Inter-comparison of wind measurements in the atmospheric boundary layer with Aeolus and a ground-based coherent Doppler lidar network over China

. After the successful launch of Aeolus which is the first spaceborne wind lidar developed by the European Space 15 Agency (ESA) on 22 August 2018, we deployed several ground-based coherent Doppler wind lidars (CDLs) to verify the wind observations from Aeolus. By the simultaneous wind measurements with CDLs at 17 stations over China, the Rayleigh-clear and Mie-cloudy horizontal-line-of-sight (HLOS) wind velocities from Aeolus in the atmospheric boundary layer are compared with that from CDLs. To ensure the quality of the measurement data from CDL and Aeolus, strict quality controls are applied in this study. Overall, 52 simultaneous Mie-cloudy comparison pairs and 387 Rayleigh-clear comparison pairs from this 20 campaign are acquired. All of the Aeolus-produced L2B Mie-cloudy HLOS, Rayleigh-clear HLOS and CDL-produced HLOS are compared individually. For the inter-comparison result of Mie-cloudy HLOS wind and CDL-produced

ms  respectively, while the "y=ax" slope, the "y=ax+b" slope and the "y=ax+b" intercept are 1.00, 0.96 and -1.2 -1 ms  . It is found that the standard deviation, the scaled MAD and the bias on ascending tracks are slightly better lower than that on descending tracks. Moreover, to evaluate the accuracy of Aeolus HLOS wind measurements under different product baselines, the Aeolus L2B Mie-cloudy HLOS wind data and L2B Rayleigh-clear HLOS wind data under Baselines 07/08, Baselines 09/10, and Baseline 11 are compared against the CDL-retrieved HLOS wind data separately. From 30 the comparison results, marked misfits between the wind data from Aeolus Baselines 07/08 and wind data from CDL in the planetary atmospheric boundary layer are found. With the continuous calibration and validation and product processor updates, of Aeolus in the Atlantic Ocean west of the African continent was conducted by using the RV Polarstern cruise PS116 carried 65 radiosondes (Baars et al., 2020). In China, the wind observations from Aeolus are were compared with the results from groundbased Radar wind profiler network and radiosonde over China (Guo et al., 20202021b;Liu et al., 2021). There are were some significant validation campaigns as well using airborne instruments and radiosondesin the worldwide (e.g., Bedka et al., 2020;Martin et al., 2020).
As a member of the CAL/VAL teams, Ocean University of China (OUC) has performed one long-term observation 70 campaigns with 1550 nm coherent Doppler wind lidars (CDLs) all over China. During these campaigns, 439 simultaneous measurement cases are acquired with the CDLs of types Wind3D 6000 and WindMast PBL, which are manufactured by Qingdao Leice Transient Technology Co., Ltd (http://www.leice-lidar.com/en/index.html). During the data processing, it was found that the atmospheric vertical velocity could influence the HLOS wind velocity measured by Aeolus in the atmosphericplanetary boundary layer. Hence, it should be specially noted that the HLOS wind velocities from CDL and Aeolus 75 are different and should be corrected.
This paper provides the inter-comparison of the horizontal-line-of-sight (HLOS) wind velocities measured by CDL and Aeolus. The paper is organized as follows: in section 2 the simultaneous validation campaigns and the instrument deployed for the measurements are described. Section 3 presents the details to the inter-comparison strategy, the quality control and vertical velocity correction procedure. In section 4 we provide the HLOS wind velocity measurement examples and 80 comparison results. Section 5 summarizes the recent comparison results and compares those with ours.

Overview of the validation campaigns
Shortly after the successful launch of the Aeolus, the primary laser head FM-A (Flight Model -A) was switched on and an initial laser pulse energy of 65 mJ was achieved (Lux et al. 2020b). During the period from 14 January and to 14 February 85 2019, Aeolus was in standby-mode and switched-on with FM-A. After a final test with laser FM-A of Aeolus on 17 June, the transition to laser FM-B took place. About half a year later, the validation campaign (VAL-OUC) performed by the Ocean University of China has been carried out since January 2020 at 17 stations. The comparison results of HLOS wind velocities in the atmospheric boundary layers from CDLs and Aeolus) are presented in section 4. The duration of the validation campaign (VAL-OUC) is was from January to December 2020. The locations of the CDLs, the ascending and descending orbits of 90 Aeolus are shown in Fig. 1. An overview and detailed information of the validation campaign are provided in Table 1.

The ALADIN and CDL descriptions
In this subsection, the unique payload of Aeolus, the Atmospheric Laser Doppler Instrument (ALADIN), and the ground-based reference coherent Doppler wind lidar are briefly described.

ALADIN 100
ALADIN is a direct detection high spectral resolution wind lidar which operates at the wavelength of 354.8 nm with a laser pulse energy around 65 mJ and with a repetition of 50.5 Hz (Lux et al. 2020b). It is equipped with a 1.5 m diameter telescope to collect the backscatter light from molecules and aerosol particles. The high spectral resolution design of ALADIN allows for the simultaneous detection of the molecular (Rayleigh) and particle (Mie) backscattered signals in two separate channels, each sampling the wind in 24 vertical height bins with a vertical range resolution between 0.25 km and 2.0 km. This makes it 105 possible to deliver winds both in clear and (partly) cloudy conditions down to optically thick clouds at the same time. The horizontal resolution of the wind observations is about 90 km for the Rayleigh channel and about 10-15 km for the Mie channel.
A detailed description of the instrument design and a demonstration of the measurement concept are introduced in e.g. (Reitebuch et al., 2009;2012;Straume et al., 2018;ESA 2008;Marksteiner 2013).
The data products of Aeolus are processing processed at different levels mainly including Level 0 (instrument housekeeping 110 data), Level 1B (engineering-corrected HLOS winds), Level 2A (aerosol and cloud layer optical properties), Level 2B (meteorologically-representative HLOS winds) and Level 2C (Aeolus-assisted wind vectors from ECMWF model) (Tan et al., 2008;Rennie et al., 2020a). In this study, the Level 2B HLOS wind velocities are used. Within the Level 2B processor, the Rayleigh-clear and Mie-cloudy winds are classified and the temperature and pressure correction are applied for the Rayleigh wind retrieval. 115

Coherent Doppler wind lidar instrument
Lidar is one of the most accurate optical remote sensing techniques for wind field measurements. The 1550 nm wavelength all-fiber Coherent Doppler wind Lidar (CDL) with high resolution takes advantage of the fact that the frequency of the echo signal is shifted from the local-oscillator light because of the Doppler effect which occurs from backscattering of aerosols. The Doppler frequency shift in the frequency of the backscattered signal is analyzed to obtain the LOS velocity along the lidar 120 beam direction. The CDL is based on the heterodyne technique, consisting of a single frequency seed laser source, an acoustooptic modulator, an Erbium doped fiber amplifier, optical isolators and amplified spontaneous emission noise filters, an optical switch, a transceiver telescope, a balanced detector and an analog-to-digital converter and a Fast Fourier Transform signal processor. Further information regarding the CDL is described in a separate paper (Wu et al., 2016).
The CDL of types Wind3D 6000 and WindMast PBL are lidar systems for wind measurements in the lower atmosphere. 125 The devices are were developed by the Leice Transient Technology and designed with consideration for the needs of the meteorological application, wind energy industry and aviation safety. The specifications of the CDLs are listed in Table 2.  Fig. 2. The horizontal distance between the Wind3D 6000, the WindMast PBL and the mast are around 6 m. The met mast configuration is compliant with IEC 61400-12-1 Edition 2 (IEC, 2011). All cup anemometers installed on the reference mast are class 0.9A instruments and have undergone individual rotor specific MEASNET calibration at a MEASNET certified wind tunnel. Data acquisition systems sample all input ports and connected sensors continuously with a sampling rate 135 of 1 Hz and compress the values to 10-minute-average-values. The specifications of the cup anemometers and wind vanes are listed in Table 3. The measurement heights selected for comparison are 50 m, 100m. Figure 3 shows The the comparison results at 50 m, which are wind speed and wind direction for Wind3D 6000 and WindMast PBL are shown in Fig. 2 (a) and (b), respectively. By performing ordinary least squares linear regressions of the CDLs and cup anemometers and wind vines wind measurements, the slopes, offsets, standard deviations and correlation coefficients are acquired and they are within the 140 acceptable limits. The statistic results of the validation are shown in Table 34. Hence, the CDLs of types Wind3D 6000 and WindMast PBL can be act as reference instruments for the validation of Aeolus in the atmospheric boundary layer.    0.049 (°)

Inter-comparison of Aeolus and CDL measurements
During the validation campaigns of VAL-OUC, the wind field measurements at the sites over China are continuously performed, except during the period of the CDL maintenances. 155

Inter-comparison strategy
In Fig. 34, we provide the flowchart of the comparison between Wind3D 6000/WindMast PBL measurements against Aeolus measurements. To ensure the quality of the measurement data from Wind3D 6000 and WindMast PBL, we only used the CDL data with SNR>-10 dB. For Aeolus, only observations with the corresponding "validity _flag" of TRUE, which is provided in the Aeolus L2B product, are considered. For the comparison, only the Mie-cloudy and Rayleigh-clear wind velocities from 160 the L2B product with estimated errors lower than 4 -1 ms  and 8 -1 ms  , respectively, are selected (Witschas et al., 2020).
Moreover, the Aeolus lowest atmospheric bins close to the ground are also removed from the comparison because lowest atmospheric range bins from Aeolus could be contaminated with ground. In this study, the horizontal separations between the locations of CDLs and Aeolus measurement ground track should be less than 80 km. Since the CDL provide continuous atmospheric observations, there is no time difference between CDL and simultaneous Aeolus measurements. Vertical 165 averaging of the CDL measurements over 1 Aeolus range bin is also performed.

Figure 34. Sketch of the comparison between CDL and Aeolus in the atmospheric boundary layer.
During the measurement procedure, to To observe the three-dimensional wind speed and direction, the Doppler Beam Swing 170 (DBS) scanning mode of CDLs is applied. The five-beam DBS scanning technique is mainly used to retrieve the wind profiles by measuring the LOS wind speeds in the vertical, the north, the east, the south, and the west directions. The original wind product of CDL is one second average results (Level 2 product). By considering the low horizontal spatial resolution of Aeolus data (about 90 km for the Rayleigh-clear wind velocities and 10 km for the Mie-cloudy wind velocities), 30 minutes (±15 minutes) average of CDL wind product is applied and the nearest observations profile provided by CDL and Aeolus is selected 175 by using the geolocation information in each measurement case. Besides, since Since Aeolus can only deliver the HLOS winds data, the simultaneous wind measurements from CDL have to be projected onto the Aeolus HLOS direction using the azimuth angle from Aeolus. The CDL-HLOS wind (HLOSCDL) is calculated as CDL-EW V VCDL-EW, CDL-SN V VCDL-SN are the east-west wind speed and the south-north wind speed measured by CDL respectively, 180 Aeolus Azi AziAeolus is the azimuth angle of ALADIN provided by with the Aeolus products.

Influence of vertical velocity in the atmospheric boundary layer
In the atmosphericplanetary boundary layer, the vertical velocity of air mass has shows pronounced impact on the HLOS wind velocity measured by Aeolus. The schematic diagram of the vertical velocity impacts on the HLOS velocity retrieval is presented in Fig. 45. Hence, the difference between the HLOS wind velocities from CDL and Aeolus should be specially 185 noticed during the data processingtreated for the comparison.

Results and discussion
In the validation campaign, the CDLs of types Wind3D 6000 and WindMast PBL are deployed at different observations sites.
According to the sketch illustrated in Fig. 34, the measurement data from CDLs and Aeolus are processed. In this section, some examples of single profiles and a statistical analysis is presented.

Profiles comparison 225
In Fig. 6, a measurement case of wind field observed with Wind3D 6000 in Zhangye, Gansu Province on 10 May 2020 is provided. In this figure, the Aeolus Mie-cloudy HLOS wind velocity ( Fig. 6 (a)) and Rayleigh-clear HLOS wind velocity ( Fig.  6 (b)) on 10 May 2020 are shown. The red dashed lines in Fig. 6 (a) and (b) show the nearest observation profiles on this orbit to Zhangye. Meanwhile, the profiling of the SNR (Fig. 6 (c)), wind velocity ( Fig. 6 (d)), wind direction (Fig. 6 (e)) and vertical velocity ( Fig. 6 (f)  To compare the measurement results in ZhangyeQingdao, the simultaneous profiles of HLOS wind velocities observed with Aeolus and Wind3D 6000 are provided in Fig. 7. Firstly, it should be introduced that the Aeolus L2B data of this case was produced by the processor Baseline 11. In this figure, the CDL-retrieved HLOS wind velocities with and without vertical 250 velocity correction are compared against the Aeolus Mie L2B products and Rayleigh L2B products. The vertical velocity profile is plotted as well. From this figure, it is found that the Aeolus L2B Rayleigh-clear HLOS products profile in the atmosphericplanetary boundary layer are not always available but are almost trustable , when they are providedexcept for the lowest height bin of Aeolus Rayleigh-clear HLOS profile, which has a large bias compared with the CDL-retrieved HLOS wind. In the aspect of tThe Aeolus L2B Mie-cloudy HLOS profile, in the atmospheric planetary boundary layer, the 255 synchronous CDL measurements are in the range of Aeolus estimated error. fit well with the synchronous CDL measurements.
Additionally, the 30-minute averaged vertical velocity profile shows that the vertical velocity is around -0.16in the range of  Fig. 8 (a), it is found that the Aeolus L2B Mie-cloudy products in the atmosphericplanetary boundary layer fit well with the CDL-retrieved HLOS wind velocities. In Fig. 8 (b), (c) and (d), the CDL-retrieved HLOS wind velocities and the Aeolus L2B Rayleigh-clear HLOS products agree well from the atmosphericplanetary boundary layer to the altitude of around 4 6 km while the CDL-retrieved profile are all in the range of 275 Aeolus estimated errors. It should be emphasized that the Xidazhuangke (Beijing) case uses the Aeolus L2B HLOS wind data on 21 January, which is from Baseline 07, while the processor of the Aeolus data in the Lanzhou (Gansu Province) case on 11 April is Baseline 08 and the processor of the Aeolus data in the Wuwei (Gansu Province) case and Huludao (Liaoning Province) case on 15 November and 16 November are Baseline 10 and Baseline 11, respectively. It is because the adaptive bias correction based on ECMWF data and M1 telescope temperatures which was added after Baseline 09 was not yet in place for Baseline 280 07 that there is the noticeable bias for the Rayleigh channel winds in the Xidazhuangke (Beijing) case (Rennie et al., 2020b).
Besides, in the inter-comparison case of Huludao (Fig. 8 (cd)), in the altitude of around 1.2 km to 1.7 km, the vertical velocity measured by CDL is larger than 1.00 -1 ms  , which could produce introduce the error of about 1.33 -1 ms  if it is not considered.
The vertical velocity corrected results (the yellow line) show the better agreement with the Aeolus L2B Rayleigh-clear HLOS wind velocities than the original CDL-retrieved HLOS wind velocities. 285

Statistics Statistical comparison
In this section, we compare the HLOS wind velocity results from Aeolus observations with the accompanying ground-based CDLs measurements. During the time period of January to December 2020 within the VAL-OUC campaign, 52 simultaneous Mie-cloudy comparison pairs and 387 Rayleigh-clear comparison pairs at 17 stations are acquired. Figure 9 shows the counts numbers of the comparison data pairs at different detection height ranges of Mie-cloudy channel and Rayleigh-clear channel. 290 It can be seen that the heights of the comparison pairs are mainly in and close to the atmosphericplanetary boundary layer.

Figure 9. Counts of data pairs at different height ranges of (a) Mie-cloudy vs CDL and (b) Rayleigh-clear vs CDL.
In Fig. 10, the Mie-cloudy HLOS wind velocities and Rayleigh-clear HLOS wind velocities from Aeolus are compared with that from CDL, respectively. Figure 10  ms  respectively, while the "y=ax" slope, "y=ax+b" slope and "y=ax+b" intercept are 0.93, 0.92 and -0.33 -1 ms  . In Fig. 10 (c), the scatter diagram of Aeolus L2B Rayleigh-clear HLOS and CDL HLOS data are plotted. There are 387 comparisons are taken into consideration. Accordingly, the correlation coefficient, the standard deviation, the scaled MAD, and the bias are 0.62, 7.07 -1 ms  , 5.77 -1 ms  , 300 -1.15 -1 ms  respectively, while the "y=ax" slope, "y=ax+b" slope and "y=ax+b" intercept are 1.00, 0.96 and -1.20 -1 ms  . Table   4 5 summarizes the statistical results of the comparison. It should be emphasized that before these comparisons process, the outlier control is conducted firstly. the The data with HLOS differences larger than one the original standard deviation (5.89  against that from CDL. In Fig. 10 (a) and (c), the red dotted lines represent the "y=ax" fitting lines; the blue lines represent the "y=ax+b" fitting lines; the black lines represent the "y=x" reference line. Figure 10 (b) and (d) show the histogram of counts of HLOS wind velocities, where the blue columns represent the count of CDL HLOS wind velocities and the red columns represent the count of Aeolus HLOS wind velocities.   Additionally, the scatter plots and the statistics histograms of Aeolus Rayleigh-clear HLOS against the CDL-retrieved HLOS according to the measurements made on ascending and descending tracks are presented individually in Fig. 11. Figure 11 (a) indicates the comparison between the Aeolus ascending measurements against that from CDL. It is found that the correlation coefficient, the standard deviation, the scaled MAD and the bias are 0. ms  , respectively, while the "y=ax" slope, "y=ax+b" slope and "y=ax+b" intercept are 1.02, 1.03 and -0.23 -1 ms  . Figure 11 (bc) shows the 325 comparison between the Aeolus descending measurements against that from CDL. The correlation coefficient, the standard deviation, the scaled MAD and the bias are 0.51, 7.47 -1 ms  , 6.06 -1 ms  and -2.00 -1 ms  , respectively. Besides, the "y=ax" slope, "y=ax+b" slope and "y=ax+b" intercept are 0.97, 0.78 and -2.61 -1 ms  . Consequently, the standard deviation, the scaled MAD and the bias on ascending tracks are slightly better lower than that on descending tracks. Especially, it can be found that the significantly negative bias exists in the comparison result of descending HLOS and CDL-retrieved HLOS. The statistic 330 results are summarized in Table 56. The count comparison histograms of Aeolus Rayleigh ascending and descending HLOS winds against CDL-retrieved HLOS winds are presented in Fig. 11 (b) and (d) individually.  Since the time periodduration for the comparison of Aeolus and CDL synchronous measurements lasted during 2020, the baselines of the Aeolus product changed accordingly during this period (Rennie et al., 2020b). From baseline 07 to baseline 08 of the L2B wind product processor, the associated new auxiliary file carrying the parameters for needed for the M1 mirror 340 temperature correction were provided, but not used in the L2B processing. After the deployment of baseline 09, the new auxiliary file with the M1 mirror temperature correction parameters were used, thus correcting for the associated biases in the L2B wind product. The bias corrected dataset consists of baseline 09 data from 1 to 20 April 2020 and baseline 10 data from 20 April 2020 to 8 October 2020 and the FM-B low bias reprocessed dataset of 28 June 2019 to 31 December 2019.The baseline 10 dataset consists of baseline 09 data from April 2020 to October 2020 and the FM-B low bias reprocessed dataset 345 of 2019. With the Baseline 11 processor deploying, different SNR thresholds for classification of Mie and Rayleigh and an option to transfer Mie SNR results to the Rayleigh channel were added, which allows to do SNR based classification for the Rayleigh channel, resulting in a clear quality improvement. Therefore, to evaluate the impact of updating baseline products on the HLOS measurements, the Aeolus L2B Mie-cloudy HLOS data is compared with the corresponding CDL HLOS data in Fig. 12 (a), meanwhile the Aeolus L2B Rayleigh-clear HLOS data from Baseline 07 and 08, Baseline 09 and 10, and Baseline 350 11 are compared against the CDL-retrieved individually in Fig. 12 (ba), (c) and (de).
During the comparison period, the Aeolus L2B HLOS measurements between January and April 2020 are produced with the baseline 07 and 08, the measurements between May and September 2020 are with the baseline 09 and 10, and the rest measurements from October 2020 are supported by Baseline 11.In Fig. 12 (a) ms  , respectively. Hence the Aeolus products with Baseline 07 and 08 need toshould be calibrated furtherly. From the contrast results shown in Fig. 12 (c) and (e), thanks to the M1 mirror temperature correction from baseline 08 processor to baseline 09 processor and continuous calibration and validation activities carried out by the CAL/VAL team of Aeolus, the correlation coefficients, the standard deviations, the scaled MAD and the biases are significantly improved than that from Baselines 07/08. The correlation 360 coefficient reaches to 0.75 (0.86) for scatter plot with Baselines 09/10 (Baseline 11). The corresponding standard deviation and scaled MAD decreases to 4.66  Table 67. From Fig. 12(b), (d) and (f), the count histograms of comparison also show the significant improvement of the comparison results from baseline 07/08 to baseline 09/10 and baseline 11. 365     (Guo et al., 2021b) 0.94 4.2 / -0.28 1.01 -0.41 RS over China (Guo et al., 2021) 0.90 / / 0.09 0.92 -0.22  (Belova et al., 2021) in summer 0.63 (Ascend); 0.72 (Descend) 6.8 (Ascend); 6.5 (Descend) / 6.6 (Ascend); 2.3 (Ascend); 0.5 (Descend) WPR over Japan (Iwai et al., 2021) Table 7a 8a and Table 7b8b. From Table 78, the statistical parameters including correlation coefficient, SD, MAD, bias, slope and intercept of recent calibration and validation campaigns show consistent tendency and similar comparison results. The deviations among all of these studies may result from the differences 385 in operation strategies, spatial distances and temporal gaps and so on. In summary, considering that this study conducts the inter-comparison with the data pairs mainly in heterogeneous atmosphericplanetary boundary layer, the statistical results of this study are reasonable and significant due to the long time period and large number of ground-sites included for the comparison over China.significative.

Summary and Conclusion 390
To evaluate the accuracy and precision of the Aeolus-retrieved wind results, ground-based coherent Doppler wind lidars are ms  , respectively, are selected. Moreover, the Aeolus lowest atmospheric bins close to the ground are removed from the comparison. In this study, the horizontal distance between the locations of CDLs and the Aeolus footprints must be less than 80 km. Since the CDL provide continuous atmospheric observations with a temporal resolution of 1 min, theoretically, there is no time difference between CDL and simultaneous Aeolus measurements.
Vertical averaging of the CDL-produced wind measurements over Aeolus range bins is performed. Overall, after the strict 400 quality control and assurance introduced above, 52 simultaneous Mie-cloudy comparison pairs and 387 Rayleigh-clear comparison pairs from this campaign are acquired. ms  respectively, while the 410 "y=ax" slope, the "y=ax+b" slope and the "y=ax+b" intercept are 1.00, 0.96 and -1.20 -1 ms  . Besides, the scatter diagrams and the count histogram of Aeolus Rayleigh-clear HLOS according to the measurements made on ascending and descending tracks against the synchronous CDL-retrieved HLOS are plotted individually. It is found that the standard deviation and bias on ascending tracks are slightly better lower than that on descending tracks. Especially, the significantly negative bias exists in the Aeolus Rayleigh-clear descending HLOS comparison results. Moreover, to evaluate the accuracy of Aeolus HLOS wind 415 measurements with the baselines update, the Aeolus L2B Mie-cloudy and Rayleigh-clear HLOS wind data under Baseline 07 and 08, Baseline 09 and 10, and Baseline 11 are compared against the CDL-retrieved HLOS wind data respectively. From the comparison results, marked misfits between the wind data from Aeolus Baselines 07/08 and wind data from CDL in the atmosphericplanetary boundary layers are found. After the M1 mirror temperature bias correction processor was deployed and new Rayleigh channel threshold were added, resulting in that the performances of Aeolus wind measurements under Baselines 420 09/10/11 are were improved significantly. It has to be emphasized that the misfit may result from the contamination of Mie backscatter signal to Rayleigh backscatter signal which introduces errors to the retrieval of Rayleigh-clear HLOS velocity.
Additionally, the distance between the CDL sites and the footprint of Aeolus and the strong small-scale dynamics field may cause differencesother reasons for this misfit. Finally, the statistical results of recent Aeolus wind-products calibration and validation campaigns that have been reported all over the worldso far are summarized and compared. It is figured out that this 425 study acquired similar results compared with other recent inter-comparison campaigns and all the comparison results show consistent tendency.
In the atmospheric boundary layer, the vertical velocity from convection and turbulence could influence the comparison, due to the impact of vertical velocity on the HLOS wind velocity retrieval from Aeolus.In planetary boundary layer, the vertical velocity from convection and turbulence could influence the comparison. The vertical velocity could an impact on the HLOS 430 wind velocity retrieval from Aeolus. Hence, a method is described to use the vertical velocity measured with the CDL to project onto the Aeolus LOS direction and consider it for the comparison.

Data availability
The Aeolus data are downloaded via the website of https://aeolus-ds.eo.esa.int/oads/access/collection (last accessed on 23 August 2021). The presented work includes preliminary data (not fully calibrated/validated and not yet publicly released) of 435 the Aeolus mission that is part of the European Space Agency (ESA) Earth Explorer Programme. This includes wind products from before the public data release in May 2020 and/or aerosol and cloud products, which have not yet been publicly released.
The preliminary Aeolus wind products will be reprocessed during 2020 and 2021, which will include in particular a significant L2B product wind bias reduction and improved L2A radiometric calibration. Aerosol and cloud products will become publicly available by spring 2021. The processor development, improvement and product reprocessing preparation are performed by 440 the Aeolus DISC (Data, Innovation and Science Cluster), which involves DLR, DoRIT, ECMWF, KNMI, CNRS, S&T, ABB and Serco, in close cooperation with the Aeolus PDGS (Payload Data Ground Segment). The analysis has been performed in the frame of the Aeolus Scientific Calibration & Validation Team (ACVT). To get the CDL data please contact to wush@ouc.edu.cn at Ocean University of China.