The quality and distribution in time and space of available atmospheric observations are crucial for the accuracy of semi-empirical thermosphere models. However, datasets can be inconsistent, and their qualities and resolutions are often unequal. The main thermospheric density datasets of this century are briefly described and then compared to each other when possible in order to quantify differences. Total mass densities used in the comparisons include all high-resolution CHAMP, GRACE and GOCE data, Swarm A, daily-mean Stella, global daily mean TLE densities, and the SET HASDM density database. The temperature data from TIMED-SABER are also reviewed. The recently updated daily-mean TLE densities (TLE2021) are 2–10 % smaller on average than the previous version (TLE2015). The differences are not constant offsets per altitude level, but fluctuations of up to 5 % are present. Compared to HASDM densities for 6 altitudes from 250 to 675 km, TLE2021 is 15–20 % smaller at 250 km, and then the difference diminishes with altitude to reach the same average value at 575 km. These mean differences also fluctuate by a few percent on time scales of months, to 10 % over half a solar cycle at 575 km. The TLE2021 and HASDM densities are larger than the accelerometer-inferred CHAMP, GRACE and GOCE densities and average offsets are 10–15 % and 10–20 %, respectively. The comparison to Swarm-A and Swarm-B showed mean offsets of 10 % and less, with significant positive trends seen in the comparison with HASDM. Finally, largest differences are found for Stella and HASDM at 800 km, up to 45 % with strong semiannual variations. This study clearly shows that the available density data cannot be simply assimilated or combined without first accurately calibrating the data. The HASDM database is a valuable asset due to its considerable coverage in space and time, but its uncertainty and true resolution are not well understood and are still being evaluated. Data compatibility requires employing physically accurate and harmonized aerodynamic force models in the density derivation procedure, which is presently not achieved. The accuracy of the procedure, independent of the quality of the instrument (GNSS receiver, ground-based orbit determination, or accelerometer), inevitably decreases with altitude due to weakening of the drag signal to noise ratio. The TIMED-SABER instrument provides measurements of pressure and temperature in the lower thermosphere. SABER temperature uncertainty is well-known. The SABER dataset now exceeds twenty years and has been continuously operating that entire time. It was ingested in NRLMSIS 2.0 and comparisons show the much-improved fit in comparison with NRLMSISE-00. The lower thermosphere temperatures significantly modify density at higher altitudes, and its measurement is essential for modeling and assessment.

This paper presents measurements of the amplitudes and timings of the combined, annual, and semiannual variations of thermospheric neutral density, and a comparison of these density variations with measurements of the infrared emissions from carbon dioxide and nitric oxide in the thermosphere. The density values were obtained from measurements of the atmospheric drag experienced by the Challenging Minisatellite Payload, Gravity Recovery and Climate Experiment A, Gravity field and Ocean Circulation Explorer, and three Swarm satellites, while the optical emissions were measured with the Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) instrument on the Thermosphere Ionosphere Mesosphere Energetics and Dynamics satellite. These data span a time period of 16 years. A database containing global average densities that were derived from the orbits of about 5,000 objects (Emmert, 2009, https://doi.org/10.1029/2009JA014102, 2015b, https://doi.org/10.1002/2015JA021047) was employed for calibrating these density data. A comparison with the NRLMSISE-00 model was used to derive measurements of how much the density changes over time due to these seasonal variations. It is found that the seasonal density oscillations have significant variations in amplitude and timing. In order to test the practicality of using optical emissions as a monitoring tool, the SABER data were fit to the measured variations. Even the most simple fit that used only filtered carbon dioxide emissions had good correlations with the measured oscillations. However, the density oscillations were also well predicted by a simple Fourier series, contrary to original expectations. Nevertheless, measurements of the optical emissions from the thermosphere are expected to have a role in future understanding and prediction of the semiannual variations.

This study is focused on improving the poorly understood seasonal dependence of northern high-latitude F region thermospheric winds under active geomagnetic conditions. The gaps in our understanding of the dynamic high-latitude thermosphere are largely due to the sparseness of thermospheric wind measurements. With current observational facilities, it is infeasible to construct a synoptic picture of thermospheric winds, but enough data with wide spatial and temporal coverage have accumulated to construct a meaningful statistical analysis. We use long-term data from eight ground-based and two space-based instruments to derive climatological wind patterns as a function of magnetic local time, magnetic latitude, and season. These diverse data sets possess different geometries and different spatial and solar activity coverage. The major challenge is to combine these disparate data sets into a coherent picture while overcoming the sampling limitations and biases among them. In our previous study (focused on quiet time winds), we found bias in the Gravity Field and Steady State Ocean Circulation Explorer (GOCE) cross-track winds. Here we empirically quantify the GOCE bias and use it as a correction profile for removing apparent bias before empirical wind formulation. The assimilated wind patterns exhibit all major characteristics of high-latitude neutral circulation. The latitudinal extent of duskside circulation expands almost 10^∘ from winter to summer. The dawnside circulation subsides from winter to summer. Disturbance winds derived from geomagnetic active and quiet winds show strong seasonal and latitudinal variability. Comparisons between wind patterns derived here and Disturbance Wind Model (DWM07) (which have no seasonal dependence) suggest that DWM07 is skewed toward summertime conditions.

This paper investigates the large-scale seasonal dependence of geomagnetically quiet time, northern high-latitude F region thermospheric winds by combining extensive observations from eight ground-based (optical remote sensing) and three space-based (optical remote sensing and in situ) instruments. To provide a comprehensive picture of the wind morphology, data are assimilated into a seasonal empirical vector wind model as a function of season, latitude, and local time in magnetic coordinates. The model accurately represents the behavior of the constituent data sets. There is good general agreement among the various data sets, but there are some major offsets between GOCE and the other data sets, especially on the duskside. The assimilated wind patterns exhibit a strong and large duskside anticyclonic circulation cell, sharp latitudinal gradients in the duskside auroral zone, strong antisunward winds in the polar cap, and a weaker tendency toward a dawnside cyclonic circulation cell. The high-latitude wind system shows a progressive intensification of wind patterns from winter to equinox to summer. The latitudinal extent of the duskside circulation cell does not depend strongly on season. Zonal winds show a mainly diurnal variation (two extrema) around polar and middle latitudes and semidiurnal variation (four extrema) at auroral latitudes; meridional winds are primarily diurnal at all high latitudes. The strength of zonal winds channeling through the auroral zone on the duskside is strongest in the summer season. The vorticity of the wind pattern increases from winter to summer, whereas divergence is maximum in equinox. In all three seasons, divergence is weaker than vorticity.