The impact of Aeolus wind retrievals on ECMWF global weather forecasts

• Published in: Quarterly journal of the Royal Meteorological Society Vol. 147; no. 740; pp. 3555 - 3586
• Main Authors: Rennie, Michael P., Isaksen, Lars, Weiler, Fabian, Kloe, Jos, Kanitz, Thomas, Reitebuch, Oliver
• Format: Journal Article
• Language: English
• Published: Chichester, UK John Wiley & Sons, Ltd 01.10.2021; Wiley Subscription Services, Inc
• Subjects: Aeolus; Algorithms; assimilation; Backscatter; Data assimilation; Data collection; Doppler sonar; Doppler Wind Lidar; Global weather; Lidar; Lower stratosphere; numerical weather prediction; NWP impact; observation biases and errors; observations; Statistical analysis; Statistical methods; Statistics; Stratosphere; Temperature gradients; Tropical climate; Troposphere; Upper troposphere; Weather forecasting; Wind; Winds
• ISSN: 0035-9009; 1477-870X
• DOI: 10.1002/qj.4142

Aeolus is the world's first spaceborne Doppler Wind Lidar, providing profiles of horizontal line‐of‐sight (HLOS) wind retrievals. Numerical weather prediction (NWP) impact and error statistics of Aeolus Level‐2B (L2B) wind statistics have been assessed using the European Centre for Medium‐range Weather Forecasts (ECMWF) global data assimilation system. Random and systematic error estimates were derived from observation minus background departure statistics. The HLOS wind random error standard deviation is estimated to be in the range 4.0–7.0 m·s−1 for the Rayleigh‐clear and 2.8–3.6 m·s−1 for the Mie‐cloudy, depending on atmospheric signal levels which in turn depend on instrument performance, atmospheric backscatter properties and the processing algorithms. Complex systematic HLOS wind error variations on time‐scales less than one orbit were identified, most strongly affecting the Rayleigh‐clear winds. NWP departures and instrument housekeeping data confirmed that it is caused by temperature gradients across the primary mirror. A successful bias correction scheme was implemented in the operational processing chain in April 2020. In Observing System Experiments (OSEs), Aeolus provides statistically significant improvement in short‐range forecasts as verified by observations sensitive to temperature, wind and humidity. Longer forecast range verification shows positive impact that is strongest at the day two to three forecast range: ∼2% improvement in root‐mean‐square error for vector wind and temperature in the tropical upper troposphere and lower stratosphere, and polar troposphere. Positive impact up to 9 days is found in the tropical lower stratosphere. Both Rayleigh‐clear and Mie‐cloudy winds provide positive impact, but the Rayleigh accounts for most tropical impact. The Forecast Sensitivity Observation Impact (FSOI) metric is available since 9 January 2020, when Aeolus was operationally assimilated, which confirms Aeolus is a useful contribution to the global observing system, with the Rayleigh‐clear and Mie‐cloudy winds providing similar overall short‐range impact in 2020. The Aeolus Doppler wind lidar satellite's winds are assessed in global NWP at ECMWF. Estimates of HLOS wind retrieval error statistics are obtained by comparison to ECMWF equivalents. A significant systematic error was discovered to vary with instrument temperatures; however, a correction was determined with the aid of NWP. The effects of Aeolus data assimilation on the analysis are assessed and we demonstrate forecast improvements through Observing System Experiments into the medium‐range in the tropics and to 3–4 days in the polar regions.