Correction of wind bias for the lidar on-board Aeolus using telescope temperatures

Abstract

Abstract. The European Space Agency satellite Aeolus provides continuous profiles of the horizontal line-of-sight wind component at a global scale. It was successfully launched into space in August 2018 with the goal to improve numerical weather prediction (NWP). Aeolus data has already been successfully assimilated into several NWP models and has already helped to significantly improve the quality of weather forecasts. To achieve this major milestone the identification and correction of several systematic error sources was necessary. One of them is related to small temperatures fluctuations across the 1.5 m diameter primary mirror of the telescope which cause varying wind biases along the orbit of up to 8 m/s. This paper presents a detailed overview of the influence of the telescope temperature variations on the Aeolus wind products and describes the approach to correct for this systematic error source in the operational near-real-time (NRT) processing. It was shown that the telescope temperature variations along the orbit are due to changes of the top-of-atmosphere short- and long-wave radiation of the Earth and the response of the telescope’s thermal control system to that. To correct for this effect ECMWF model-equivalent winds are used as bias reference to describe the wind bias in a multiple linear regression model as a function of various temperature sensors located on the primary telescope mirror. This correction scheme has been in operational use at ECMWF since April 2020 and is capable of reducing a large part of the telescope-induced wind bias. In cases where the influence of the temperature variations is particularly strong it was shown that the bias correction can improve the orbital bias variation by up to 53 %. Moreover, it was demonstrated that the approach of using ECMWF model-equivalent winds is justified by the fact that the global bias of models u-component winds w.r.t to radiosondes is smaller than 0.3 m/s. However, this paper also presents the alternative of using Aeolus ground return winds which serve as zero wind reference in the multiple linear regression model. The results show that the approach based on ground return winds only performs 10.8 % worse than the ECMWF model-based approach and thus has good potential for future applications for upcoming reprocessing campaigns or even in the NRT processing of Aeolus wind products.

Exploiting Aeolus Level-2B Winds to Better Characterize Atmospheric Motion Vector Bias and Uncertainty

“…Therefore, changes in the mirror shape due to thermal variations result in perceived frequency shifts of the signal. The operational M1 bias correction uses instrument temperatures as predictors and innovation departures from ECMWF backgrounds as a reference, and is shown to improve the quality of the Rayleigh and Mie signal levels, reducing the Aeolus HLOS wind bias relative to ECMWF background winds by over 80%: the global average Rayleigh-clear bias decreased to near-zero and the Mie bias decreased to -0.15 m s -1 (Abdalla et al, 2020; information regarding the limitations of the operational M1 correction are presented in Weiler et al, 2021). In this study, profiles of Aeolus Rayleigh-clear HLOS winds (hereafter RAY winds) and Mie-cloudy HLOS winds (hereafter MIE winds) are collocated with AMVs.…”

“…MIE minus IFS winds have a global mean of -0.16 m s -1 and a smaller standard deviation of 3.8 m s -1 (Abdalla et al, 2020). It is noted that the ECMWF model, the Integrated Forecasting System (IFS), is used as a reference in the calculation of the reprocessed Aeolus L2B winds, and thus a model dependency is introduced into the dataset (Weiler et al, 2021). Related NWP impact assessments show that Aeolus has a positive impact on operational global forecasts (Cress 2020;Rennie and Isaksen 2020b) at major NWP centers including ECMWF, the German Weather Service (DWD), Météo-France, and the UK Met Office.…”

“…Despite the high quality and positive impacts, limitations remain with the Aeolus L2B dataset (Abdalla et al, 2020;Weiler et al, 2021). Mie and Rayleigh random errors could be further improved, as the Mie error standard deviations average to approximately 3.5 m s -1 and Rayleigh error standard deviations increase from 4 m s -1 to over 5 m s -1 from July to December 2019 (Abdalla et al, 2020).…”

“…Further, MIE winds exhibit a slow (fast) wind speed dependent bias for high HLOS speeds of negative (positive) sign. Moreover, there is currently an ECMWF model dependency in the reprocessed Aeolus L2B wind dataset (Weiler et al, 2021). Additionally, at the time of writing, issues thought to be due to instrumentation or software https://doi.org/10.5194/amt-2021-277 Preprint.…”

Reply on AC2

“…Further, MIE winds exhibit a slow (fast) wind speed dependent bias for high HLOS speeds of negative (positive) sign. Moreover, there is currently an ECMWF model dependency in the reprocessed Aeolus L2B wind dataset (Weiler et al, 2021). Additionally, at the time of writing, issues thought to be due to instrumentation or software malfunctions have become apparent that affect the quality of the winds.…”