Monthly Climatologies of Zonal‐Mean and Tidal Winds in the Thermosphere as Observed by ICON/MIGHTI During April 2020–March 2022

Version 5 (v05) of the thermospheric wind data from the Michelson Interferometer for Global High‐resolution Thermospheric Imaging (MIGHTI) instrument on the Ionospheric Connection Explorer (ICON) mission has been recently released, which largely avoids local‐time dependent artificial baseline drifts that are found in previous versions of the ICON/MIGHTI wind data. This paper describes monthly climatologies of zonal‐mean winds and tides based on the v05 ICON/MIGHTI data under geomagnetically quiet conditions (Hp30 < 3o) during April 2020–March 2022. Green‐line winds in the lower thermosphere (90–110 km) and red‐line winds in the middle thermosphere (200–300 km) are analyzed, as these data cover both daytime and nighttime. The latitude and height structures of zonal‐mean winds and tides are presented for each month, and the results are compared with the widely used empirical model, Horizontal Wind Model 2014 (HWM14). The ICON/MIGHTI and HWM14 results are in general agreement, providing a validation of the v05 ICON/MIGHTI data. The agreement is especially good for the zonal‐mean winds. Amplitudes of lower thermospheric tides from ICON/MIGHTI tend to be larger than those from HWM14 as well as from an empirical model, Climatological Tidal Model of the Thermosphere (CTMT). This could be due to the influence of interannual variability of the tides. The amplitude structure of lower thermospheric tides in HWM14 does not match those from ICON/MIGHTI and CTMT in some months. Also, HWM14 underestimates the meridional‐wind amplitude of the migrating diurnal tide in the middle thermosphere. These results highlight the need for improved tidal representation in HWM14.

Thermospheric winds can have a significant impact on ionospheric dynamics (e.g., Rishbeth, 1998) and electrodynamics (e.g., Heelis, 2004), and thus are important for the accurate description of space weather.
There are several ways to observe thermospheric winds. For instance, wind velocities can be measured using an accelerometer onboard a low-Earth-orbit satellite. Past satellite missions like Dynamic Explorer 2 (DE2) (Spencer et al., 1982), CHAllenging Minisatellite Payload (CHAMP) (H. Liu et al., 2006;Sutton et al., 2007), and Gravity Field and Steady State Ocean Circulation Explorer (GOCE) (Doornbos et al., 2010; provided global in-situ observations of thermospheric winds. Wind velocities can also be measured with a sounding rocket, which can reach the thermosphere. For example, the chemical release technique (e.g., Larsen, 2002;Pfaff et al., 2020) uses measurements of trails of a chemical tracer released by a rocket to derive thermospheric wind velocities. Moreover, optical measurements of Doppler shifts in airglow emissions, such as the 557.7 nm O( 1 S) green line and the 630.0 nm O( 1 D) red line, have also been used to observe thermospheric wind velocities from ground stations (e.g., Makela et al., 2012;Meriwether, 2006;Shiokawa et al., 1999) as well as from satellites such as DE2 (Hays et al., 1981), Upper Atmosphere Research Satellite (UARS) (Hays et al., 1993) and Thermosphere-Ionosphere-Mesosphere Energetics and Dynamics (TIMED) . Ground-based meteor radars can be used to measure wind velocities in the mesosphere and lower thermosphere around 80-100 km (e.g., Chau et al., 2019;Hocking et al., 2001). Thermospheric wind velocities at E-region and F-region heights can also be estimated using incoherent scatter radar measurements of ionospheric parameters (e.g., Harper, 1977;Hysell et al., 2014;Salah & Holt, 1974).
Global empirical models of thermospheric winds have been developed based on the measurements obtained through the techniques mentioned above and others. The most commonly used empirical model is the Horizontal Wind Model (HWM) series (e.g., Drob et al., 2008Drob et al., , 2015Hedin et al., 1991;Hedin et al., 1996). HWM is constructed by fitting analytical functions to a large volume of historical data. It predicts the zonal and meridional components of the neutral wind velocity at a given location (latitude, longitude and altitude) and time (day of year and UT). The latest version is HWM14 (Drob et al., 2015), and since its release, the model has been widely used in the space physics community. The validation of HWM is a community effort. Thermospheric wind measurements are often compared against HWM for a validation of the observational data as well as for a performance evaluation of HWM (e.g., Englert et al., 2012;Jiang et al., 2018;Li et al., 2021;Okoh et al., 2022;Tang et al., 2021). The present study shows comparisons of HWM14 with thermospheric wind observations from the Ionospheric Connection Explorer (ICON) mission, which was launched on 11 October 2019 (Immel et al., 2018).
The Michelson Interferometer for Global High-Resolution Thermospheric Imaging (MIGHTI) instrument onboard ICON measures the horizontal wind velocity by observing Doppler shifts of the atomic oxygen airglow emissions (e.g., Englert et al., 2017;Harding et al., 2017). The green-line wind measurements extend from an altitude of 90-300 km during the daytime but to only 110 km at night, as the strength of the green-line emission varies considerably from day to night. The red-line wind data cover the height range approximately 160-300 km during day and 200-300 km at night. These wind data are useful not only for studying the neutral dynamics of the thermosphere (e.g., Cullens et al., 2020;Englert et al., 2017;Forbes et al., 2022;He et al., 2021;Triplett et al., 2023;Yiğit et al., 2022) but also for investigating atmosphere-ionosphere coupling processes, which can be realized by combining the ICON/MIGHTI wind data with ionospheric measurements made by ICON (e.g., England et al., 2021;Immel et al., 2021;Forbes et al., 2021;Park et al., 2021;Heelis et al., 2022; or by other missions (e.g., Aa et al., 2022;Gasperini et al., 2021;Gasperini et al., 2022;Harding et al., 2022;Le et al., 2022;Oberheide, 2022;Yamazaki et al., 2021;Yamazaki, Arras, et al., 2022).
The studies mentioned above used version 4 (v04) or an earlier version of the ICON/MIGHTI wind data. The v04 wind data, especially during the early period of the mission, showed reasonable agreement with other independent observations (e.g., Chen et al., 2022;Dhadly et al., 2021;Harding et al., 2021;Makela et al., 2021). However, later it became clear that the baseline of the v04 data has slowly drifted over time, leading to errors of 50-100 m/s for some cases in 2021. This issue was described in detail by Englert et al. (2023). The baseline drift was found to be dependent on the local time and height, which has made the reliable assessment of zonalmean winds and tides difficult. Version 5 (v05) of the ICON/MIGHTI wind data has been recently (in November 2022) released. A new calibration method for the so-called "zero wind" has been developed for v05, which uses a long-term (96 days) comparison of the ascending-and descending-orbit data to perform a self-calibration of the zero baseline, independent of external data or models . The present study evaluates, for the first time, zonal-mean winds and tides using the v05 ICON/MIGHTI wind data for the height ranges of 90-110 and 200-300 km, where wind measurements are made during both day and night. Monthly climatologies derived from the ICON/MIGHTI observations during April 2020-March 2022 are compared with HWM14 predictions.
It is noted that in atmospheric science, the term "climatology" is often used to represent the average state over many decades, so that it is robust against interannual variability. In this study, we describe the average state of zonal-mean winds and tides using all the available v05 data, which are referred to as "climatology." However, since the averaging period is relatively short (i.e., 2 years), it needs to be kept in mind that our climatologies could be contaminated by the interannual variability. This aspect will be further discussed later.

Method to Determine Zonal-Mean Winds and Tides
The v05 ICON/MIGHTI wind data (level 2.2, cardinal vector winds) during the 24-month period from April 2020 to March 2022 are analyzed to determine zonal-mean winds and tides. The estimated accuracy of the v05 wind data is generally 10-25 m/s . Only the data that are flagged as "Good" (Wind_Quality = 1 are used. This largely eliminates (a) the observations from the South Atlantic Anomaly where the retrieval of wind velocities is difficult due to increased radiation, (b) the observations with little airglow signal, and (c) the observations from the day-night terminators where mode changes of the instrument take place. Cullens et al. (2020), using synthetic data sampled along the ICON/MIGHTI measurement points, demonstrated that (c) does not have a large impact on the estimation of tidal amplitude.
We use only the measurements made during geomagnetically quiet periods. Our criterion for the geomagnetically quiet periods is Hp30 < 3o, where Hp30 is the geomagnetic activity index described by Yamazaki, Matzka, et al. (2022). Briefly, Hp30 is a planetary geomagnetic activity index, similar to Kp  but with a higher temporal resolution of 30 min in contrast to the 3-hourly Kp index. The higher temporal resolution has an advantage in accurately selecting quiet-time data. Hp30 is produced at the GeoForschungsZentrum (GFZ) Potsdam and distributed at their website: https://kp.gfz-potsdam.de/en/hp30-hp60.
The green-line winds are given at every ∼3 km for the height range 91-112 km, while the red-line winds are given at every ∼10 km for 203-301 km. At each height, the data were binned in hourly UT bins, in 5° latitude bins every 2.5° latitude from 10°S to 40°N, and in 15° longitude bins every 15° longitude. This was done separately for each month of the year (but without distinction of different years) and for the zonal and meridional components of the wind. The mean value and standard deviation were computed for each bin. The standard deviation is used, in a later step, to evaluate 1-σ uncertainties in zonal-mean winds and tides. The bin-mean values at given latitude and height were expressed as a function of UT (t in hours), longitude (λ in degrees), and month (M = 1, 2, …, 12) using the following analytical representation: Formula 1 takes into account zonal-mean winds, tides and stationary planetary waves, and their seasonal variations. n represents the tidal frequency. That is, n = 1, 2, 3, 4 correspond to the 24-hr (or diurnal) tide, 12-hr (or semidiurnal) tide, 8-hr (or terdiurnal) tide and 6-hr tide, respectively. The higher order tides are generally not as important in the thermosphere (e.g., Oberheide et al., 2011). |s| denotes the zonal wavenumber, and the sign of s indicates the direction of the zonal propagation of tides. That is, s > 0 and s < 0 correspond to eastward-and westward-propagating tides, respectively. The standard tidal nomenclature is used throughout this paper, such as DE3 and SW2, where the first letter indicates the period (i.e., "D" for diurnal and "S" for semidiurnal), the second letter represents the propagation direction (i.e., "E" for eastward and "W" for westward), and the last number is the zonal wavenumber |s|. Going back to formula 1, the zonal-mean winds are represented by the terms with n = 0 and s = 0, while stationary planetary waves are represented by the terms with n = 0 and |s| > 0. The seasonal variations of the zonal-mean winds, tides and stationary planetary waves are represented by m = 1, 2, 3, corresponding to the annual, semiannual and terannual cycles. The coefficients a nsm and b nsm were determined in such a way that the deviation of formula 1 from the binned values of the ICON/MIGHTI data will be the smallest in a least-squares sense.
As previous studies have discussed (e.g., Englert et al., 2023), it takes longer than a month (48 days) for ICON/ MIGHTI to achieve full local-time coverage at all latitudes. Despite that, our method described above allows to determine tides for each month for the following reasons. (a) Each monthly bin contains observations from two different years, thus covering approximately 60 days (b) The results for a given month is constrained not only by the data of that month but also by the data of other months, as we assume seasonal variation consisting of the annual, semiannual and terannual cycles.
The goodness-of-fit was evaluated using two statistical metrics. One is the correlation coefficient between the observations (X) and fit (Y): where Cov is the covariance. The other is the root-mean-square error: where N is the number of the observations. The latitude and height distributions of the correlation coefficient and root-mean-square error are presented in Figure 1. The correlation coefficient is generally higher in the middle thermosphere (r = 0.8-1.0, based on red-line winds) than in the lower thermosphere (r = 0.65-0.85, based on green-line winds). This is mainly due to the fact that the lower thermosphere is more strongly influenced by the waves that are not described by formula 1 such as acoustic waves, gravity waves, lunar tides, Kelvin waves, and Rossby waves (e.g., H.-L. Liu, 2016;Yiğit & Medvedev, 2015). These waves are generated in the lower layers of the atmosphere and propagate into the thermosphere. They can interact with tides and other waves to produce secondary waves, which makes the spatial temporal variability of the lower thermosphere rather complex (e.g., Chang et al., 2011;H.-L. Liu, 2014;Nystrom et al., 2018). The waves from the lower atmosphere get strongly dissipated before reaching the middle thermosphere (200-300 km), where the red-line measurements are made. The middle thermosphere is dominated by the diurnal tide that is locally generated by solar heating (Hagan et al., 2001), which can be represented well by formula 1.
In Figure 1, the latitude and height distributions of the correlation coefficient reflect the structure of tides. That is, the correlation coefficient tends to be small where the amplitude of the dominant tide is small. For instance, as will be shown later, the amplitude of the migrating diurnal tide in meridional wind in the middle thermosphere is small at low latitudes. Accordingly, the correlation coefficient is smaller at low latitudes than at middle latitudes.
Also in Figure 1, RMS is somewhat larger in the middle thermosphere (20-30 m/s) than in the lower thermosphere (15-25 m/s). This reflects generally larger wind velocities in the middle thermosphere. These RMS values are much smaller than those reported for HWM14 (40-80 m/s) by Drob et al. (2015). This is not surprising given that HWM14 involves more diverse sources of data from many different years with various degrees of accuracy.
For a given month M, formula 1 can be rewritten in the following form, which more explicitly represents the zonal-mean winds and waves: Here, is the zonal-mean wind velocity (in m/s), A ns and P ns are the amplitude (in m/s) and phase (in rad) of a tide, respectively. ′ and ′ are the amplitude and phase of a stationary planetary wave, respectively. 1-σ uncertainties in the zonal-mean winds, tides and stationary planetary waves were evaluated using the standard devia tion obtained during the binning procedure described earlier. A Monte Carlo method was used for this purpose. That is, random noise was generated for each bin based on the standard deviation, and the noise was superimposed on the corresponding mean value. Fitting of formula 1 was repeated for 250 Monte Carlo samples, and 1-σ uncertainties were computed for the zonal-mean wind velocity, and the amplitude and phase of tides and stationary planetary waves at each latitude and height. The derived 1-σ uncertainty in the zonal-mean wind velocity is typically 1.0-3.5 m/s for both green-line and red-line winds. The 1-σ uncertainties in the amplitude and phase of tides and stationary planetary waves are typically less than 4.5 m/s and 20°, respectively. These uncertainty values are appreciably smaller compared to the features discussed in this paper.
Zonal-mean winds, tides and stationary planetary waves were also evaluated using HWM14 for the purpose of comparison. Hourly values of the zonal and meridional wind velocities were derived from HWM14 for each month by running the model for the 15th day of the month without including disturbance winds . At each latitude and height, the zonal-mean wind velocity, and the amplitude and phase of tides and stationary planetary waves were determined by least squares fitting of formula 4, which can be directly compared with the ICON/MIGHTI results.

Results
First, we examine seasonal climatologies of zonal-mean winds. Figure 2 depicts the zonal-mean zonal and meridional winds in the lower thermosphere (91-110 km) as derived from the ICON/MIGHTI green-line measurements. Below ∼105 km, the zonal-mean zonal wind in the equatorial region (10°S-10°N) tends to be weakly westward throughout the year. An eastward jet can be seen at 30°N during the Northern Hemisphere (N.H.) summer. The reversal of the zonal-mean zonal wind is often seen around 105 km, which was also noted by Yiğit et al. (2022) based on the analysis of the v04 ICON/MIGHTI data. The zonal-mean meridional wind is generally weak with little seasonal variation. The corresponding results obtained from HWM14 are presented in Figure 3. HWM14 captures the salient features of the observed zonal-mean zonal and meridional winds well, including the wind reversal near 105 km in the zonal-mean zonal wind. Figure 4 shows the zonal-mean zonal and meridional winds in the middle thermosphere (203-300 km) as derived from the ICON/MIGHTI red-line measurements. An annual variation of the zonal-mean zonal wind is evident. That is, the zonal wind in the N.H. is largely eastward and westward during the local winter and summer, respectively. The seasonal variation of the zonal-mean meridional wind is also dominated by an annual cycle. That is, the meridional wind is primarily northward during the N.H. winter and southward during the N.H. summer. The seasonal transitions occur in March and September. The annual variations in the zonal-mean zonal and meridional winds are reproduced well by HWM14 as shown in Figure 5.
Next, we examine seasonal climatologies of tides and stationary planetary waves. Different wave components, as expressed by different combinations of (n, s), have varying degrees of significance in the thermosphere (e.g., Forbes et al., 2014;Truskowski et al., 2014). Figures 6 and 7 depict wave spectra for two representative altitudes. Figure 6 shows the amplitude of different wave components for the green-line winds over the equator at an altitude of 106 km. At this particular latitude and height, the eastward-propagating diurnal tide with zonal wavenumber 3 (DE3; n = 1, s = 3) dominates the tide in zonal wind, especially during July-November, and the migrating semidiurnal tide (SW2; n = 2, s = −2) dominates the tide in meridional wind, especially during April-September. Figure 7 is similar to Figure 6 but for the red-line winds at 30°N at an altitude of 273 km. In the middle thermosphere, the migrating diurnal tide (DW1; n = 1, s = −1) is by far dominant. Since DW1, SW2 and DE3 are found to be dominant within the latitudinal and altitudinal range of the ICON/MIGHTI wind measurement, we further analyze these specific tides. Figure 8 shows the amplitude and phase of DW1 in meridional wind in the lower thermosphere as derived from the ICON/MIGHTI green-line measurements (The corresponding results for the zonal wind can be found in Figure  S1 in Supporting Information S1.) The amplitude is largest at 15°-20°N and 95-97 km, and it shows a semiannual variation with equinoctial maxima of ∼60 m/s. The phase of DW1 tends to decrease with increasing height. This 'downward phase propagation' is a fundamental feature of upward-propagating tides (e.g., Forbes, 1995). The results suggest that DW1 in the lower thermosphere originate from lower layers of the atmosphere. The latitude-height pattern of the DW1 phase does not vary much with the season. DW1 derived from HWM14 is presented in Figure 9. The DW1 amplitude in HWM14 is largest at 15°-20°N, which is in agreement with the ICON/MIGHTI results. HWM14 also reproduces the semiannual variation in the DW1 amplitude. However, the DW1 amplitude in HWM14 is generally smaller than in ICON/MIGHTI, and the height structure does not agree well with the observations. The latitude and height structures of the DW1 phase are reproduced by HWM14 during equinoctial months. Figure 10 presents the SW2 amplitude and phase in meridional wind in the lower thermosphere as derived from the ICON/MIGHTI green-line measurements (The corresponding results for the zonal wind can be found in Figure S2 in Supporting Information S1.) The amplitude is relatively large over the equator (10°S-10°N) and at N.H. middle latitudes (>30°N). In the equatorial region, the amplitude grows with height, reaching 60 m/s at 110 km during August-September. The maximum amplitude probably occurs above 110 km. At middle latitudes, the amplitude peaks at 105 km. The downward phase propagation is seen at both equatorial and middle-latitude      Figure S1 in Supporting Information S1.  Figure S2 in Supporting Information S1.  Figure S3 in Supporting Information S1.
regions, indicating that the SW2 energy propagates upward at these heights. The corresponding results derived from HWM14 are shown in Figure 11. Again, the amplitude in HWM14 is generally smaller than in ICON/ MIGHTI, and the height structure does not agree well with the observations. Interestingly, there is remarkable agreement in the phase of SW2 in the lower thermosphere between the ICON/MIGHTI and HWM14 results. As one possible explanation, HWM14 might involve different wind measurements at different heights that are not uniformly calibrated. This could lead to the good agreement in phase but poor agreement in amplitude. Figure 12 shows the amplitude and phase of DE3 in zonal wind in the lower thermosphere as derived from the ICON/MIGHTI green-line measurements (The corresponding results for the meridional wind can be found in Figure S3 in Supporting Information S1.) DE3 is the largest non-migrating (i.e., non-sun-synchronous) tidal component found in the green-line data. The zonal-wind amplitude is largest over the equator at a height of 105-110 km. The maximum amplitude exceeds 30 m/s during July-October. The downward phase propagation is visible, indicating upward energy propagation of DE3. DE3 is nonexistent in HWM14, as the model does not take into account any non-migrating tide.
We now look at DW1 in the middle thermosphere. Figure 13 shows the amplitude and phase of DW1 in zonal wind in the middle thermosphere as derived from the ICON/MIGHTI red-line observations (The corresponding results for the migrating semidiurnal tide SW2 can be found in Figure S4 in Supporting Information S1.) It is noted that the scale range for the amplitude is different from those used for the green-line results (Figures 8,  10, and 12). The DW1 amplitude grows with height from ∼50 m/s at 200 km to ∼90 m/s at 300 km. It exceeds 100 m/s in some months. The phase does not vary with height, indicating that DW1 in the middle thermosphere is a vertically trapped (evanescent) tidal mode that is locally generated, rather than an upward-propagating mode from below. The corresponding results derived from HWM14 are presented in Figure 14. HWM14 reproduces the latitude and height structures of the amplitude and phase well. Figure 15 also shows the amplitude and phase of DW1 from the ICON/MIGHTI red-line measurements, but for the meridional wind (The corresponding results for the migrating semidiurnal tide SW2 can be found in Figure S5 in Supporting Information S1.) The amplitude is small over the equatorial region but can exceed 100 m/s at middle latitudes (>30°N) above 280 km. The phase depends strongly on latitude. The phase structure is well captured by HWM14 (Figure 16)

Discussion
We have presented seasonal climatologies of the zonal-mean winds and tides derived from the v05 ICON/MIGHTI data, and compared the results with those from HWM14. Here we compare the ICON/MIGHTI results with those presented in earlier work based on other observations and models. Also, we discuss physical mechanisms behind some of the features observed in the ICON/MIGHTI winds, referring to previous theoretical studies. Wang et al. (1997) created an empirical model of lower thermospheric winds (90-120 km) using the measurements from the wind imaging interferometer (WINDII; Shepherd et al., 1993) onboard UARS. They presented the zonal-mean zonal and meridional winds for different seasons, which can be compared with our ICON/MIGHTI results (Figures 2 and 4). S. P. Zhang et al. (2007) later analyzed an updated version of UARS/WINDII data and obtained similar results as Wang et al. (1997). The UARS/WINDII results showed a westward jet of 10-30 m/s over the equator at ∼100 km throughout the year. This is also seen in the ICON/MIGHTI results (Figure 2), as well as in HWM14 (Figure 3). It is noted that the UARS/WINDII data are already incorporated in HWM. The westward jet over the equator is considered to result from the westward momentum deposition by dissipating migrating (thus westward-propagating) tides (e.g., Jones Jr et al., 2014;Lieberman & Hays, 1994;Miyahara, 1978Miyahara, , 1981. Wang et al. (1997) and S. P. Zhang et al. (2007) noted that the equatorial westward jet is sandwiched by eastward jets centered around ±40° latitudes. The eastward jets were reported to be stronger in the summer hemisphere, with the magnitude of 30-40 m/s. The ICON/MIGHTI results (Figure 2) clearly capture the N.H. part of the eastward jets. The mechanism for the middle-latitude eastward jets is not well understood. The numerical work by Forbes et al. (1993) predicted that the equatorial westward jet induced by tidal dissipation is accompanied by eastward jets at higher latitudes. However, the eastward jets due to tidal dissipation are predicted to be much weaker than the westward jet, which agrees with neither ICON/MIGHTI nor UARS/WINDII observations. Besides tides, Figure 13. Amplitude (top 12 panels) and phase (bottom 12 panels) of the migrating diurnal tide (DW1) in zonal wind in the middle thermosphere as derived from the Version 5 Ionospheric Connection Explorer/Michelson Interferometer for Global High-resolution Thermospheric Imaging red-line data. For the amplitude, contour intervals are 10 m/s. The corresponding results for the migrating semidiurnal tide (SW2) can be found in Figure S4 in Supporting Information S1. Figure 15. Same as Figure 13 but in meridional wind. The corresponding results for the migrating semidiurnal tide (SW2) can be found in Figure S5 in Supporting Information S1. Figure 16. Same as Figure 14 but in meridional wind. Miyoshi and Fujiwara (2006) numerically demonstrated that the momentum deposition by eastward-propagating equatorial Kelvin waves also plays a significant role for the zonal-mean zonal wind in the equatorial lower thermosphere. More studies are required to determine the relative importance of different waves to explain the observed westward and eastward jets.
The zonal-mean meridional wind in the lower thermosphere as derived from the ICON/MIGHTI data is generally weak (Figure 2), which is consistent with the UARS/WINDII results presented by Wang et al. (1997) as well as HWM14 (Figure 3). S. P. Zhang et al. (2007) noted that the zonal-mean meridional wind sometimes shows a cell-like structure in the low latitude region, which is characterized by poleward winds on both sides of the equator at altitudes of 95-105 km and equatorward winds at 105-115 km. There is some hint of such a cell-like structure in the ICON/MIGHTI zonal-mean meridional wind (see, e.g., August-September), but it is not well resolved because of the small magnitude. The cell-like structure in the zonal-mean meridional wind in the lower thermosphere is sometimes found in numerical models and is considered to be driven by tidal dissipation (e.g., Forbes et al., 1993;Miyahara et al., 1993).
ICON/MIGHTI red-line measurements revealed seasonal climatologies of the zonal-mean zonal and meridional winds at 200-300 km (Figure 4). HWM14 reproduces the ICON/MIGHTI observations well for both the zonal and meridional components ( Figure 5). HWM14 at this height range is well constrained by UARS/WINDII red-line winds as well as observations by ground-based Fabry-Perot interferometers. The zonal-mean meridional wind in the middle thermosphere is directed from the summer to the winter hemisphere, which is not surprising given the higher temperature and pressure in the summer hemisphere. The seasonal transition in the meridional circulation occurs in March and September. Using a numerical model, Roble et al. (1977) showed that the seasonal transition of the zonal-mean circulation takes place within a few weeks of equinox. Such an abrupt seasonal transition is not fully resolved in our monthly analysis. The zonal-mean zonal wind in the middle thermosphere arises mainly from the correlation between diurnal variations of pressure gradient and ion drag (Dickinson et al., 1975. That is, the wind is weaker on the dayside than the nightside as the ion drag is larger on the dayside due to higher plasma concentration. Since the wind in the middle thermosphere undergoes a diurnal cycle due to day-night pressure differences, an imbalance between the daytime and nighttime winds leads to the zonal-mean winds. The three most dominant tidal components in the ICON/MIGHTI green-line winds are DW1, SW2, and DE3 (Figures 8, 10 and 12). This is as expected from previous studies on tides in the lower thermosphere (e.g., Forbes et al., 2008;Oberheide et al., 2011). DW1 and SW2 are sun-synchronous, while DE3 is non-sun-synchronous. In the lower thermosphere, they consist mainly of upward-propagating modes, which can be seen from their downward phase propagation. They are driven by radiative heating through insolation of H 2 O in the troposphere and O 3 in the stratosphere (e.g., Forbes, 1982a;Forbes, 1982b) as well as by latent heating in the troposphere (Hagan & Forbes, 2002;Hagan & Forbes, 2003;X. Zhang et al., 2010aX. Zhang et al., , 2010b. DW1 in meridional wind, as derived from the ICON/MIGHTI observations, shows an amplitude maximum at 15°-20°N at an altitude of 95-98 km (Figure 8), which are consistent with those from UARS/WINDII (McLandress et al., 1996;S. P. Zhang et al., 2007) and TIMED/TIDI (Wu et al., 2008a). The amplitude is larger during the equinoxes than the solstices, which is well known from previous studies (e.g., Burrage et al., 1995;Xu et al., 2009). McLandress (2002a, 2002b examined the mechanism for the semiannual variation of DW1 using a numerical model, and concluded that the change in the latitudinal shear of the zonal-mean zonal wind plays a leading role for the seasonal variation of DW1 in the lower thermosphere. HWM14 reproduces the semiannual variation of DW1 (Figure 9), but the overall amplitude of DW1 tends to be smaller than in ICON/MIGHTI. SW2 in meridional wind, as derived from the ICON/MIGHTI observations, is relatively strong during May-September (Figure 10), which is consistent with the UARS/WINDII observations (S. P. Zhang et al., 2007). HWM14 reproduces the seasonal variation of SW2 but with somewhat smaller amplitude ( Figure 11). The mechanism for the seasonal variation of SW2 is not well established. DE3 in the lower thermosphere has characteristics of a Kelvin wave (e.g., Forbes et al., 2003). In classical theory, a Kelvin wave travels eastward, and its zonal wind component has a Gaussian-shaped latitudinal profile with maximum amplitude over the equator (e.g., Forbes, 2000). The latitude and height structures of DE3 and its seasonal variation in the ICON/MIGHTI green-line zonal wind ( Figure 12) are consistent with those from the UARS/WINDII  and TIMED/TIDI observations (Oberheide et al., 2006;Wu et al., 2008b). As the zonal wind amplitude of DE3 reaches its maximum in the equatorial dynamo region at 105-110 km, it has a significant impact on the equatorial zonal electric field and current (e.g., England et al., 2006;Fejer et al., 2008) as well as on the F-region plasma concentration (e.g., Immel et al., 2006;Lin et al., 2007). Despite the importance of DE3 in low-latitude ionosphere-thermosphere coupling, it is not included in HWM14 like other non-migrating tides.
For further validation of lower thermospheric tides, we have compared our results with those from an empirical model, Climatological Tidal Model of the Thermosphere (CTMT; Oberheide et al., 2011). CTMT is based on fitting of Hough Mode Extensions (e.g., Forbes & Hagan, 1982) to the temperature and wind observations by the TIMED satellite during 2002-2008. Tidal amplitudes and phases from CTMT can be compared with those from ICON/MIGHTI green-line measurements in the lower thermosphere. Since CTMT does not take into account tidal forcing due to in-situ solar heating in the middle and upper thermosphere, CTMT tides are not comparable with those from ICON/MIGHTI red-line measurements. Figure 17 shows the amplitude and phase of DW1 in meridional wind in the lower thermosphere as derived using CTMT, which can be compared with the ICON/ MIGHTI green-line results in Figure 8. The CTMT results for SW2 and DE3 can be found in Figures S6 and S7 in Supporting Information S1, which are comparable with the ICON/MIGHTI results in Figures 10 and 12. The comparison between the ICON/MIGHTI and CTMT results reveal that they are generally in good agreement, except that the tidal amplitudes in ICON/MIGHTI tend to be larger than those in CTMT. For instance, the maximum amplitude of DW1 in meridional wind during March is about 60 m/s in ICON/MIGHTI ( Figure 8) while it is about 40 m/s in CTMT ( Figure 17). One explanation is the interannual variability of tides. For example, previous studies reported that the amplitude of DW1 in the lower thermosphere can change by a few tens of m/s from 1 year to the next (e.g., Burrage et al., 1995;Hagan et al., 1999). The interannual variability of DW1 (or tides in general) is not well understood, although some sources of the variability have been identified, such as the quasi-biennial oscillation (e.g., Xu et al., 2009) and the El Niño-Southern Oscillation (e.g., Pedatella & Liu, 2012). It is possible that our tidal climatologies are contaminated by the influence of tidal interannual variability due to the short averaging period (i.e., 2 years). Another explanation for larger tidal amplitudes in ICON/ MIGHTI is the underestimation of tidal amplitudes in CTMT. Further studies are needed to identify the cause of the difference in tidal amplitudes between ICON/MIGHTI and CTMT.
DW1 in the middle thermosphere (Figures 13 and 15) is predominantly a vertically trapped tide that is excited by in-situ solar heating (e.g., Forbes, 1982b;Hagan et al., 2001). This contrasts with DW1 in the lower thermosphere (Figure 8), which is primarily upward-propagating. The latitude and height structures of DW1 in the middle thermosphere are not well documented, particularly those based on observations. The simulation results by Hagan et al. (2001) showed that (a) the amplitude of DW1 at 200-300 km grows with height at all latitudes, (b) both zonal and meridional wind amplitudes are largest at high latitudes, (c) the meridional wind amplitude is vanishingly small over the equator but it increases with latitude, (d) the zonal wind amplitude does not depend strongly on latitude over the middle-and low-latitude regions, (e) both zonal and meridional wind phases do not depend strongly on height, (f) the zonal wind phase does not vary strongly with latitude, and (g) the meridional wind phase also does not vary strongly with latitude except that the phase reversal occurs at the equator. The ICON/MIGHTI results (Figures 13 and 15) are consistent with these numerical predictions.
Some previous studies have addressed a potential impact of the solar flux, mainly at the wavelengths of extreme ultraviolet (EUV), on neutral winds in the middle and upper thermosphere (e.g., Hedin et al., 1994). The ICON/ MIGHTI observations examined in this paper are obtained during the period April 2020-March 2022. The mean value of the F 10.7 index (Tapping, 2013), which is often used as a proxy of the EUV flux, was 82.8 sfu (1 sfu = 10 −22 W⋅m −2 ⋅Hz −1 ), with the minimum and maximum monthly values of 69.2 sft in May 2020 and 117.8 sfu in March 2022, respectively. We have ignored possible variations in the wind velocities associated with the change in the solar flux, as the ICON/MIGHTI data used in this study are not sufficient for evaluating the solar activity effect on the thermospheric winds. HWM14 also does not take into account the dependence of wind velocities on solar activity. Hedin et al. (1994) reported that although the solar flux significantly influences the temperature, the zonal-mean winds in the middle thermosphere do not strongly depend on solar activity. Hagan et al. (2001) noted in their simulation results that the solar flux has a marked effect on the temperature amplitude of DW1 in the middle thermosphere but not on the wind amplitudes. H. Liu et al. (2006) examined the effect of the solar flux on thermospheric winds (∼400 km) using the CHAMP accelerometer data. They found that the solar flux effect can be significant depending on the season and local time. The solar flux effect on tides is generally small below about 130 km, where tidal waves are mainly of lower atmospheric origin (Dhadly et al., 2018;Oberheide et al., 2009). More observational studies are required to establish the solar activity dependence of zonal-mean winds and tides in the thermosphere.

Conclusions
Monthly climatologies of quiet-time zonal-mean winds and tides are derived using the recently released v05 of the ICON/MIGHTI thermospheric wind measurements during April 2020-March 2022 at the altitude ranges of 90-110 and 200-300 km. Earlier versions of the ICON/MIGHTI wind data suffered from artificial baseline drifts that depend on local time. Thus, it was previously difficult to obtain reliable climatological estimates of zonal-mean winds and tides. The v05 data avoids this issue by the use of a renewed baseline calibration technique .
The ICON/MIGHTI results are compared with those from the latest version of HWM (i.e., HWM14) as well as previous studies. Salient features of zonal-mean winds and tides in the lower and middle thermosphere are in general agreement between ICON/MIGHTI and HWM14, including latitude and height structures and their seasonal variations. This provides a validation of the v05 ICON/MIGHTI data. HWM14 reproduces the zonalmean zonal and meridional winds well in both the lower and middle thermosphere. There are some discrepancies especially in tidal amplitudes. For example, latitude and height structures of DW1 amplitude in the lower thermosphere in HWM14 do not agree well with those observed by ICON/MIGHTI or those derived from an empirical tidal model CTMT. Also, HWM14 does not include non-migrating tides such as DE3, which is especially important in the equatorial lower thermosphere. In the middle thermosphere, HWM14 underestimates the meridional-wind amplitude of DW1 especially during the N.H. winter. A more accurate description of tides in the thermosphere is key to the future improvement of HWM, which can benefit from the inclusion of ICON/ MIGHTI winds.
Structures and seasonal variations of lower thermospheric tides in ICON/MIGHTI and CTMT are in good agreement. However, tidal amplitudes are often larger in ICON/MIGHTI than CTMT. One possibility is the influence of the interannual variability of tides, which is difficult to evaluate and remove from the 2-year observations used in this study. Another possibility is that CTMT underestimates tidal amplitudes in the lower thermosphere. Identifying the cause of the discrepancy requires further studies.

Data Availability Statement
The ICON/MIGHTI Level 2.2 product Cardinal Vector Winds (Version 5) is accessible from the ICON website https://icon.ssl.berkeley.edu/Data. The Hpo indices including Hp30 used in this study are available at the GFZ website https://kp.gfz-potsdam.de/en/hp30-hp60/data. See also data publication . The monthly F10.7 index is available at the website of the Canadian Space Weather Forecast Centre https://spaceweather.gc.ca/forecast-prevision/solar-solaire/solarflux/sx-5-mavg-en.php. The CTMT data used in this study are available from the website https://globaldynamics.sites.clemson.edu/articles/ctmt.html. See also data publication (Oberheide, 2011).