The acquisition of atmospheric wind profiles on a global
scale was realized by the launch of the Aeolus satellite, carrying the
unique Atmospheric LAser Doppler INstrument (ALADIN), the first Doppler wind lidar in space. One major component of ALADIN is its high-power, ultraviolet
(UV) laser transmitter, which is based on an injection-seeded,
frequency-tripled Nd:YAG laser and fulfills a set of demanding requirements
in terms of pulse energy, pulse length, repetition rate, and spatial
and spectral beam properties. In particular, the frequency stability of the
laser emission is an essential parameter which determines the performance of
the lidar instrument as the Doppler frequency shifts to be detected are on
the order of 10
The launch of the ESA's Aeolus mission in August 2018 was an influential event
in the history of spaceborne active remote sensing (Stith et al., 2018;
Kanitz et al., 2019; Parrinello et al., 2020). Since then, the first Doppler
wind lidar in space has offered the acquisition of global wind profiles from the
ground up to the lower stratosphere, which helps to fill observation gaps in
the global wind data coverage, particularly over the oceans, poles, tropics,
and the Southern Hemisphere (Stoffelen et al., 2020). In this manner, the
lack of wind data on a global scale, which represented a major deficiency in
the Global Observing System (GOS) (Baker et al., 2014; Andersson, 2018; NAS,
2018), was mitigated, thus contributing to the improvement of the accuracy of numerical
weather prediction (NWP) (Straume et al., 2020). On 9 January 2020, the
operational assimilation of the Aeolus wind data started at the European
Centre for Medium-Range Weather Forecasts (ECMWF), followed by the German,
French, and British weather services (Deutscher Wetterdienst, DWD;
Météo France; and Met Office) in May, June and December 2020,
respectively. Recent assessments of the significance of the Aeolus data for
NWP have demonstrated a statistically positive impact, especially in the
tropics and at the poles, thus providing a useful contribution to the GOS
(Rennie and Isaksen, 2020; Rennie et al., 2021; Martin et al., 2021). This
was made possible by the identification of and correction for large
systematic errors which had strongly degraded the wind data quality in the
initial phase of the mission (Kanitz et al., 2020; Reitebuch et al., 2020).
Firstly, dark current signal anomalies on single (“hot”) pixels of the
Aeolus detectors which had led to wind errors of up to 4 m s
Apart from these two major issues, the performance of Aeolus and particularly the precision of the Rayleigh wind results are impaired by the lower-than-expected atmospheric return signal levels, which deviated from end-to-end simulations by a factor of 2.0 to 2.5 already shortly after launch (Reitebuch et al., 2020). The situation was aggravated by a progressive decrease in emit energy of the laser transmitter during the first year of operation. Consequently, a switchover to the redundant laser onboard Aeolus was performed in summer of 2019. The second laser showed higher emit energy at a significantly lower decrease rate (Lux et al., 2020a) so that as of September 2021 it provides more than 70 mJ of pulse energy.
A common characteristic of both lasers is the occurrence of periods with
significantly increased frequency fluctuations that were not observed in the
same manner during on-ground tests. The laser frequency stability is a
crucial parameter for the Aeolus mission and many Doppler wind lidar
instruments in general given the fact that a wind speed of 1 m s
In this context of frequency stability, the topic of micro-vibrations and
their influence on the stability of spaceborne lasers is highly relevant.
Micro-vibrations are defined as vibrations at frequencies greater than
This research article aims to provide a comprehensive overview of the frequency stability of the Aeolus laser transmitters and to explore its influence on the quality of the Aeolus wind data. After a short description of the ALADIN instrument and measurement principle (Sect. 2.1), the results from on-ground tests of the laser spectral properties are recapitulated (Sect. 2.2). In Sect. 2.3, the utilization of the Mie receiver channel for the assessment of the laser frequency stability is explained, followed by an introduction of the periods that were analyzed for the present study (Sect. 2.4). Section 3 comprises the results of the analysis, starting with a presentation of the ALADIN laser frequency stability over 1 selected week of the mission (Sect. 3.1). Section 3.2 then elaborates on the correlation of this parameter with the geolocation of the satellite, leading to the detrimental influence of the reaction wheels on the spectral characteristics of the laser (Sect. 3.3) and the identification of micro-vibrations as the most likely root cause (Sect. 3.4). The fourth section addresses the question of to what extent the Aeolus data quality is diminished by the temporally increased frequency noise. After a short assessment regarding the number of affected data in different phases of the mission, the impact of the degraded frequency stability on the accuracy and precision of the wind results (Sect. 4.1) as well as on the ground velocities (Sect. 4.2) is evaluated. Finally, a summary and conclusion of the study are provided in Sect. 5 together with an outlook on possible improvements in future space lidar missions.
This section provides a brief description of the ALADIN instrument and its operating principle. After summarizing the results from on-ground tests of the laser frequency stability, the approach to assess the frequency stability in orbit by using the Mie channel is explained. Afterwards, the periods of the Aeolus mission that were selected for analysis are presented in the context of the overall performance of the ALADIN laser transmitters over the course of the 2 years after launch in 2018.
The direct-detection Doppler wind lidar ALADIN onboard Aeolus is composed of a pulsed, frequency-stabilized UV laser transmitter, transmit–receive optics (TRO), a 1.5 m diameter Cassegrain-type telescope, and a dual-channel receiver which is sensitive to both molecular and particle backscatter from clouds and aerosols (ESA, 2008; Stoffelen et al., 2005; Reitebuch, 2012; Reitebuch et al., 2018). A schematic diagram of the instrument is illustrated in Fig. 1.
Schematic of the direct-detection Doppler wind lidar ALADIN on board Aeolus. The instrument consists of two fully redundant, switchable UV laser transmitters (FM-A, FM-B), a Cassegrain telescope, transmit–receive optics (TRO) and a dual-channel receiver. The latter is composed of a Fizeau interferometer and sequential Fabry–Pérot interferometers for analyzing the Doppler frequency shift from particulate and molecular backscatter signals, respectively. HR: highly reflective mirror; FFM: flip-flop mechanism; BS: beam splitter; PBS: polarizing beam splitter; HWP: half-wave plate; QWP: quarter-wave plate; IF: interference filter; LT: light trap; LCM: laser chopper mechanism; FS: field stop; ACCD: accumulation charge coupled device. Numbers indicate the sequential light path in the receiver.
A flip-flop mechanism provides the ability to switch between the two fully redundant laser transmitters, referred to as flight models A and B (FM-A, FM-B). Both lasers are realized as diode-pumped Nd:YAG lasers in a master oscillator power amplifier configuration that are frequency-tripled to 354.8 nm emission wavelength (ESA, 2008; Cosentino et al., 2012, 2017; Lux et al., 2020a). A nonplanar Nd:YAG ring laser, frequency-locked to an ultra-low-expansion cavity, provides narrowband seed radiation that is fiber-coupled into the folded cavity of the 80 cm long Q-switched master oscillator (MO). The MO cavity length is actively controlled by means of a piezo actuator in order to find the optimal condition for single longitudinal mode operation for each laser pulse (Cosentino et al., 2017). The cavity control scheme is based on the ramp-hold-fire technique (Henderson et al., 1986), which involves the detection of cavity resonances of the injected seed radiation while sweeping the cavity length and firing the Q-switch at the detected resonance position of the piezo actuator. The actual implementation of this technique in the ALADIN MO is capable of achieving a cavity control length of better than a few nanometers. However, it has the drawback of a delay in the millisecond regime between the detection of the cavity resonance and the laser pulse emission (Trespiti et al., 2017).
The infrared (IR) single longitudinal mode output pulses from the MO (energy: 5 to 10 mJ; full width at half maximum (FWHM) pulse duration: 20 ns; pulse repetition frequency: 50.5 Hz) are amplified in a double-pass pre-amplifier and subsequent single-pass power amplifier, which are each realized by side-pumped and conductively cooled Nd:YAG zigzag slabs. The amplification stage boosts the energy of the IR pulses to more than 250 mJ, before they are guided to the harmonic generation stage of the laser. The latter comprises a set of nonlinear lithium triborate (LBO) crystals to generate UV output pulses with a conversion efficiency of about 25 %, resulting in an in-flight emit energy in excess of 60 mJ.
The UV beam from one of the two switchable laser transmitters is then
directed to the telescope, which is used in a monostatic configuration; i.e.,
signal emission and reception are realized via the same primary and
secondary mirror. A small portion (0.5 %) of the beam is separated at a
beam splitter (BS) within the TRO configuration and, after being attenuated,
guided to the instrument field stop (FS) and receiver channels. This portion
is referred to as the internal reference path (INT) signal and serves the
determination of the frequency of the outgoing beam as well as the
calibration of the frequency-dependent transmission of the receiver
spectrometers. The INT signal is thus essential for the wind measurement
principle of ALADIN, which relies on detecting frequency differences between
the emitted laser pulses and those backscattered from the atmospheric
particles and molecules moving with the ambient wind. The frequency shift
The ALADIN receiver consists of two complementary channels which
individually derive the Doppler frequency shift from the narrowband
(FWHM
The Mie and Rayleigh signals are finally detected by two accumulation
charge-coupled devices (ACCDs) with an array size of 16 pixels
According to the above equation for the Doppler frequency shift, a LOS wind
speed of 1 m s
Measurement of the ALADIN laser absolute frequency and its temporal
stability was done during pre-flight tests under vacuum conditions by means
of an external wavelength meter (High Finesse WSU10) that was calibrated
using a helium neon laser (Mondin and Bravetti, 2017). The experimental
setup was provided by the German Aerospace Center (DLR) and represents an integral part of the diagnostics
for determining the spectral properties of the ALADIN Airborne Demonstrator
(A2D) (Lemmerz et al., 2017). Using it for characterization of the ALADIN
laser, it was found that the laser frequency stability was well within the
specification requirement, which states that the root mean square (RMS)
variation in the frequency stability over 14 s should be below 7 MHz. The
14 s time period was chosen to ensure an adequate signal-to-noise ratio (SNR) over one Aeolus
observation (12 s). In addition, sensitivity tests with a thermal cycle of
An external wavelength meter is not available in space. However, the
spectrometer data gained from the Fizeau interferometer of the Mie channel
can be exploited for deriving the spectral properties of the narrowband
laser emission. For this purpose, the INT signal that is usually analyzed
after accumulation of multiple pulses on the measurement level and contained in
the L1A Aeolus data product (Reitebuch et al., 2018) is evaluated for each
individual pulse. As stated above, the Mie channel response is represented
by the centroid position of the fringe that is imaged onto the Mie detector
and then integrated over the 16 lines of the ACCD. Figure 2 shows the
vertically summed ACCD counts over the 16 horizontal pixels for one laser
pulse. The fringe centroid position is calculated by a Nelder–Mead downhill
simplex algorithm (Nelder and Mead, 1965) to optimize a Lorentzian line
shape fit of the signal distribution (Reitebuch et al., 2018). In this
manner, the Mie response is derived with high accuracy. Conversion of the
Mie response into relative laser frequency is based on dedicated in-flight
calibrations of the Mie channel from which a sensitivity of
Internal reference path Mie signal for one laser pulse (24 November 2020, 0:23:47 UTC). After vertical integration of imaged fringe on the Mie ACCD (see also Fig. 1), the signal is distributed over 16 pixels (blue bars). A Nelder–Mead downhill simplex algorithm is applied to determine the centroid position from a Lorentzian line shape fit (dark-blue line).
The laser frequency stability was analyzed for different periods of the
Aeolus mission. Since the satellite is circling around the Earth with an
orbit repeat cycle of 1 week, it was decided to study the performance over
selected 7 d periods. This approach was also motivated by the observed
correlation between the frequency stability and the satellite's geolocation
(see Sect. 3.2). After 1 week the maximum coverage of the globe is reached,
and the operation timeline, particularly the attitude control sequence of
the platform, approximately repeats. In total, 5 weeks between December
2018, when the mission was still in the commissioning phase after its launch on
22 August 2018, and October 2020 were chosen for investigation, as listed in
Table 1. The table contains information on the
operated laser transmitter as well as on the MO output energy (IR), the
laser emit energy (UV) and the laser frequency stability, which is
discussed later in the text. The energy values represent the respective mean
values and standard deviations from laser internal photodiode readings. It
should be noted that the IR energy reported by the MO photodiode of the FM-A
is considered inaccurate. Here, Q-switch discharges influence the energy
monitoring as they result in IR radiation with the wrong polarization that is circulating in the MO. This light is partially incident on the MO
photodiode, thus corrupting the energy measurement. As can be seen from the
table, the UV emit energy of FM-A decreased significantly between December
2018 and May 2019. The degradation of the laser performance was traced back
to a progressive misalignment of the MO (Lux et al., 2020a) and led to the
decision to switch to the second laser FM-B. This was necessary to ensure a
sufficient SNR of the backscatter return and thus a low random error in the
wind observations. The FM-B not only delivered a higher initial energy
after switch-on in late June 2019 (67 mJ compared to 65 mJ after FM-A
switch-on) but also has been showing a much slower power degradation. After
an initial drop by 6 mJ between July 2019 and October 2019, the UV emit
energy has decreased by less than 0.08 mJ per week. Thanks to an
optimization of the laser cold-plate temperatures in March 2020, which
increased the UV energy by about 4 mJ, the energy has remained above 60 mJ
as of the writing of this paper. Despite the better overall performance
compared to the FM-A, the FM-B behavior was observed to be more affected by
orbital and seasonal temperature variations in the satellite platform that
were transferred to the laser optical bench. As a consequence, laser
anomalies associated with larger energy variations of about
Overview of the periods that were studied in terms of the laser frequency stability. The active laser transmitter operated in the respective period (FM-A or FM-B) and the corresponding master oscillator IR output energy and UV emit energy (as reported by the laser internal photodiodes) are provided. Note that the reading of the MO photodiode for FM-A is considered inaccurate (see text). The frequency stability is given as the mean of the standard deviations from all wind observations of the respective week.
The periods listed in Table 1 are chosen such that they represent different phases of the mission: the early FM-A phase in December 2018, when the instrument parameters had settled after launch, but the degradation of the MO was already ongoing; the late FM-A phase in May 2019, when the degradation had progressed; the early FM-B phase in October, after completed thermalization of the second laser; and later FM-B periods in August and September/October 2020, after optimization of the laser cold-plate temperatures. The latter two periods were also chosen to identify the variability in the frequency stability performance decoupled from the power performance of the laser, which was stable in summer and autumn 2020 and free of temperature-related laser anomalies as stated above.
The frequency stability of the ALADIN laser is first presented with one example, namely the week in October 2019, to illustrate the main temporal characteristics of the spectral behavior as well as the relation to the geolocation of the satellite. This leads to the correlation of the laser frequency stability with platform parameters, particularly the reaction wheel speeds, which is elaborated on subsequently. This correlation is additionally analyzed for the FM-A period in May 2019 to allow for a comparison between the two flight model lasers. The section concludes with a discussion of micro-vibrations as the root cause of the frequency noise.
Figure 3a depicts a typical time series of the
laser frequency on the pulse-to-pulse level over about one orbit that was
measured for FM-B on 14 October 2019 between 01:27 and 02:57 UTC. The plot
contains the calculated Mie responses from 243 000 pulses distributed over
450 observations of 12 s each. The mean Mie response is 7.25 pixels,
corresponding to a fringe centroid position close to the center of the ACCD,
as shown in Fig. 2. The response variations are converted into relative
frequency fluctuations considering the sensitivity of the Mie channel of
ALADIN laser frequency stability:
It should be pointed out that, apart from laser frequency variations caused by cavity length changes, the measured Mie response can, in principle, also be altered by angular variations in the laser beam incident on the Fizeau interferometer. In order to estimate the contribution of angular variations, a potential correlation of the Mie response with the far-field beam position was investigated. For this purpose, the horizontal position of the two spots from the internal reference signal that are imaged onto the Rayleigh ACCD (see also Fig. 1) was analyzed during periods of enhanced noise. The studies showed a small spot motion correlated with the Mie response, which is most likely due to the influence of the Fizeau reflection that is promoted by the sequential configuration of the two receiver channels. The change in spot position was in line with the motion that was observed at times when the laser frequency was deliberately tuned, e.g., during instrument calibrations. This result strongly suggested that the contribution of angular variations to the Mie response fluctuations is negligible and that variations in the internal path Mie response are largely due to changes in the laser frequency.
Analysis of the entire week from 14 to 21 October 2019 reveals a mean frequency stability of 8.1 MHz over the 49 209 observations, as depicted in Fig. 4a, where the standard deviation is plotted for each observation of the regarded period. Data gaps in the timeline are due to special operations that are regularly performed in each week, such as instrument spectral registration (ISR; Reitebuch et al., 2018), the so-called down under dark experiment (Weiler et al., 2021a) or orbit correction maneuvers. The figure also illustrates the percentage of observations that are affected by enhanced frequency fluctuations. While the frequency stability is better than 15 MHz for the vast majority of observations (about 93 %), it is worse than 20 MHz for 2.4 % and even worse than 25 MHz for almost 1 % of the observations. However, there are also a considerable number of observation periods (19 %) for which the frequency stability is better than 5 MHz, i.e., comparable to the A2D laser performance.
To estimate the potential impact of the enhanced fluctuations on the wind
accuracy, the following calculation is performed. According to the above
equation for the Doppler shift, a frequency difference of 10 MHz is
introduced by a LOS wind speed of about 1.8 m s
Interestingly, the distribution of measured Mie responses (see Fig. 4b) indicates that the frequency fluctuations are not symmetrically distributed. Instead, the frequency tends to jump to lower values (i.e., lower responses). This behavior is also visible in Fig. 3, where the largest departures from the mean are negative. As a result, the higher the standard deviation over one particular observation is, the larger the negative shift in the respective mean from the mean over all observations will be. This relationship is shown in Fig. 4c and can most likely be traced back to disturbances of the active stabilization of the MO cavity length during the periods of enhanced frequency jitter, which results in frequency jumps preferentially in one direction. For instance, such jumps occur when the interference signal produced by the seed laser circulating in the MO features parasitic peaks that are erroneously detected as MO cavity resonances. A similar behavior was observed for the A2D in a highly vibrational environment or in the case of imperfect alignment of the MO.
The enhanced frequency fluctuations in the laser transmitter were detected very early in the mission and attributed to potential vibrations introduced by the satellite platform, which affects the MO cavity length (Lux et al., 2020a). However, a correlation to platform parameters, particularly the rotation velocity of the satellite's reaction wheels, was not found initially. This was mainly due to the fact that only short timelines were analyzed, typically covering only several minutes to hours, as shown in Fig. 3. Since the platform parameters vary slowly over the orbit, a relationship to the fast changes in the laser frequency stability within several seconds was considered unlikely.
During the first year of operation, the assessment of the frequency
stability was then focused on the weekly instrument response calibrations
(IRCs; Reitebuch et al., 2018). IRCs are required to determine the
relationship between the Doppler frequency shift in the backscattered light,
i.e., the wind speed, and the response of the Rayleigh and Mie
spectrometers. The procedure involves a frequency scan over 1 GHz in steps
of 25 MHz to simulate well-defined Doppler shifts in the atmospheric
backscatter within the limits of the laser frequency stability.
During the IRC, which takes about 16 min (two observations (12 s) for
each of the 40 frequency steps), the contribution of (real) wind related to
molecular or particular motion along the instruments' LOS is virtually
eliminated by rotating the satellite by an angle of 35
Time series of the laser frequency fluctuations over periods of
selected IRCs over Antarctica
The laser frequency stability was studied for each of the 60 IRCs conducted
between 7 September 2018 and 9 December 2019, most of them over Antarctica,
while IRCs no. 31 to no. 43 were performed over the Arctic. Here, it was
found that the frequency stability was degraded during the IRCs compared to
operation in nominal wind velocity mode (WVM). While it was on the order of
8 to 10 MHz in WVM, the mean stability over the 16 min IRC period
accounted for 12 to 14 MHz when the satellite was pointing nadir, suggesting
an influence of the platform attitude on the laser. This conclusion was
strengthened by the circumstance that the laser temperatures and energies
were strongly varying during the nadir and off-nadir slews before and after
the IRC, respectively. Moreover, these orbit maneuvers involved thruster
firings, which also caused increased frequency noise during the slews before
and after the weekly IRC, thus pointing to mechanical disturbances as the
root cause. Furthermore, the frequency stability was shown to be significantly
better over Antarctica ((11.6
Geolocation of wind observations with enhanced frequency noise for
Following these observations, the laser frequency stability was studied over
1-week periods to review the influence of the satellite's geolocation (see
Sect. 2.4). The performance from the week between 14 and 21 October
2019 (Fig. 4a), based on the Mie response data from more than 27 million
laser pulses, is illustrated in Fig. 6. Each dot in the two maps represents
one observation, whereby the color and opacity indicate the frequency
stability in terms of
The geolocational patterns were found to be reproducible for the
investigated FM-B periods with only slight variations (
Similar correlation of the laser frequency stability with the satellite position was also obvious for the periods in December 2018 and May 2019 when FM-A was operated. However, the geolocational patterns for ascending and descending orbits markedly differed from those of the FM-B periods, suggesting that the mechanism introducing the enhanced frequency noise acts differently on the two laser transmitters, potentially due to the different locations in the payload. The underlying reason for the observed dependence on geolocation could be traced back to the reaction wheels of the satellite, as is explained in the following section.
Precise three-axis attitude control of the Aeolus satellite is accomplished by a set of reaction wheels (RWs) which rotate at different speeds, thereby causing the spacecraft to counter-rotate proportionately through the conservation of angular momentum. Due to external disturbances, mainly aerodynamic drag, the total angular momentum is periodically modified so that magnetorquers are additionally required to generate an effective external torque. Otherwise the wheel speed would gradually increase in time and reach saturation (Markley and Crassidis, 2014). The attitude and orbit control system of Aeolus additionally consists of thrusters which allow for larger torque to be exerted on the spacecraft.
A sketch illustrating the orientation of the reaction wheels within the
spacecraft is shown in Fig. 7. The reaction wheel assemblies (RWAs) are
mounted on the spacecraft such that they form a tetrahedron whose axis of
symmetry lies along the
Correlation between the laser frequency stability and the speeds
of the three active reaction wheels: the plots in panel
Based on the dataset from the week in October 2019, the frequency stability on the observation level, as depicted in Fig. 6, was correlated with the rotational speed of the three active reaction wheels on board Aeolus (RWA 4 serves as a backup). The same procedure was performed for the FM-A period in May 2019 (see Table 1). The resulting six correlation plots, which can also be considered to be spectra in terms of the wheel rotation frequency (rotations per second, RPS), are shown in Fig. 8. Note that the plot includes data from both ascending and descending orbits and that RWA 1 and RWA 3 rotate counterclockwise, while RWA 2 rotates clockwise. For the sake of better comparability of the three spectra, the absolute wheel speeds are plotted in the figure, and the negative sign for the wheel speeds of RWA 1 and RWA 3 was omitted. The six spectra exhibit pronounced peaks which demonstrate that the laser frequency fluctuations are enhanced at specific rotational speeds of the reaction wheels. Thus, the latter are subsequently referred to as critical wheel speeds or critical frequencies.
For both periods, i.e., operated lasers, the frequency stability is primarily influenced by RWA 1 and RWA 2, which show a multitude of critical wheel speeds with comparable strength in their common operating range between 19 and 24 RPS. In contrast, the correlation of the frequency stability with the speed of RWA 3 is rather poor, especially for FM-A, which can be attributed to being located farther away from RWA 3 than FM-B. For RWA 1 and RWA 2 a stronger correlation to their nearby FM-A rather than to FM-B, located on the opposite side of the instrument, is not possible to demonstrate. Since the performances of FM-A and FM-B are compared for different periods of operation, and these units showed very different behavior in terms of laser performance, degradation and susceptibility to temperature variations, it is difficult to judge if one reaction wheel has a greater disturbing effect on one laser or another due to being located closer by.
The center frequencies and
Critical wheel speeds for the analyzed period between 13
and 20 May 2019 (FM-A period), as derived from the correlation with the
frequency stability depicted in Fig. 8a. The values represent the center
frequency and
Same as Table 2 but for the analyzed period between 14 and 21 October 2019 (FM-B period).
The variability in the center frequency of the common critical wheel speeds is on the order of 0.1 RPS, which is comparable to the average width of the fitted peaks. Note that the individual wheel speeds are known with an accuracy of about 0.01 RPS. The fact that the peaks are relatively narrow explains the rather short duration of the high-noise periods of several tens of seconds as the critical wheel speeds are usually passed on these timescales. Analysis of the other three periods listed in Table 1 yields that the center frequencies and widths of the peaks are constant over time for each laser. Interestingly, the peak heights in the spectra for the early FM-A period in December 2018 are lower compared to the May 2019 period plotted in Fig. 8a. This is due to the better overall performance of FM-A, particularly the better MO alignment, at the beginning of the mission (see also Table 1) so that the laser was less prone to external perturbations. Consequently, the frequency stability was less degraded at the critical wheel speeds, which manifests in smaller peaks in the spectra.
Geolocation of the most critical frequencies of the three active
reaction wheels for
Steering of the satellite pointing by means of the reaction wheel speeds
involves regular and repeated patterns over the 1-week orbit repeat cycle,
which differ only slightly depending on the elapsed time since the last
orbit correction maneuver. The global occurrence of the most critical wheel
speeds during the analyzed week in October 2019 is illustrated in Fig. 9.
The two maps show those observations during ascending (panel a) and
descending orbits (panel b), for which one of the wheels operates at one of
the critical frequencies that are printed in bold type in
Table 3. Comparison of the geolocational patterns
with those of the laser frequency stability (Fig. 6) underlines the strong
correlation between the reaction wheel speeds and the laser frequency noise.
For instance, the manifestation of high noise along a linear structure that
extends from South and North America across the Pacific Ocean to East Asia
for ascending orbits can be traced back to the influence of RWA 1 operating
at the critical frequency of
The occurrence of critical frequencies from different wheels along the orbit suggests that the three wheels act independently on the laser. This hypothesis was confirmed by further analysis, which revealed that enhanced noise is observed almost every time when one of the wheels rotates at a critical speed, regardless of the speed of the other two wheels. Hence, there is no entanglement of the critical frequencies, even though the speeds of the wheels are related among each other. Additional studies also showed that the frequency stability is not correlated with the wheel acceleration.
Geolocation of observations with enhanced laser frequency
variations and their relation with the reaction wheel speeds of RWA 1 and
RWA 2 in the period from 16:30 to 17:30 UTC on 18 October 2019. Each dot on
the map corresponds to one observation, whereby the color coding describes
the frequency stability in terms of
The impact of the reaction wheel speeds on the laser frequency stability is
finally demonstrated at an example scene, which clearly illustrates the
origination of the geolocation patterns. Figure 10
shows a map with the color-coded frequency stability per observation (dots
on the map) for the period between 16:30 and 17:30 UTC on 18 October
2019. Within this hour the satellite crossed Africa and Europe on an
ascending orbit and passed the north pole before flying over Alaska and the
Pacific on a descending orbit. The inset of the figure depicts the temporal
evolution of the wheel speeds of RWA 1 (red) and RWA 2 (green), which are
most relevant regarding the laser frequency stability, as stated above. The
timeline of the latter is plotted below. The critical speeds of RWA 1 and
RWA 2 that were identified from the correlation plots in Fig. 8b are
indicated by dashed horizontal lines of the respective color in the top
panel of the inset. The nine marked spots indicate events when the wheels
rotated at their critical speeds. As can be seen from the figure, such
events are correlated with enhanced frequency variations that result in
standard deviations of more than 15 MHz. The frequency noise was especially
high in the period between 16:50 and 17:15 UTC, when the wheel speed of
RWA 1 remained very close to
Laser frequency stability on the pulse-to-pulse level for the period
from 16:30 to 17:30 UTC on 18 October 2019 (see also Fig. 10).
As shown in this example, the existence of critical wheel speeds explains
the enhanced frequency jitter in the following situations.
The laser frequency fluctuations for the presented scene are additionally
shown on the pulse-to-pulse level in Fig. 11a together with the Allan
deviation calculated for three selected time series. The Allan deviation
(Allan, 1966) is a statistical means to determine the frequency stability
over a wide range of averaging times, which allows the identification of different
types of noise sources and drift components which affect the stability on
different timescales. While the black data points represent the entire
1 h period, the green and red data points correspond to 20 min
subperiods with comparatively low and high frequency noise, respectively.
The Allan deviation provides further information on the frequency stability
on those timescales that are relevant for the wind retrieval. Apart from
the observation level (12 s, 540 pulses) the frequency variations on
the measurement level (0.4 s, 18 pulses; see Sect. 2.1) are crucial for the wind
data quality. This is due to the fact that the signal data from individual
measurements are classified into “clear” and “cloudy” bins by using
estimates of the backscatter ratio before the on-ground accumulation to
so-called groups in the L2B processor (Rennie et al., 2020). This grouping
algorithm enables the distinction between measurements that are better
analyzed with the Rayleigh or Mie spectrometer, respectively, and hence to
minimize the effect of crosstalk between the two receiver channels that is
detrimental to the wind data quality. In particular, Mie wind results are
usually provided on timescales shorter than 12 s as they require fewer
measurement bins to achieve a given level of precision compared to the
Rayleigh winds, with the typical levels of backscatter, e.g., from clouds
(Rennie et al., 2020).
For the low-noise period, the Allan deviation on the measurement level (0.4 s
sampling period) is 1.1 MHz, which is about 50 % better than for the entire
1 h period (1.6 MHz) and almost 2 times better compared to the
high-noise period (2.1 MHz). The differences are less pronounced when
sampling on the pulse-to-pulse or observation levels is regarded. This
underlines that the enhanced frequency fluctuations typically occur in the
second regime (see also Fig. 3b) so that the mean laser frequency varies
strongly from one measurement to the other. On longer timescales, the
enhanced noise has a minor impact on the mean frequency; i.e., the variation
is smaller from observation to observation. As a result, the Allan deviation
on the observation level is around (0.7
Although the root cause of the enhanced frequency noise is understood to be introduced by the reaction wheels, there is an apparent correlation with other platform parameters. In particular, a link to the data obtained from the magnetometer on board Aeolus was discovered. A thorough investigation revealed a strong influence from the magnetic fields generated by the platform magnetorquers, which serve the regulation of the reaction wheel speeds. Due to the complex coupling between magnetorquer currents and reaction wheel speeds, there is an indirect relationship between the critical wheel speeds, i.e., the frequency stability, and the magnetometer data via the superimposed magnetic fields of the magnetorquers and the Earth.
The observation of critical speeds of the reaction wheels that give rise to enhanced frequency noise strongly suggests that micro-vibrations are the root cause of the degraded laser frequency stability in orbit. This hypothesis is further strengthened by the fact that the frequency noise levels also increase during periods of heavy thruster firings, e.g., during orbit maneuvers and wheel speed set point changes before and after IRCs (see Sect. 3.2).
When the thrusters are not used, the dominating micro-vibrations occurring in the satellite are a result of its structure responding to disturbing forces and moments generated by the reaction wheels. These disturbances result mainly from static and dynamic imbalances of the flying wheel, bearing imperfections, and the structural modes of the wheel assembly and are exerted on the satellite structure as a combination of harmonics, each with a frequency being a constant multiple of the wheel speed. Typically, these disturbances become most intense at the wheel speeds at which the critical harmonics excite the structural modes of the wheel assembly. Please refer to Le (2017) for a detailed characterization and study of the mechanical disturbances generated by reaction wheels of the class of those embarked on Aeolus.
For Aeolus the reaction wheels were mounted on micro-vibration isolation suspensions, which filter the generated disturbance forces and moments. Such suspension was found to reduce the disturbances by more than 1 order of magnitude at the most critical frequencies, while the amplifications occurring at the lower frequency due to the suspension modes are minimized by the presence of viscoelastic elastomeric mounts.
The on-ground micro-vibration verification activities were quite extensive within the Aeolus project. These included disturbing the laser transmitter with representative mechanical excitation spectra, thus identifying susceptibility in the 400 to 600 Hz frequency band as well as around 250 Hz. Moreover, micro-vibration tests were performed at the satellite level, first with the aim of characterizing the micro-vibration environment throughout the satellite. This was achieved by operating the reaction wheels all over their operational speed range and having the satellite mounted on doughnut-shaped cushions to isolate it from external disturbances (Lecrenier et al., 2015). The tests demonstrated that, despite the fact that the peak of disturbances from the reaction wheels coincided with the most susceptible frequencies of the lasers, the vibration levels were lower than the danger levels previously attributed to them thanks to the isolation suspension. The margins observed between the micro-vibration levels measured during the full-satellite test and the danger levels characterized during the preceding laser tests were considered large enough to cover the mechanical changes that occur when passing from the ground to the orbital environment. Due to the different support boundary conditions of the satellite and the effects of gravity and air, the structural damping and natural frequencies of the satellite structure are expected to differ in orbit from their characterization on the ground. Moreover, a worsening of the disturbance signature of the reaction wheels as a result of their exposure to the vibrations of the satellite-level tests and the launch environment is also commonly observed.
These margins were finally confirmed by an in situ test during the thermal vacuum campaign with the flight model of the spacecraft. This included operation of the laser whilst the wheels were running at speeds previously identified as critical and after wheel power-off. The Mie and Rayleigh frequency response data from these tests clearly showed significant peaks at several harmonic frequencies (roughly 4.5, 10, 13, 16, 18.5, 21.5, 23, 24, 24.5 RPS) when the wheels were operating. In fact, the pulse-to-pulse frequency stability varied along the thermal plateau more strongly than during the dedicated reaction wheel test. Unfortunately, due to programmatic constraints, no tests were ever run with the spacecraft mounted in an isolated configuration from the ground and with flight-representative time-varying speed profiles whilst operating the laser to check its frequency stability.
After having identified the root cause of the temporally degraded frequency stability in orbit, the question is to what extent the Aeolus wind data quality are diminished during the periods of enhanced frequency noise. Before answering this question, it is meaningful to consider how often these periods occur and how long they last. For this purpose, the percentage of observations for which the frequency stability is significantly worse in comparison to the overall performance over a certain period can be regarded. This information is summarized in Table 4 for the 5 selected weeks of the mission that were already introduced in Sect. 2.4 as being representative of different periods of the Aeolus mission.
Laser frequency stability during different phases of the Aeolus
mission. The table provides the standard deviation of the relative frequency
at the pulse-to-pulse level (mean over all observations from 1 week) as well
as the percentage of observations (%obs) for which
The table provides the standard deviation over the entire week and the
portion of observations for which
In the following two sections, the influence of the frequency stability on the systematic and random error is assessed for the Mie and Rayleigh wind results as well as for the respective ground velocities. While the latter study is performed on the observation level (540 pulses), the impact on the wind results is discussed on the measurement level (18 pulses) for the reasons related to the wind retrieval processing that were explained at the end of Sect. 3.3.
The wind error assessment is based on the data from 2 weeks of FM-B operation in August and September/October 2020, as listed in Tables 1 and 4. These datasets already include the correction for the influence of the temperature variations across the primary telescope mirror on the wind results (M1 correction), which was implemented prior to the operational assimilation of the Aeolus wind data in NWP by various weather services (Rennie and Isaksen, 2020; Weiler et al., 2021b). The Rayleigh and Mie wind observations were extracted from the L2C product, which, in addition to a copy of the Aeolus L2B product, includes ECMWF model winds (analysis and background) provided on the same horizontal and vertical grid. Using the background model winds as the reference, the wind speed differences between observation and background (O–B) can be interpreted as the wind errors.
The L2B/C wind data are provided on a different temporal grid as the L1A data used for the calculation of the Mie responses and corresponding
frequency fluctuations. This is due to the classification of measurement
bins into “clear” and “cloudy” bins (Sect. 3.3), which results in
so-called groups of varying horizontal length. Consequently, for
investigating the influence of the frequency stability on the wind accuracy
and precision, an adaptation of the different temporal grids of the L1A/B
and L2B/C products has to be performed. Moreover, it has to be considered
that each wind profile measured within a certain period of time generally
comprises multiple wind results from the adjacent vertical bins to be
compared with the respective frequency stability within the regarded time
interval. Finally, only reliable wind results with low estimated error
should enter the statistics. The estimated wind error is included in the
L2B/C product and, in the case of the Rayleigh channel, is derived from the SNR
and the pressure and temperature sensitivity of the Rayleigh responses
(Rennie et al., 2020). For the Mie channel, it is primarily linked to the
SNR. The estimated error also considers the impact of the solar background
on the wind accuracy, which is mainly relevant for the Rayleigh winds. In
the analysis presented here, estimated error thresholds of 8 m s
The results of the statistical analysis for one of the two studied periods
are shown in Fig. 12. Here, panel (a) depicts a histogram of the laser
frequency stability on the measurement level, i.e., calculated as the standard
deviation over 18 pulses within 0.4 s, i.e.,
In a next step, the Mie and Rayleigh wind results were separated into those
for which the frequency stability was better or worse than 10 MHz on
the measurement level. The resulting probability density functions (PDFs) of the
(O–B) horizontal line-of-sight (HLOS) wind speed differences are provided in
the middle and right columns of Fig. 12. Due to the much smaller number of
wind results for
The widths of the two distributions, however, are identical (standard
deviation: 4.21 m s
The same analysis was performed for the week between 28 September and 5 October 2020 and yielded similar results. Here, the percentage of Mie and
Rayleigh wind results for which the frequency stability on the measurement level
is better than 10 MHz is 78.6 % and 78.9 %, respectively. The Mie bias
increases from (0.079
Wind bias
In order to further evaluate the impact of the laser frequency noise on the
Aeolus wind data quality, a variable frequency stability threshold was
applied, and the bias and random error with respect to the ECMWF model
background were calculated for those winds that were measured during periods
with frequency fluctuations below the threshold. The statistical results are
depicted in Fig. 13. The mean Mie and Rayleigh wind biases are plotted in the
left column, while the respective random errors for the 2 investigated
weeks in August and September/October 2020 are shown on the right. Both
values are presented relative to the statistical parameters that are
obtained for a threshold of 8 MHz as this allows for a direct comparison of
the two different datasets regardless of the absolute values, which differ
among the analyzed periods and receiver channels due to other error
contributions. The reference bias and random error values are indicated in
the round boxes. A frequency stability of 8 MHz was chosen as a reference
threshold since this value is close to the mean stability over the entire
week and thus represents the average conditions in terms of laser frequency
noise. The plots additionally present the number of winds that entered the
statistics depending on the applied threshold. As can be seen from the
figure, the number of wind results increases considerably when the threshold
is relaxed from 8 to 15 MHz, whereas only few results are added at even
higher thresholds. Overall, the number of Rayleigh (
In conclusion, the temporally degraded frequency stability of the ALADIN
laser transmitter has only marginal influence on the wind data quality on a
global scale. This is primarily due to the small percentage of measurements
for which the frequency fluctuations are considerably enhanced. The biggest
impact is observed for the Mie wind bias, which is increased by more than
0.3 m s
Since wind observations with enhanced frequency jitter occur over specific
geolocations (see Fig. 6), it is interesting to study whether the wind data
are significantly degraded in such affected locations. For this purpose,
several areas where one or multiple critical reaction wheel speeds regularly
occur, i.e., in East Asia or central Africa, were analyzed with regard to
the deviations from Aeolus observations from ECMWF model winds (O–B). This
preliminary study revealed that, although the percentage of measurements
with frequency stability worse than 15 MHz is increased by a factor of 2
to 3 compared to the global percentage, the change in wind bias and
random error is below 0.5 m s
Analogous to the approach presented in the previous section, the correlation between the laser frequency stability and the apparent velocity of the ground returns was investigated. Ground return signals are generally crucial for airborne and spaceborne radar and lidar systems that rely on the Doppler effect as they can be exploited for identifying systematic errors that are, for instance, caused by improper knowledge of the platform attitude or variations in the instrument's alignment (Bosart et al., 2002; Kavaya et al., 2014; Chouza et al., 2016). For the Aeolus mission, the ground surface could be used as a zero wind reference, which allows for the estimation of unknown wind biases from the measured ground velocities. This method, however, requires precise differentiation between atmospheric and ground return signals in order to prevent erroneous ground velocities (also referred to as zero wind calibration, or ZWC, values), which is particularly challenging due to ALADIN's coarse vertical resolution of several hundred meters. The ground detection scheme and its limitations are very similar for ALADIN and the A2D and are explained in detail in Weiler (2017) and Lux et al. (2018).
The ground velocities, obtained separately for the Rayleigh and Mie channel,
are contained in the AUX_ZWC product, which is generated by
the Aeolus L1B processor. The same 2 weeks in 2020 as in the previous
section were studied. However, in contrast to the wind results, the ground
velocities which are part of the L1B baseline 10 data are not bias-corrected using the M1 telescope mirror temperatures as for the L2B product
so that they show rather large deviations from the “ideal value” of
0 m s
Histograms of the laser frequency stability on the observation level
for those observations that yielded valid Mie ground returns during the
combined periods from 17 to 24 August 2020 and from 28 September to 5 October 2020.
Analogous to Fig. 12a and d, histograms of the laser frequency stability
on the observation level are depicted in Fig. 14a and c for those
observations that yielded valid Mie and Rayleigh ground velocities,
respectively. The number of ZWC values is very similar for the two channels
(
Like for the wind results, the number of ZWC values strongly decreases when
considering only those observations for which the frequency stability is
considerably degraded. The portion of ground velocities with frequency
stability worse than 15 MHz is as low as 8.2 % for both channels. In
accordance with the findings in Sect. 4.1, there is no significant influence
of the enhanced frequency noise on the Rayleigh channel, whereas the
determined Mie ground velocities change almost linearly with the frequency
stability. When regarding only observations with 29 MHz
Mean ground velocity
The mean ZWC values are additionally calculated for a varying frequency
stability threshold, analogous to Fig. 13, in order to consider the
different weighting of the observations with low and high frequency noise.
The results are plotted in Fig. 15 for the Mie and Rayleigh channel for both the mean and the standard deviation of the ground velocities of the
respective data subset after applying the threshold. Here, a similarly weak
dependency is evident for the Mie and Rayleigh channels, where the mean ZWC
values (relative to the 8 MHz threshold) increase to 0.15 m s
From these results it can be concluded that, like for the atmospheric winds,
the enhanced frequency noise has only minor influence on the Mie and
Rayleigh ground velocities. Although the impact is slightly larger, it is
still hardly noticeable in the statistics derived from the complete ZWC
dataset and certainly represents a smaller issue than, e.g., blowing snow
affecting the Mie and Rayleigh responses. Nevertheless, the approximately
linear relationship between the frequency stability and the Mie mean ZWC
value with a slope of about 0.08 m s
The Doppler wind lidar ALADIN on board Aeolus has set new technological standards in the field of spaceborne remote sensing. In particular, the design and performance of the frequency-stabilized UV laser transmitter is unprecedented for a space laser. By the end of 2020, the two ALADIN lasers had accumulated around 3.5 billion high-energy laser pulses in 27 months of operation. The present study has shown that the frequency stability of the emitted pulses is better than 10 MHz, except for the late FM-A period in spring 2019, most probably because of the advanced misalignment of its master oscillator. The excellent stability is achieved despite the permanent occurrence of short periods during which the frequency fluctuations are considerably enhanced to 30 MHz RMS on the observation level (12 s), 50 MHz RMS on the measurement level (0.4 s), and even up to 150 MHz peak to peak on the pulse-to-pulse level.
The investigation of the frequency stability during instrument response calibrations has revealed that it is correlated with the satellite's geolocation. This correlation entails a clustering of observations with enhanced frequency variations in specific regions of the Earth, forming linear and circular structures around the globe. The patterns differ for ascending and descending orbits and the two flight model lasers but are stable over the mission lifetime.
The underlying reason for the dependency on geolocation is the existence of critical rotation speeds of the satellite's reaction wheels, which suggests that micro-vibrations are the root cause for the deterioration of the laser stability on timescales of a few tens of seconds. This hypothesis is supported by the fact that the laser stability is also degraded during thruster firings of the satellite, which introduce high vibration levels also at the location of the power laser head. The identified detrimental frequencies of the reaction wheels range between 14 and 28 RPS and are consistent among the three active wheels, although the relative impact on the two lasers is different. Owing to the dependency of the reaction wheel speeds on the magnetorquer control authority and the strong influence of the latter on the magnetometer readings, there is an indirect link and hence decent correlation between the frequency stability and the magnetic field measured by the onboard magnetometer.
In the context of the enhanced frequency noise, the Aeolus wind error with
respect to ECMWF model background winds was studied, pointing out that the
temporally degraded stability of the ALADIN laser transmitter has only marginal influence on the global wind data quality. For 2 studied weeks in 2020,
the Mie wind bias is increased by 0.3 m s
Despite the small effect on the wind data, application of a QC in the Aeolus processor could be foreseen for the Mie winds to filter out measurements during which the frequency stability is worse than 20 MHz. The same threshold is suggested for the processing of the Mie and Rayleigh ground velocities as this will slightly improve the accuracy of the ground velocity values, while their number is not considerably reduced. These QC approaches become more important if the atmospheric return SNR is further decreasing within the mission lifetime as this will increase the share of Mie wind results among the overall Aeolus wind observations. For the Rayleigh winds, a QC based on the frequency stability is not considered useful since the error is largely dominated by shot noise. Hence, the discarding of any measurements, even with poor frequency stability, will rather diminish the Rayleigh wind precision given that it is primarily limited by shot noise.
Concerning the use of Aeolus data on smaller geographical scales, e.g., for its validation by on-ground or airborne instruments, the geolocational dependence of the laser frequency stability should be considered. Regions where the reaction wheels repeatedly rotate at their critical frequencies suffer from enhanced frequency noise during every overpass along Aeolus' ascending or descending orbit. Consequently, the percentage of wind observations with degraded stability is considerably higher than on a global scale and is expected to have a stronger impact on the wind accuracy, especially for the Mie channel.
With a view to future space lidar missions, particularly those which have
strict requirements in terms of the laser frequency stability like
EarthCARE, ACDL, MERLIN or Aeolus follow-on missions, several lessons learned
are derived from the current study: first, the cavity control scheme of the
laser transmitter plays an important role in ensuring a sufficiently short
dead time of the feedback loop and in turn in filtering out vibration
frequencies of several hundreds of hertz, which potentially deteriorate the
laser stability. A modified ramp-delay-fire technique that was developed for
the A2D (Lemmerz et al., 2017) allows for fast responses in the
microsecond regime, thus providing high frequency stability (
Second, although the on-ground micro-vibration tests led to the important implementation of isolation suspensions which effectively attenuated the micro-vibration levels by about 1 order of magnitude, the tests did not clearly reveal the existence of the critical reaction wheel speeds, which would affect the performance in a measurable way. This is mainly due to the too short operation of the wheels at their critical speeds, which did not allow for sufficient statistics of the acquired laser data depending on wheel speed. Therefore, extended tests with the spacecraft mounted in an isolated configuration and with flight-representative time-varying reaction wheel speeds whilst operating the laser are recommended. Moreover, it is proposed to use an appropriate number of accelerometers at different locations of the laser bench during such tests to properly characterize the impact of micro-vibrations on the laser behavior, particularly the frequency stability. Finally, a higher temporal resolution of the laser housekeeping telemetry data than currently available for Aeolus (0.25 Hz) with a focus on cavity control parameters is advisable to better evaluate the laser performance both on the ground and in space.
Furthermore, it is considered that the installation of accelerometers in the instrument flight model and the acquisition of micro-vibration measurements in orbit, even resulting in modest numbers of data, are extremely beneficial. These not only will allow a more definitive identification of the spacecraft elements responsible for deviations in scientific measurements but will also provide a reference for correlation of mechanical models representing orbital conditions.
Passive suspensions provide a well-proven solution for the mitigation of micro-vibrations generated by reaction wheels in orbit. With their simple configuration, they have low mass, consume no electrical power and require no computational resources. However, inherent to their mechanical architecture, they produce a non-negligible amplification of the disturbances at the frequencies of the suspension modes. Also, extreme isolation at high frequencies, which can be required by some applications, is often compromised by the presence of spurious mechanical modes whose effects in orbit are difficult to predict. As an alternative, active closed-loop vibration isolation systems based on sensing and counteracting the mechanical disturbances arising in the spacecraft show great promise in overcoming the limitations of passive systems. These can be implemented at the level of the source (Preda et al., 2018) and/or the payload (Sanfedino et al., 2021) and can significantly mitigate the micro-vibration environment in Aeolus follow-on and other future space lidar missions.
The presented work includes data of the Aeolus mission that is part of the European Space Agency (ESA) Earth Explorer program. This includes the L2C wind product (baseline 10) (de Kloe et al., 2020,
OLu analyzed the ALADIN laser frequency stability, its correlation with the reaction wheel speeds and its impact on the Aeolus wind data quality. CL, FW, TK, DW and OR supported the data analysis and interpretation of the results. GR and AH studied the platform parameters related to the Aeolus attitude and orbit control system. GR investigated the results from the on-ground micro-vibration tests with regard to the in-orbit instrument performance. OLe, PM and FF were involved in the on-ground micro-vibration tests of the ALADIN instrument and provided insight into the test procedure and findings at that time. PB provided information of the ALADIN MO cavity control scheme. TP is the Aeolus mission manager. OR is the scientific coordinator of the Aeolus DISC. The paper was written by OLu with contributions from all co-authors.
The authors declare that they have no conflict of interest.
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article is part of the special issue “Aeolus data and their application (AMT/ACP/WCD inter-journal SI)”. It is not associated with a conference.
The Doppler wind lidar ALADIN was built by Airbus SAS in Toulouse, France; the satellite by Airbus Ltd in Stevenage, UK; and the laser transmitters by Leonardo S.p.A. in Florence and Pomezia, Italy. The authors acknowledge Anne Grete Straume (Aeolus mission scientist) and Jonas von Bismarck (Aeolus data quality manager) as well as the Aeolus Mission Advisory Group, the Aeolus Space and Ground Segment Operations teams, and the Aeolus DISC for their invaluable contributions. The authors are also grateful for the insightful discussions with colleagues from the ESA's SWARM mission, particularly Nils Olsen, Anja Strømme, Rune Floberghagen, Laurent Maleville and Jerome Bouffard, regarding the interaction between the reaction wheels, magnetorquers and magnetometers on board Aeolus, which helped to rule out a direct influence of the Earth's magnetic field on the laser frequency stability.
This research has been supported by the European Space Agency (grant no. 4000126336/18/I-BG). The first author was partly funded by a young scientist grant by the ESA within the DRAGON 4 program (grant no. 4000121191/17/I-NB).The article processing charges for this open-access publication were covered by the German Aerospace Center (DLR).
This paper was edited by Ad Stoffelen and reviewed by Weibiao Chen and one anonymous referee.