). The temperature of the thermosphere can be retrieved from spectral broadening characteristics of atomic lines (Biondi & Meriwether, 1985;Chakrabarty et al., 2002;Fagundes et al., 1996;Pant & Sridharan, 1998) or molecular bands (Aksnes et al., 2006;Evans et al., 2018;Meier et al., 2015;Zhang et al., 2019). Ground based observations of the thermosphere have good local time coverage, at a cadence of minutes to hours for over at least 10 h per day, but they are mostly limited to the night-time sector and are available from limited ground stations (Biondi & Meriwether, 1985;Chakrabarty et al., 2002;Fagundes et al., 1996;Pant & Sridharan, 2001). Satellite based remote-sensing observations on the other hand have poor local time coverage and are mostly from limb viewing geometry (Aksnes et al., 2006;Meier et al., 2015). In-situ temperature measurements are also very limited, for example, those from Dynamic Explorer 2 (DE-2) mission (Spencer et al., 1981) or those retrieved from satellite drag (e.g., Mehta et al., 2017, and references therein). As most of the earlier satellite missions measuring thermospheric parameters were in quasi-sun synchronous orbits, they lack local time coverage. However, a constellation of quasi-sun-synchronous low earth orbits or very low inclination orbits can provide good local time coverage, but such configurations have not been used so far for thermospheric measurements. Thus, the earlier studies provided mostly a seasonally averaged or near-single local time behavior of the thermospheric temperature variability during geomagnetic activities. Therefore, the local time variations of the geomagnetic storm effects on thermospheric temperatures with global coverage have been limited mainly to model simulations . For a wide spatial and local time coverage one would require ground based stations covering the globe, or a constellation of sun-synchronous satellites in low-earth-orbit, or measurements from the geo-stationary orbit. The Global-scale Observations of the Limb and Disk (GOLD) mission, launched on January 14, 2018, is in geostationary orbit and provides Far-Ultraviolet (FUV) emission measurements, which can be used to retrieve thermospheric neutral temperatures . Though GOLD disk measurements are limited over a fixed hemisphere, they have very good coverage in latitude (  69 S to  69 N), longitude (  25 E to  120 W; covering America and parts of Europe and Africa), and local time (a minimum of 6-18 h near nadir longitude, which extends to even more local times depending on location).
Evidence of geomagnetic storm related changes in thermospheric temperatures, densities, and winds are ample in the literature (Aksnes et al., 2007;Astafyeva et al., 2020;Bagiya et al., 2014;Burns & Killeen, 1992;Burns et al., 1995;Crowley et al., 2006;Karan & Pallamraju, 2018;Mandal & Pallamraju, 2020;Mayr et al., 1978;Pallamraju et al., 2004;Pant & Sridharan, 2001;Strickland et al., 1999;Zhang et al., 2019). Due to particle precipitation, the temperatures at high-latitude are increased and the resulting circulation changes alter the whole thermospheric temperature (Burns & Killeen, 1992;Burns et al., 1995;Fuller-Rowell et al., 1994) with some delayed response at different altitudes and latitudes . Some recent investigations using GOLD O/ 2 N (ratio of atomic oxygen and molecular nitrogen column densities) (Cai, Burns, Wang, Qian, Pedatella, et al., 2021;Cai, Burns, Wang, Qian, Solomon, et al., 2021), oxygen-I (OI) 135.6 nm emission intensities (Gan et al., 2020), and limb temperature (Evans et al., 2020) data showed that even minor to moderate geomagnetic activities can impact the TI system significantly. But the majority of the earlier investigations of daytime thermospheric variability lack the local time and spatial coverage provided by the GOLD mission from geosynchronous orbit. The current study aims to use GOLD neutral disk temperature (T disk ) data to investigate synoptic and local time behavior of the thermospheric temperature during periods of enhanced geomagnetic activity of varying magnitudes.

Data and Methods
The primary data set used in this investigation is the GOLD retrieved neutral disk temperatures, T disk . Solar and geomagnetic indices are also used. GOLD observes the Earth's thermosphere in the FUV for over 18.5 h each day, from 0610 to 0040 UT of the next day Laskar et al., 2020;McClintock et al., 2020). The day-disk measurements cover about 0610 UT to 2300 UT, of which the majority of the disk falls in the day-light sector for about 9 h. GOLD daytime disk scans of the 2 N Lyman-Birge-Hopfield (LBH) bands are used to retrieve the T disk data. As the GOLD 2 N LBH emission measurements are column integrated quantities, the retrieved T disk products are a representative of the corresponding 2 N LBH layer. The peak altitude of the layer has a range of 150-220 km which varies with solar zenith angle (SZA) and emission angle. But the peak altitudes remain below 200 km for SZA and emission angles less than  70 (Evans et al., 2018;Laskar et al., 2021). GOLD scans each full disk in about 30 min. The retrieval algorithm is an improvement of the code that was used previously to derive temperature from limb measurements of 2 N LBH intensity from the High-resolution Ionospheric and Thermospheric Spectrograph (HITS) instrument (Aksnes et al., 2006;Krywonos et al., 2012). GOLD measurements have a higher spectral resolution that includes 2 N LBH band emissions within 132-162 nm range. Effective neutral temperatures are retrieved by fitting the observed rotational structure of the 2 N LBH bands using an optimal estimation routine (Evans et al., 2018;Lumpe et al., 2002;Rodgers, 2000). Five parameters are retrieved from each measured spectrum: rotational temperature (K), wavelength shift and dispersion (nm), background (counts/bin), and a forward model scale factor. The current investigation used Level 2 (L2) T disk version 3 (V03) data that are retrieved from 2  2 binned level-1C 2 N LBH spectra, which are available at the GOLD web-page, https://gold.cs.ucf. edu/. The 2  2 binned data have a spatial resolution of 250-km  250-km near nadir. Typical random errors in the 2  2 binned T disk data varies with signal-to-noise ratio (SNR) of the 2 N LBH emission and it ranges from 20 K (for high SNR) to 600 K (for low SNR). The arithmetic mean of the errors over the disk and for all local times is about 45 K. An earlier version (V02) of T disk data were retrieved from unbinned L1C data. The random error in those data were more than seven times higher than the current version, so we are using the V03 2  2 data for the current investigation.

Results
Sub-satellite local time (Sub-Sat. LT) versus day-to-day variation of GOLD T disk (from here on temperature or in short T) for about 2.5 years (October 14, 2018-March 15, 2021 of observations with ap index and solar F10.7 flux are shown in Figure 1a. These are averaged between  21 N to  53 N and  43 W to  54 W from the disk observations. This spatial bin is chosen arbitrarily and any other spatial combination also shows similar behavior. But selection of a very narrow bin results in a noisy signal. The average random uncertainty, which varies with LBH emission SNR, within the above chosen spatial bin is about 12 K. Notable features in this figure are: (a) the temperature is highest near 13-14 LT and afternoon is warmer than morning, and (b) the temperature increases over all local times in response to an increase in ap index or F10.7 flux. Figure 1b shows temperature deviations (T) from a baseline local time behavior. The baseline levels are calculated using all the quiet days (with ap  6 nT) that are within a 30-day running window of observation. Since the mean of the ap values during these 2.5 years of observations is 6.1  4.8 nT, we use ap  6 nT to define geomagnetically quiet times. We choose a 30-day window so that the seasonal changes are also removed once we subtract the 30-day mean. A shorter window will have fewer days with ap  6 nT and also we do not want to remove variations shorter than 30 days, where most of the dominant geomagnetic variations fall.
Though the solar flux has not exceeded 120 sfu (1 sfu =    22 2 1 10 Wm Hz ) during the period covered in this study, the temperatures show some increase with the slight enhancement in F10.7 cm flux during the last quarter of 2020, where it was above 110 sfu for a couple of days. This temperature increase with F10.7 cm flux is above the random uncertainty, which is about 12 K. A notable feature of T, as shown in Figure 1b, is that it increases with increasing ap index. Thus, observations in Figure 1 demonstrate that the GOLD temperature responds to both solar flux and geomagnetic activity. These findings are consistent with the response observed in GOLD exospheric temperatures due to minor geomagnetic activity reported by Evans et al. (2020). In addition to variations that are related to geomagnetic activity, annual variations can also be seen, for example, the winter (summer) morning temperatures are relatively colder (warmer) in Figure 1a. These annual variations are very interesting and will be addressed in a separate investigation. There is a gap of about 10 days of data during April 16-26, 2019, where the GOLD channel-A detector gain was low, which needed a grating yaw maneuver to overcome this. The data during this interval are available but should be interpreted with caution. Thus, they are not used in the present investigation. Due to this data gap the effective number of days having ap  6 nT will be smaller, but there are sufficient number of days with ap  6 nT within these 30 days windows to calculate the quiet-time background.
To quantify the relationship between ap and T, a scatter plot and a linear regression analysis are shown in Figure 2a. The scatter plot and the correlation analysis are done with the averaged T values between  21 N to  53 N and  43 W to  54 W and 10-14 h local time for all the days. The choice of local time range does not change the results significantly, but only impacts the random uncertainty (not presented here) but the choice of latitude range impacts the T. Each point in the scatter plot represents a day averaged within the above local-time and spatial bin. A correlation coefficient of  0.64 is observed between ap and T, which indicates that they are positively, though weakly, correlated. The not so strong positive correlation between ap and T can be due to other sources of temperature variability, for example, lower atmospheric waves  and non-linear response of temperature to geomagnetic-forcing (Connor et al., 2016), and solar flux variability. To further quantify their variabilities, Lomb-Scargle periodograms (Horne & Baliunas, 1986) of ap, F10.7, and T disk are shown in Figure 2b. Almost all the dominant periodicities that are seen in ap, such as, 6, 7, 9, 13-15, and 23-30 days can also be seen in T disk . These results demonstrate that the thermospheric temperature responds positively to geomagnetic activity. The F10.7 cm flux show some dominant periodicities around 27-day, the solar rotational period, which can also be seen in T disk and ap periodicities. As GOLD provides good latitude coverage from  69 S to  69 N, a latitudinal variation of correlation coefficients between ap and T is shown in Figure 3a   the two geomagnetic activity ranges is shown in Figure 3b. The temperature retrieval algorithm is not optimized to take into account the changes in LBH emissions due to energetic particle precipitation, so latitudes higher than  60 are not considered in this analysis. Note that the correlation coefficients are positive at all latitudes and are higher for stronger geomagnetic events. Also, the temperature enhancements are always positive and are greater at higher mid-latitudes. The greater T enhancements at higher mid-latitudes are in accordance with the fact that most of the energy deposition of the storm time particle precipitation occurs at high-latitudes. As the northern geomagnetic pole, for the current longitude sector, is closer to geographic equator the temperature enhancements are greater for the northern latitudes compared to southern hemisphere higher latitudes. It also demonstrates that the temperature enhancements occur over all latitudes. The percentage increase in T varies from 10% to 20% for the stronger events. These percentages are for an average of all the 27 individual events having ap  14 nT. However, the strongest event on August 31 to September 1, 2019 with a daily average ap of 43 nT showed about 25%-35% enhancement in temperature at low-to mid-latitudes. As the T values are calculated using a reference quiet-day that is, obtained from a 30-day running mean around the active day, they should be independent of any artifact that may arise from SZA and emission angle dependence of T disk (Evans et al., 2018). The higher temperatures at mid-latitudes as reported in the current work can induce thermospheric pole to equatorward circulation that transports oxygen rich air toward low-latitudes (Mayr et al., 1978), giving rise to enhanced OI 135.6-nm emission intensities at low-latitudes and decreased intensities at mid-and high-latitudes (e.g., Gan et al., 2020).
A unique feature of the GOLD mission is that it provides an unprecedented local time coverage, in addition to a wide spatial coverage. To investigate the local time variabilities, T and T variation at northern mid-latitudes (  32 N to  53 N) for the days with ap  20 nT are shown in Figure 4. Note that the number of individual active events with ap  20 nT was 16 in the 2.5 years period. Of the 16 active events, many of them lasted for several days. For this analysis only the first day of the active events is considered when comparing them with a pre-active quiet day. For a particular event, the pre-active quiet day is selected from five consecutive days just prior to the first active day. The day with minimum ap value among these 5 days is regarded as the quiet day. For most of the events the pre-active quiet day falls 2-3 days before the first active day. As LASKAR ET AL. there are 16 individual active events having ap  20 nT, the calculations in Figure 4 used 16 active and 16 quiet days. Average ap indices for the 16 quiet and 16 active days were 2.9 and 25.4 nT, respectively. Also, the average F10.7 cm flux for the quiet and active days were 72.1 and 73.0 sfu, that is, they are nearly the same. It can be seen in Figure 4 that there is more than 90 K difference between active and quiet times at all local times, which cannot be attributed to the 1 sfu increase in F10.7 cm flux. Note that the morning time (8-12 LT) temperature deviations are 23.5  2.5 K larger compared to afternoon time (12.5-16.5 LT). Such morning and afternoon temperature differences are seen primarily at mid-latitudes and the differences are very small or absent at low-latitudes (not shown here). This morning and afternoon difference is higher for the stronger events and they are nearly absent, even at mid-latitudes, for events having ap index less than 20 nT. This is the reason why ap  20 nT is chosen as a lower limit for the morning and afternoon difference investigation. Also, the latitude range is changed to mid-latitude only as these differences are prominent at mid-latitudes  and are very small or absent at low-latitudes for majority of the events. Note that the strongest geomagnetic event during the current observation period is of moderate strength that occurred during August 31 to September 1, 2019. In the future, when we expect a greater number of moderate and severe geomagnetic storms, a quantitative investigation of the relationship between strength of geomagnetic activity and corresponding morning to afternoon temperature difference at low and mid-latitudes could be performed.

Discussion
Heating due to solar Extreme-Ultra-Violet (EUV) absorption and cooling due to downward heat transport are the primary source and sink of energy in the daytime thermosphere. The diurnal tidal circulation in the thermosphere is mainly driven by in-situ differential heating due to solar EUV. As a result of this circulation there occur regions of convergence and divergence that produce vertical motions (Laskar et al., 2017), which are upward in the daytime and are downward in the majority of the night-sector at low-and mid-latitudes LASKAR ET AL.
10.1029/2021GL093905 6 of 10  Mayr et al., 1978). From numerical model simulations, Burns et al. (1995) showed that under quiet geomagnetic conditions the usual vertical winds during late-night to late-morning are downward and that they are upward in the afternoon and late-evening sector. During geomagnetically active events the usual thermospheric equator to pole circulation gets disturbed due to high-latitude energy deposition. Under such altered circulation, the morning time vertical winds at low-and mid-latitudes become increasingly downward due to pole to equator circulation thus creating greater compressional heating compared to quiet time. Whereas, in the afternoon and evening sectors, the storm time circulation makes the vertical winds weaker, resulting in less expansion of the thermosphere and thus less cooling. Due to this change in circulation, the storm-time temperatures at low-and mid-latitudes are higher than quiet time. This mechanism suggests that the pre-noon sector enhancement in thermospheric temperature would be greater than in the afternoon. Figure 8 of Burns et al. (1995) shows a numerical simulation result that provides a more detailed explanation of this mechanism. The DE-2 data used in Burns et al. (1995) were extremely limited and thus could not demonstrate the effect unambiguously. The numerical model simulation results of Burns et al. (1995) provide a plausible explanation of the GOLD observed morning and afternoon temperature difference as shown in Figure 4. Therefore, the GOLD results presented in Figure 4 provide a first experimental demonstration of this effect during geomagnetically active conditions. This has been possible due to the unprecedented local-time and latitude coverage of the GOLD mission. Further studies are warranted on the possible association of the morning and afternoon difference to the day-night difference of the magnetospheric energy input to the high-latitude thermosphere. In the future, a more detailed investigation of GOLD observations during stronger geomagnetic storms could be performed to quantify the temporal relationship between different phases of the storms and T.
LASKAR ET AL.