Abstract. Ground-based observations of horizontal winds have been performed in Leipzig (51.12 N, 12.43 E), Germany, and at Punta Arenas (53.35 S, 70.88 W), Chile, in the framework of the German initiative EVAA (Experimental Validation and Assimilation of Aeolus observations) with respect to the validation of the Mie and Rayleigh wind products of Aeolus (L2B data). In Leipzig, at the Leibniz Institute for Tropospheric Research (TROPOS), radiosondes have been launched on each Friday for the Aeolus overpasses (ascending orbit) since mid of May 2019. In Punta Arenas, scanning Doppler cloud radar observations have been performed in the frame of the DACAPO-PESO campaign (dacapo.tropos.de) for more than 3 years from end 2018 until end 2021. We present two case studies and long‐term statistics of the horizontal winds derived with the ground-based reference instruments compared to Aeolus Horizontal Line-of-Sight (HLOS) winds. It was found that the deviation of the Aeolus HLOS winds from the ground-reference is usually of Gaussian shape which allowed the use of the median bias and the scaled median absolute deviation (MAD) for the determination of the systematic and random error of Aeolus wind products, respectively. The case study from August 2020 with impressive atmospheric conditions in Punta Arenas shows that Aeolus is able to also capture strong wind speeds up to more than 100 m/s. The long-term validation has been performed for all product baselines since the change to the second laser (called FM-B) in June 2019 until summer 2022 and also partly for the era of the first laser (FM-A). The long-term validation showed that the systematic error of the Aeolus wind products could be significantly lowered with the changes introduced into the processing chain (different baselines) during the mission lifetime. While in the early mission phase, systematic errors of more than 2 m/s (absolute values) were observed for both wind types (Mie cloudy and Rayleigh clear), these biases could be reduced with the algorithm improvements, such as the introduction of the correction for temperature fluctuations at the main telescope of Aeolus (M1 temperature correction) with Baseline 09. Hence, since Baseline 10, a significant improvement of the Aeolus data was found leading to a low bias (close to 0 m/s) and nearly similar values for the mid-latitudinal sites on both hemispheres. The random errors for the wind products were first decreasing with increasing baseline but later increasing again due to the performance losses of the Aeolus emitter. However, the systematic error is only slightly affected by this issue, so that one can conclude that the uncertainty introduced by the reduced atmospheric return signal received by Aeolus is mostly affecting the random error. Even when considering these issues, we can confirm the general validity of Aeolus observations during its lifetime. This proves the general concept of this space explorer mission to perform active wind observations from space.