Observation-constrained reconstruction and validation of an integrated zonal wind dataset over Langfang

Zonal wind in the mesosphere and lower thermosphere (MLT) is a primary control parameter for gravity-wave filtering, momentum deposition, and the propagation of tides and planetary-scale disturbances. In this work, an observation-constrained hourly zonal-wind dataset for 2024 is constructed by combining medium-frequency radar (MF), meteor radar (MR), and the Specified-Dynamics Whole Atmosphere Community Climate Model with thermosphere and ionosphere extension (SD-WACCMX) on a common 80–100 km grid with 2 km vertical spacing. The reconstructed products are evaluated against the JAGUAR-DAS whole neutral atmosphere reanalysis (JAWARA), whose pressure-level winds are interpolated to the target grid by using JAWARA’s own native geometric-height variable Z. Two symmetric calibration frameworks are examined, with MF and MR alternately treated as the reference. Local linear transfer functions are estimated in ±5 d windows at each height from samples satisfying n_overlap > 160, and the resulting slope, intercept, and correlation coefficient r fields are completed by using robust two-dimensional time–altitude interpolation before grid-point recalibration and three-source fusion. Meteor radar provides the strongest observational constraint (91.42% coverage), and radar–radar consistency is markedly stronger than radar–model consistency. Relative to native-height JAWARA, both fused products outperform the raw SD-WACCMX; the MR-reference solution yields zonal-wind bias, mean absolute error, root mean square error, and r values of 16.223 m s⁻¹, 29.250 m s⁻¹, 37.736 m s⁻¹, and 0.518, compared with 18.427 m s⁻¹, 35.969 m s⁻¹, 45.955 m s⁻¹, and 0.265 for the raw SD-WACCMX. These results show that a single-station, observation-constrained reconstruction can preserve the dynamically relevant seasonal, amplitude, and vertical-shear structure while improving external skill relative to an unconstrained whole-atmosphere model background.