Quality control and error assessment of the Aeolus L2B wind results from the Joint Aeolus Tropical Atlantic Campaign

• DOI: 10.5194/amt-15-6467-2022
• Publication Date: 2022-11-11T11:11:58Z
• Authors: Oliver Lux; Benjamin Witschas; Alexander Geiß; Christian Lemmerz; Fabian Weiler; Uwe Marksteiner; Stephan Rahm; Andreas Schäfler; Oliver Reitebuch
• Keywords: Comparability; Quantile; Statistical power

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