Comment on amt-2022-223

Abstract

Since the start of the European Space Agency’s Aeolus mission in 2018, various studies were dedicated to the evaluation of its wind data quality, and particularly to the determination of the systematic and random errors of the Rayleigh-clear and Mie-cloudy wind results provided in the Aeolus Level-2B (L2B) product. The quality control (QC) schemes applied in the analyses mostly rely on the estimated error (EE), reported in the L2B data, using different and often subjectively chosen thresholds for rejecting data outliers. This work gives insight into the calculation of the EE for the two receiver channels and reveals its limitations as a measure of the actual wind error. The spatial and temporal variability of the Rayleigh and Mie EE values, together with the inconsistency in the QC approaches, hampers the comparability of the results across different validation studies. It is demonstrated that a precise error assessment of the Aeolus winds necessitates a careful statistical analysis, including a rigorous screening for gross errors to be compliant with the error definitions formulated in the Aeolus mission requirements. To this end, the modified Z-score and normal quantile plots are shown to be useful statistical tools for effectively eliminating gross errors and for evaluating the normality of the wind error distribution in dependence on the applied QC scheme, respectively. The influence of different QC approaches and thresholds on key statistical parameters is discussed in the context of the Joint Aeolus Tropical Atlantic Campaign (JATAC), which was conducted in Cabo Verde in September 2021. Aeolus winds are compared against model background data from the European Centre for Medium-range Weather Forecasts (ECMWF) before assimilation of Aeolus winds and against wind data measured with the 2-µm heterodyne-detection Doppler wind lidar (DWL) onboard the Falcon aircraft. The two studies make evident that the error distribution of the Mie-cloudy winds is strongly skewed with a preponderance of positively biased gross errors distorting the statistics if not filtered out properly. Effective outlier removal is accomplished by applying a two-step QC based on the EE and the modified Z-score, thereby ensuring an error distribution with a high degree of normality while retaining a large portion of wind results from the original dataset. The same is true for the Rayleigh-clear winds, although the errors are more homogeneously distributed. After utilization of the described QC approach, the systematic errors of the L2B Rayleigh-clear and Mie-cloudy winds are determined to be below 0.3 m s-1 with respect to both the ECMWF model background and the 2-µm DWL. Differences in the random errors relative to the two reference datasets (Mie vs. model: 5.3 m s-1, Mie vs. DWL: 4.1 m s-1; Rayleigh vs. model: 7.8 m s-1; Rayleigh vs. DWL: 8.2 m s-1) are elaborated in the text.