Thermospheric Responses to the 3 and 4 November 2021 Geomagnetic Storm During the Main and Recovery Phases as Observed by NASA's GOLD and ICON Missions

Leveraging observations by two NASA missions—GOLD (Global‐scale Observations of the Limb and Disk) and ICON (Ionospheric Connection Explorer), we investigate concurrent responses of thermospheric composition, temperatures, and neutral winds to the geomagnetic storm on 3–4 November 2021, as well as their interplay at low and middle latitudes. The synergetic observations reveal remarkable depletions up to 60%–70% in GOLD O/N2, along with large enhancements in GOLD temperatures poleward of 30° in the middle thermosphere. Meridional winds from ICON observations are altered by ∼100 m/s equatorward of 25°N latitude and at 250 km, characterized by a reversal of prevailing northward winds to geomagnetic storm‐driven southward winds. This study fills a need, after a decade‐long gap, for observing concurrent and co‐located responses of composition, temperatures, and neutral winds in the thermosphere to geomagnetic storms.

Supporting Information may be found in the online version of this article.

Citation:
Gan, Q., Eastes, R. W., Wu, Y.-J., Qian, L., Cai, X., Wang, W., et al. (2024).Thermospheric responses to the 3 and 4 November 2021 geomagnetic storm during the main and recovery phases as observed by NASA's GOLD and ICON missions.Geophysical Research Letters, 51, e2023GL106529.https://doi.org/10.1029/2023GL106529Liu & Luhr, 2005;Sutton et al., 2005;Zhang et al., 2014Zhang et al., , 2022)).Part of the reason for that is the lack of concurrent observations of crucial state variables in the thermosphere.Previous satellite missions have been largely devoted to measuring thermosphere neutral density or composition, such as O/N 2 (the column density ratio of atomic oxygen to molecular nitrogen) from GUVI/TIMED (Global Ultraviolet Imager/Thermosphere, Ionosphere, Mesosphere Energetics and Dynamics), which is a good tracer of thermospheric dynamics (Evens et al., 1995;Strickland et al., 1995;Zhang et al., 2014).Relative to composition, little is known about how thermospheric temperatures respond to a geomagnetic storm.This further hinders the understanding of changes in neutral density because temperature is a key parameter that governs compositional scale height.Moreover, global observations made by low-Earth-orbiting (LEO) satellites are subject to tangled spatial and temporal variations, which lead to ambiguity in distinguishing time-dependent variations from spatial-dependent variations or vice versa.Of most importance was the lack of global neutral wind measurements over a wide altitude range in the thermosphere.
Fortunately, such data are now available, thanks to, recent NASA missions-ICON (Ionosphere Connection Explorer, Immel et al., 2018) and GOLD (Global-scale Observations of the Limb and Disk, Eastes et al., 2017).One of the key state parameters that ICON measures is wind profiles spanning 90-300 km at low and middle latitudes (Englert et al., 2023;Harding et al., 2021;Makela et al., 2021), and GOLD provides unparalleled synoptic observations of disk O/N 2 and temperatures (Eastes et al., 2020).The two missions have been revolutionizing the understanding of the ionosphere-thermosphere dynamics during solar and geomagnetically active and quiet times (e.g., Cai et al., 2023;Eastes et al., 2020;England et al., 2020;Gan et al., 2020aGan et al., , 2020bGan et al., , 2023aGan et al., , 2023b;;Goel et al., 2023;Immel et al., 2023;Laskar et al., 2022;McGinness et al., 2023;Oberheide et al., 2020).The current investigation is motivated by a previous study- Gan et al. (2020a)-that reported the first observational changes in atomic oxygen (OI 135.6 nm brightness was used as a proxy) from the GOLD mission during the 5 November 2018 CME (Coronal Mass Ejection) storm.Gan et al. (2020a) also conducted a data-model comparison, and they found that the simulation from a suite of models (TIE-GCM and GLOW) can reproduce the observed changes in OI brightness reasonably well.Of special noted is the simulation revealed a close connection between thermospheric composition and neutral winds during the main and recovery phases of geomagnetic storms.Specifically, the TIE-GCM simulates pronounced upwelling associated with the divergence of horizontal winds due to Joule heating, which transports nitrogen-rich molecules from lower to higher altitudes and substantially depletes O/ N 2 at high latitudes.Subsequently, significant equatorward winds at higher latitudes are initiated by equatorward pressure gradients due to Joule heating.The equatorward winds further corotate to the dayside and rapidly expand the reduced O/N 2 toward middle latitudes.Meanwhile, downwelling due to converged horizonal winds can enhance the O/N 2 at low latitudes.We refer readers to the paper for details.However, the model prediction had not been well tested by observations as global wind measurements were not available.To test current and model-based hypotheses, we investigate another geomagnetic storm that occurred on 3 November 2021 when both GOLD and ICON were operational.Interestingly, the previous and current geomagnetic storms occurred at the same time of year, allowing for a comparative study.
To achieve this goal, we synergistically employ observations from NASA's GOLD and ICON missions, including synoptic observations of daytime disk O/N 2 and temperatures, and asynoptic observations of neutral winds of 90-300 km at low and middle latitudes.This allows one to characterize coincident storm-driven changes in thermospheric composition, temperatures, and neutral wind profiles during geomagnetic storms.The manuscript is organized as follows.Section 2 overviews data products used in this study.Section 3 presents the main results and relevant discussion.Conclusions are given in the last section.

Data and Methodology
Thermosphere O/N 2 and temperature data from GOLD and neutral winds in the altitude range of 90-300 km are analyzed for the period of interest.Since GOLD and ICON provide distinct observational geometries to measure thermospheric parameters from space, we apply two approaches to quantify storm-driven changes in O/N 2 , temperatures, and neutral winds.A key point is that GOLD scans the Earth at the same locations and Universal Times (UTs) on a daily basis, and one can directly differentiate the storm day and non-storm day scans to obtain changes in disk O/N 2 and temperatures, as illustrated by Gan et al. (2020a).As for ICON neutral winds, we first use a 45-day window-centered on the storm day -to calculate the baseline prevailing wind at each given location, which is a 45-day average plus various tides (diurnal, semidiurnal, and higher-order harmonics).The wind deviations on the storm day, relative to the baseline wind, are a measure of storm-driven changes.Utilizing a 45-day window to compute the baseline prevailing winds effectively eliminates day-to-day variability due to the slow orbit precession (Cullens et al., 2020;Gan et al., 2014).It is worth noting that ICON's green line neutral wind measurements above 105 km are limited to daytime, and fully resolving diurnal tides is not feasible.Given that the core window for characterizing storm-induced changes is less than a week, one can assume the diurnal tides do not change significantly, and the wind deviations derived by subtracting the baseline reference-a 45-day average plus tides with a period of 12-hr or shorter-are primarily associated with the geomagnetic storm.

Results and Discussion
Figure 1 illustrates the temporal evolution of interplanetary magnetic field (IMF) Bz, solar wind velocity, and geomagnetic activity indices Kp and Dst from October 29 to 8 November 2021.Significantly disturbed Bz (Figure 1a) was seen around 22 UT of November 3rd, and within 8 hr, it dropped by 15 nT.IMF Bz remained southward until mid-day on November 4th.Meanwhile, the solar wind velocity (Figure 1b) was drastically accelerated from ∼450 to 700 km/s.Such coincident features in IMF Bz and solar wind velocity imply the arrival of interplanetary shocks at Earth.Consequently, the Kp index (Figure 1c) started to escalate around 17 UT of November 3rd and ultimately reached 8 minus (designating a G3 geomagnetic storm per NOAA/SWPC's definition) before mid-day of November 4th.The geomagnetic storm subsequently transitioned into the recovery phase, as depicted by Dst (Figure 1d), during which Bz became much less perturbed and switched from southward to northward.The scope of the current study is focused on how the thermosphere evolves during the storm main and recovery phases.
GOLD's disk O/N 2 scans at 10:30, 12:30, 14:30 UT are shown in Figure 2, for pre-storm (upper panels) and storm (lower panels) days.Overplotted vectors combine ICON zonal and meridional winds at 250 km for GOLD and ICON coincidences; that is, the ICON wind observations are selected within a 2-hr window (1-hr before and after GOLD completes a full disk scan).Quiet-time GOLD O/N 2 (upper panels) exhibits a seasonally dependent patten; that is an overall higher O/N 2 in the northern hemisphere (NH) than in the southern hemisphere (SH) due to the summer-to-winter interhemispheric circulation, consistent with the reports by Gan et al. (2023b) and Qian et al. (2022).A localized peak in O/N 2 is noticeable at 45-60° latitude of both hemispheres on November 3rd.We hypothesize that these two peaks are attributable to some prevalent tidal components during this time of year.Another possible candidate is a localized downwelling circulation.During the storm times of November 4th (bottom panels), the O/N 2 poleward of 30° in both hemispheres depleted substantially, whereas significant enhancements were observed at low latitudes.Such a typical storm-driven pattern is caused by Joule heating (upwelling), which transports molecular-rich air parcels from lower to higher altitudes, reducing O/N 2 at middle and high latitudes.Conversely, the downwelling enhances low-latitude O/N 2 .
Meanwhile, ICON observed that northward winds are prevailing within GOLD's field-of-view under the geomagnetically quiet condition, aligned with the well-known summer-to-winter interhemispheric circulation, while wind velocities vary with location.On November 4th (bottom panels), prominent southward winds of up to ∼250 m/s were seen toward the west coast of North America at 12.5 and 14.5 UT (bottom middle and right).This marks a reversal of meridional winds from season-dependent northward winds to storm-driven southward winds in the NH, indicating that the geomagnetic storm drives substantial equatorward winds that overwhelms the summer-to-winter interhemispheric circulation.However, horizonal winds are not changed very much over the Atlantic and the west coast of Africa.The high O/N 2 at ∼10°S should be ignored as it is caused by a flat-field correction issue, regardless of storm and non-storm days (see details in Correira et al., 2021).This "bump" would not introduce significant uncertainties in quantifying and interpretating storm-induced changes because it is largely eliminated in differentiations (see examples in Gan et al., 2020a).
The main goal of this study is to quantify the concurrent changes in O/N 2 , temperatures, and neutral winds.As exhibited in Figure 3, we subtract the observations of the quiet-time reference day (October 29) from the disturbed day to derive the storm-induced changes in disk O/N 2 and temperatures.The largest O/N 2 depletion, up to 55%-60%, occurred in the middle of the north Atlantic at 10.5 UT (upper left) and shifted toward the American continent between 12.5-14.5UT (upper middle and right) on November 4th.The SH exhibited a slightly weaker O/N 2 depletion of 35%-40% poleward of 45°S.A prominent hemispheric asymmetry is seen in the depletion regions, which could be linked to the hemispheric asymmetry of the magnetic poles.At low-latitude, an O/ N 2 enhancement of 30%-40% is evident across GOLD's entire field-of-view at all three UTs, and its magnitude varies significantly with longitude.The distribution of low-latitude O/N 2 is quite complex since both high-latitude forcing and waves arising from the lower atmosphere can play important roles.Disk temperatures (bottom panels) are greatly increased at higher latitudes and correspond nicely with O/N 2 depletions in both hemispheres during the main phase of the geomagnetic storm.For instance, the temperature enhancement of 60%-70% co-locates with the largest O/N 2 depletion in the NH.Low latitudes demonstrate a more complicated distribution of O/ N 2 and temperature changes, which are highly longitudinal dependent.Fully understanding the intricate spatial distribution of the changes in O/N 2 and temperatures is beyond the scope of the current paper.It is noteworthy that the O + radiative recombination (RR) in the equatorial ionization anomaly produces 135.6 nm emission that could result in an enhanced O/N 2 at low latitudes (Kil et al., 2013).Such an ionospheric contamination has not been removed from the current GOLD O/N 2 data set.The previous study by Correira et al. (2021) revealed that the RR contamination potentially leads to a 10%-20% bias in the measured O/N 2 under the high geomagnetic activity conditions.However, the RR is not adequate to explain the observed enhancement in O/N 2 at low latitudes.We thus suggest that the observed O/N 2 changes at low latitudes are mainly attributed to the storm-driven dynamics after accounting for the existing ionospheric contamination.
Horizontal wind deviations (green vectors) are derived by subtracting the baseline prevailing winds (blue vectors, the 45-day average wind plus tidal winds) from total winds (raw data).Coincident wind measurements mostly occurred over the Atlantic at 10.5 UT, and over North America and South Africa at 12.5 and 14.5 UTs.Large southward deviations of ∼250 m/s stand out toward the west coast of North America, which are intimately aligned with the greatest O/N 2 depletion and temperature enhancement, giving support of the model prediction.As noted by Gan et al. (2020a), prominent southward winds are driven by equatorward temperature gradients due to Joule heating and ion drag at higher latitudes.From the middle Atlantic through the African sector, wind deviations remain northward, possibly because of the northward pressure gradient from higher to lower latitudes in the SH.The equatorward deviations are significantly greater over the American sector than over the African sector, suggesting a strong hemispheric asymmetry in Joule heating.Combining GOLD and ICON observations provide insights into the interplay among thermospheric composition, temperatures, and neutral winds during this CME geomagnetic storm.For example, prevalent equatorward winds due to the geomagnetic storm are aligned with the latitudinal gradient of O/N 2 and temperatures, as is expected for pressure-gradient driven winds at this altitude.
To elucidate how these state variables evolve across all the stages of the geomagnetic storm, coincident observations of GOLD and ICON over the American sector during the course of 2-6 November 2021 are further examined, as displayed in Figure 4. Locations of concurrent and co-located measurements are denoted in Figures 4a and 4b (yellow), which are essentially within 0-25°N and 50°-120°W (see examples in Figures 2 and 3).Disk O/N 2 (black in Figure 4a) and temperatures (black in Figure 4b) were enhanced by ∼0.2 and 150 K at the coincidences on November 4 and then ramped down to the pre-storm state within 24 hr.In other words, both O/N 2 and temperatures seem to recover promptly at those locations.The largest southward deviation (green in Figure 4c) attained ∼250 m/s on November 4, consistent with what is seen in Figure 3; otherwise, day-to-day variability in meridional winds is not substantial.In the meantime, zonal wind deviations (green in Figure 4d) on November 4 are much more westward compared to those during other days, and this could be the result of the Coriolis force that drives equatorward meridional winds.It is noteworthy that the temperature (black in Figure 4b) is lower on November 5 than on November 3 at the given locations.This may be associated with the "overcooling" phenomenon that was first seen in the post-storm neutral density during the recovery phase of a geomagnetic storm (Lei et al., 2012).The post-storm neutral density in that study was over reduced by 23%-36% compared to the pre-storm state, and the estimated overcooling in temperature was 70-110 K. GOLD Tdisk observations appear to be consistent with the previous studies.
Next, ICON wind observations at 130-300 km are further examined.This altitude range is of special interest because it spans the altitudes that GOLD makes disk scans of O/N 2 (∼130 km) and temperatures (140-180 km) to the uppermost altitude that ICON measures neutral winds with a reasonable signal-to-noise ratio.We utilize the meridional wind profiles during the November 4 over the American sector, where the exceptional southward deviations are observed.Data analyses are performed individually for 200-300 km and 130-192 km segments, because wind measurements below 200 km are limited to daytime.As illustrated in Figure 5a, the prevailing meridional winds (solid lines), including the 45-day average and tides, vary between −20 m/s (negative is southward and vice versa) and 40 m/s from 1 day to another.An overall prevailing wind is northward due to the summer-to-winter circulation.Day-to-day variability in the prevailing wind profiles is attributable to atmospheric tides on top of the season-dependent circulation (Oberheide et al., 2011).On November 4, southward deviations (the dotted line) were predominant, reaching 90-110 m/s, and decreased drastically above 265 km (although uncertainties are too large for meaningful interpretation).Note that the wind profiles shown in Figure 5 are averages at GOLD and ICON coincidences within 0-25°N and 50°-120°W.It is therefore not unexpected that the absolute wind change is a factor of 2 smaller than that at a single location (e.g., Figures 3 and 4).
Prevailing winds remain northward between 130 and 200 km (Figure 5b), and day-to-day variability due to tides is coherent with that above 200 km.Southward wind deviations are also prominent, despite the magnitude being decreased by a factor of 2.5 (from ∼130 m/s at 190 km to ∼50 m/s at 130 km) with decreasing altitude.This is probably because air density increases exponentially from higher to lower altitudes.It seems that the lack of nighttime observations would not lead to a significant uncertainty in assessing wind changes between 130 and 190 km given that geomagnetic storms are such a dominant driver.Moreover, aliasing due to diurnal tides would not introduce significant uncertainties because the core time window in which we look for storm-induced changes is less than a week.It is noted that the analysis of ICON observations performed in this study provides a perspective to quantify the degree to which responses of thermospheric neutral winds to a geomagnetic storm versus atmospheric tidal forcing over a broad altitude range.

Conclusions
This study provides a quantitative assessment of the responses of thermospheric composition, temperatures, and neutral winds to a geomagnetic storm from an observational standpoint.Such a goal had not yet been achieved until the launch of two NASA satellite missions: GOLD and ICON.Taking full advantage of two distinct observational samplings-synoptic view by GOLD in geosynchronous orbit and asynoptic view by ICON in LEO, we quantified geomagnetic storm-induced changes in O/N 2 , temperatures, and neutral winds at GOLD and ICON coincidences in the low-and middle-latitude thermosphere.Large O/N 2 depletions and temperature enhancements, up to 60%-70%, were seen over the west coast of North America on November 4. The spatial distribution of the O/N 2 depletion and temperature enhancement correspond nicely at higher latitudes.Low latitudes display a more complicated spatial distribution of O/N 2 and temperature, which is highly longitudinal dependent.Furthermore, O/N 2 and meridional winds are closely connected at GOLD and ICON coincidences, implying that changes of O/ N 2 are primarily driven by changes of meridional winds, as predicted by model simulations (Gan et al., 2020a).Prominent southward wind deviations of ∼250 m/s are co-located with the greatest gradient in O/N 2 at the west coast of North America.This suggests that Joule heating produces a large equatorward pressure gradient and thus southward winds, which overcomes season-dependent summer-to-winter circulation.Storm-driven southward deviations are coherent across the altitudes ICON observes, where the magnitude of the southward winds increases rapidly with increasing altitude until 200 km and tends to be a constant at 200-260 km.

•
Remarkable depletions in O/N 2 and enhancements in temperature, up to 60%-70%, are seen by Global-scale Observations of the Limb and Disk during the geomagnetic storm • Averaged southward wind deviations of ∼100 m/s are observed by Ionospheric Connection Explorer, coincident with the largest gradient in O/N 2 depletions • Both O/N 2 and temperatures recover rapidly from the disturbed states to pre-storm states Supporting Information:

Figure 3 .
Figure 3.The same as above but are percentage changes in O/N 2 (top) and temperatures (bottom) on day 308 (November 4), relative to the baseline reference day 302 (October 29).Blue vectors are the baseline prevailing winds that include a 45-day average plus tidal winds at 250 km.Green vectors are wind deviations relative to the prevailing winds.

Figure 4 .
Figure 4. Temporal evolution of GOLD disk (a) O/N 2 , (b) temperatures, and ICON (c) meridional and (d) zonal winds at GOLD and ICON coincidence during November 2-6, 2021.Black squares (a) and (b) are GOLD raw data, and orange circles denote latitudes (a) and longitudes (b) of GOLD and ICON coincidence.Yellow filled represents the prevailing baseline winds of the meridional (c) and zonal (d) wind components; green circles represent wind deviations relative to the baseline winds.

Figure 5 .
Figure 5. Meridional wind deviations (orange dotted) of 200-300 km (a) and 130-190 km (b) averaged over 0-25°N and 50°-120°W at GOLD and ICON coincidences on day 308.Top is ICON red line observations, and bottom is ICON green line observations.Solid line is season-dependent prevailing winds plus tidal winds, which vary from one day to another due to orbit (local time) precession.Error bars overplotted are measurement errors (on the dotted line) and 1-sigma tidal fitting uncertainties (on the solid line).