Median-filter quality control for the recovery of Aeolus Mie-cloudy wind observations under low-signal conditions

Abstract. The Aeolus mission demonstrated the significant value of global Doppler wind lidar observations for numerical weather prediction (NWP) and atmospheric research by providing the first spaceborne measurements of wind profiles on a global scale. In particular, Aeolus Mie-cloudy winds showed a strong positive impact on weather forecasting despite their comparatively limited spatial coverage. However, the operational Level-2B processing was deliberately optimized for robust retrievals from strong cloud returns, resulting in the rejection of many potentially useful wind observations under weak backscatter conditions such as aerosol layers, optically thin cirrus clouds, and polar stratospheric clouds. Recovering additional winds in these regions is scientifically important because they are closely linked to aerosol transport and upper-tropospheric and lower-stratospheric dynamics.

To address this limitation, a median-filter-based quality-control scheme was developed to identify physically meaningful wind observations among the large fraction of Mie-cloudy winds flagged as invalid by the operational processing. The method compares each wind result with neighbouring observations in along-track distance and altitude space using a computationally efficient k-d tree approach. Invalid winds are recovered only if they are dynamically consistent with the surrounding wind field and satisfy additional plausibility criteria designed to suppress gross retrieval artefacts.

The performance of the algorithm was evaluated for representative Aeolus orbits throughout the mission and for month-long periods in September 2022 and April 2023. The method substantially increases the number of useful Mie-cloudy wind observations while maintaining acceptable error characteristics. Depending on the selected filter configuration, the number of retrieved winds increases typically by 25–60%, whereas the associated increase in random error remains below 1 m/s. Particularly strong improvements are achieved in the upper troposphere and lower stratosphere, including during measurements of the 2022 Hunga Tonga stratospheric volcanic aerosol plume, when the coverage of wind observations with departures from the model background of less than 2.5 m/s increased by up to a factor of six in the lower stratosphere. The developed approach therefore significantly enhances the scientific and meteorological value of the Aeolus dataset and provides a promising approach for future spaceborne Doppler wind lidar missions such as Aeolus-2.