One‐night (28 October 2003) temperature and horizontal wind measurements by a resonance sodium (Na) wind/temperature lidar at Maui (20.7°N, 156.3°W) and temperature measurement by a Rayleigh lidar at Mauna Loa Observatory (MLO, 19.5°N, 155.6°W), Hawaii, were used to study gravity wave (GW) propagation from the lower stratosphere to the lower thermosphere. A dominant wave mode was identified from 35 to 103 km. The wave was partially dissipated and propagating upward with a scale height of temperature amplitude at ∼14 km. A damping layer was present around the stratopause where the wave amplitude was small, which also corresponded to a low static stability layer. The vertical wavelengths were larger in the mesosphere (12–13 km) than in the stratosphere (6–7 km), consistent with the decreasing static stability with altitude. Hodograph analysis of the Na lidar wind data showed that the wave was propagating northward and the horizontal wavelength was 2140 km and intrinsic period was 15 h in the region 84–103 km. The apparent period was ∼6 h and consistent with Doppler shift of the background wind. It is suggested that the convective zone over the equator to the south of Hawaii provided a constant GW source that is responsible for the observed GW throughout the night.

Ground-based remote sensing of atmospheric parameters is often limited to single station observations by vertical profiles at a certain geographic location. This is a limiting factor for investigating gravity wave dynamics as the spatial information is often missing, e.g., horizontal wavelength, propagation direction or intrinsic frequency. In this study, we present a new retrieval algorithm for multistatic meteor radar networks to obtain tomographic 3-D wind fields within a pre-defined domain area. The algorithm is part of the Agile Software for Gravity wAve Regional Dynamics (ASGARD) and called 3D-Var, and based on the optimal estimation technique and Bayesian statistics. The performance of the 3D-Var retrieval is demonstrated using two meteor radar networks: the Nordic Meteor Radar Cluster and the Chilean Observation Network De Meteor Radars (CONDOR). The optimal estimation implementation provide statistically sound solutions and diagnostics from the averaging kernels and measurement response. We present initial scientific results such as body forces of breaking gravity waves leading to two counter-rotating vortices and horizontal wavelength spectra indicating a transition between the rotational k⁻³ and divergent $k^{-\mathrm{5}/\mathrm{3}}$ mode at scales of 80–120 km. In addition, we performed a keogram analysis over extended periods to reflect the latitudinal and temporal impact of a minor sudden stratospheric warming in December 2019. Finally, we demonstrate the applicability of the 3D-Var algorithm to perform large-scale retrievals to derive meteorological wind maps covering a latitude region from Svalbard, north of the European Arctic mainland, to central Norway.

Simultaneous sodium (Na) Doppler lidar and meteor radar measurements of horizontal winds in the mesopause region over Maui, Hawaii, were collected in July 2002 and October/November 2003. The coincident measurements span 96 hours and altitudes between 80 and 100 km. Statistical comparisons are carried out on radar/lidar winds with 1 hour and 4 km time and height resolution, respectively. The RMS radar/lidar wind component differences observed in this study are in the range 12–17 m/s at altitudes below 96 km. This is smaller than the RMS differences observed in a previous Na lidar and meteor radar comparison. Lidar wind component variances exceed radar variances, and radar/lidar covariance, is nearly equal to the radar variance. Excess variance observed by the lidar is consistent with the fact that the meteor radar cannot resolve wind perturbations with horizontal scales smaller than ∼200 km, whereas the lidar will respond to all horizontal scales. Close correspondence between the radar wind variance and radar/lidar covariance suggests that measurement errors associated with the radar winds are swamped by geophysical variation. Furthermore, the excess lidar variance exceeds lidar estimation errors by a large factor, indicating that the lidar measurement errors are also insignificant relative to geophysical variations. Together these observations suggest that the observed radar/lidar differences are a consequence of the different horizontal wave number filters associated with the techniques, and hence the differences are determined by the strength and shape of the horizontal wave number spectrum for wind perturbations at scales smaller than ∼200 km.