Extended validation of Aeolus winds with wind-profiling radars in Antarctica and Arctic Sweden

• Published in: Atmospheric Measurement Techniques
• Main Authors: Kirkwood, Sheila, Belova, Evgenia, Voelger, Peter, Chatterjee, Sourav, Satheesan, Karathazhiyath
• Format: Text
• Language: English
• Published: 2023
• Subjects: Aeolus; Arctic; Esrange; Mara; Antarc*; Antarctica
• Online Access: https://doi.org/10.5194/amt-16-4215-2023; https://amt.copernicus.org/articles/16/4215/2023/

Winds from two wind-profiling radars, ESRAD (ESrange atmospheric RADar) in Arctic Sweden and MARA (Moveable Atmospheric Radar for Antarctica) on the coast of Antarctica, are compared with collocated (within 100 km) winds measured by the Doppler lidar on board the Aeolus satellite for the time period July 2019–May 2021 (baseline 2B11). Data are considered as a whole and subdivided into summer and winter as well as ascending (afternoon) and descending (morning) passes. Mean differences (bias) and random differences are categorized (standard deviation and scaled median absolute deviation) and the effects of different quality criteria applied to the data are assessed, including the introduction of the “modified Z score” to eliminate gross errors. This last criterion has a substantial effect on the standard deviation, particularly for Mie winds. Significant bias is found in two cases, for Rayleigh winds for the descending satellite passes. at MARA ( −1.4 ( + 0.7) m s −1 ) and for all Mie winds at ESRAD ( + 1.0 ( + 0.3) m s −1 ) . For the Rayleigh winds at MARA, there is no obvious explanation for the bias in the data distribution. The Mie wind error with respect to the wind data measured at ESRAD shows a skewed distribution toward positive values (Aeolus horizontal line-of-sight wind > ESRAD wind). Random differences (scaled median absolute deviation) for all data together are 5.9 and 5.3 m s −1 for Rayleigh winds at MARA and ESRAD, respectively, and 4.9 and 3.9 m s −1 for Mie winds. When the comparison is restricted to Aeolus measurements with a mean location within 25 km from the radars, there is no change to the random differences for Rayleigh winds, but for Mie winds they are reduced to 3.3 and 3.6 m s −1 . These represent an upper bound for Aeolus wind random errors since they are due to a combination of spatial differences and random errors in both radar winds and Aeolus winds. The random errors in radar winds are < 2 m s −1 and therefore contribute little, but spatial variability clearly makes a ...