Tropospheric wind profiling from spaceborne Doppler wind lidar
ESA's Aeolus mission was the first satellite to profile tropospheric winds globally using Doppler lidar, filling a decades-old data void over oceans and the Southern Hemisphere that degraded numerical weather prediction for half the planet.
Sensors
- ALADIN (Atmospheric LAser Doppler INstrument) aboard ESA Aeolus: Single-channel UV Doppler wind lidar at 354.8 nm. Retrieves line-of-sight horizontal wind profiles from 0 to 30 km in approximately 87 range gates, with vertical resolution of 0.5 km in the boundary layer and 1–2 km in the free troposphere. Horizontal averaging length per wind profile is roughly 87 km. Aeolus operated from August 2018 to July 2023.
- Aeolus Mie channel (particulate backscatter): Detects coherent backscatter from aerosol particles and cloud droplets. Provides wind retrievals where sufficient particulate loading exists, typically below 10 km. Vertical resolution can reach 0.5 km in aerosol-rich layers. Signal-to-noise is higher than the Rayleigh channel but coverage is conditional on aerosol presence.
- Aeolus Rayleigh channel (molecular backscatter): Detects incoherent backscatter from air molecules. Provides wind retrievals throughout the full 0–30 km column, including the upper troposphere and lower stratosphere where aerosol is sparse. Vertical resolution is typically 1–2 km. Random error in the free troposphere is on the order of 2–5 m/s per profile before quality control.
- ECMWF IFS assimilation system (Aeolus L2B product ingestion): Not a sensor, but the operational analysis framework that converts Aeolus wind profiles into forecast impact. ECMWF began assimilating Aeolus data in January 2020 and published diagnostics showing forecast skill improvements, particularly for the Southern Hemisphere and tropical troposphere.
- ADM-Aeolus reanalysis and auxiliary meteorological products: ESA and ECMWF produced reanalysis-quality L2C wind products spanning the operational mission period. These are gridded, bias-corrected and quality-flagged, suitable for climatological studies of wind variability and model validation rather than real-time forecasting.
What the wind data gap actually costs a forecast
Numerical weather prediction runs on observations. Temperature and humidity are relatively well sampled by satellites from passive sounders, but wind is different. Radiosonde networks are dense over Europe, North America and parts of East Asia. Over the Southern Ocean, the tropical Atlantic and most of the Pacific, they are nearly absent. Aircraft winds fill some of the gap at cruise altitude, but below 10 km over open ocean the atmosphere is largely unobserved in the wind dimension.
That gap is not a minor inconvenience. Wind is the primary variable governing how weather systems translate, deepen and dissipate. Forecast errors in the Southern Hemisphere have historically been larger than in the Northern Hemisphere for exactly this reason. The World Meteorological Organisation identified a spaceborne wind profiler as the single highest-priority unmet observing need for global NWP, a priority it held for roughly three decades before Aeolus launched.
How ALADIN extracts a wind speed from a photon
ALADIN fires 50 mJ UV laser pulses at 354.8 nm toward the atmosphere at a fixed off-nadir angle of 35 degrees, looking roughly perpendicular to the orbital track. Backscattered photons return to a 1.5-metre telescope. The Doppler frequency shift of those photons relative to the outgoing pulse encodes the line-of-sight velocity of whatever scattered them. Two receiver channels separate the physics: the Mie channel uses a Fabry-Pérot etalon to isolate the narrow-bandwidth return from particles, and the Rayleigh channel uses a double-edge filter to measure the thermally broadened return from air molecules.
Each channel yields an independent wind estimate at each range gate. The two are combined, quality-flagged and averaged over roughly 87 km along-track to achieve acceptable signal-to-noise. The result is a vertical profile of line-of-sight wind, not a full horizontal vector. That single-look geometry is the central limitation of the entire approach: you get one component of the wind, not two. Inferring the full horizontal vector requires either a model background or a second observation from a different azimuth, neither of which Aeolus could provide alone.
Calibration problems that took years to solve in orbit
ALADIN's laser suffered unexpected UV energy degradation in the first months of operation, a known risk for high-power UV lasers in space but worse than pre-launch models predicted. ESA and the instrument team responded by switching to a second laser and by developing a systematic bias-correction scheme applied to the L2B wind product. The primary bias source is a temperature-dependent variation in the instrument's internal reference frequency, corrected using a dedicated dark-current measurement taken when the laser shutter is closed.
A second systematic error, called the Rayleigh channel temperature and pressure sensitivity, arises because the molecular backscatter spectrum depends on atmospheric temperature. Early retrievals assumed a climatological temperature profile; later processing incorporated ECMWF model temperatures as auxiliary input, reducing this error by roughly half. The lesson for any future Doppler wind lidar mission is that on-orbit calibration infrastructure needs to be designed in from the start, not retrofitted.
Demonstrated forecast impact: what the assimilation diagnostics showed
ECMWF published forecast impact assessments after operationally assimilating Aeolus L2B winds from January 2020 onward. The headline result was a statistically significant improvement in 24-hour to 120-hour wind forecasts in the Southern Hemisphere and the tropical troposphere, regions where the observing system was previously weakest. Improvements in the Northern Hemisphere were smaller, consistent with the existing density of conventional observations there.
The magnitude of the Southern Hemisphere improvement was comparable to the impact of removing all aircraft wind observations from the assimilation, which gives a sense of scale. Aeolus also improved the representation of the Quasi-Biennial Oscillation in the tropical stratosphere, a slow equatorial wind oscillation with global teleconnections that models had previously struggled to initialise correctly. These are not claims about future missions; they are published diagnostics from the operational ECMWF system during Aeolus's active life.
What Aeolus cannot do, and what comes next
The single line-of-sight constraint means Aeolus wind profiles are not wind vectors. They are projections of the horizontal wind onto one azimuth. In regions of complex flow, or where the wind direction happens to be nearly parallel to the measurement azimuth, the retrieved component carries little useful information about the dominant flow. Users assimilating these data must propagate that ambiguity correctly through their error covariance matrices, or they risk introducing biased corrections.
Cloud is the other hard limit. Optically thick cloud blocks the lidar entirely below cloud top. Aeolus's vertical coverage in a given profile depends on what the atmosphere presents: clear sky gives full-column coverage to 30 km, while a solid stratocumulus deck gives nothing below cloud top. Reported coverage statistics for the Mie channel in the lower troposphere are substantially lower than for the Rayleigh channel in the upper troposphere, precisely because cloud blocks the former more often.
ESA's Aeolus-2 mission, currently in development, aims to address some of these limitations with improved laser technology and a dual-azimuth observation geometry that would allow partial vector retrieval. The European Organisation for the Exploitation of Meteorological Satellites is also developing the Winds mission for the post-2030 Meteosat Third Generation follow-on era. Neither is operational yet.
For organisations building or buying wind-sensitive services today, Aeolus reanalysis products and the published ECMWF assimilation diagnostics are the primary public evidence base. Satellize draws on these open datasets when assessing atmospheric observation gaps for clients designing sovereign monitoring programmes, in the same way it applies open-constellation analysis to other geophysical variables such as the Tonga crop-estimation work.
Reading an Aeolus wind profile without being misled by it
The Aeolus L2B product reports wind speed in metres per second, a quality-control flag, an estimated random error, and separate flags for Mie and Rayleigh retrievals at each range gate. A common mistake is treating flagged profiles as simply absent. In practice, partially cloudy profiles may have valid Rayleigh retrievals above cloud and no Mie retrievals below it. Treating the whole profile as missing discards real information.
Random errors of 2–5 m/s per individual profile are large relative to the meteorological signals of interest. The value of Aeolus comes from the global distribution of these noisy but unbiased measurements, not from any single profile. Statistical aggregation over weeks or months produces climatologically useful wind fields with much lower effective uncertainty. Single-profile use, for instance attempting to validate an individual radiosonde ascent, requires careful error characterisation before drawing conclusions.
Typical figures
| Wavelength | 354.8 nm (UV) |
| Altitude coverage | 0 to 30 km (up to ~87 range gates, cloud-limited below) |
| Vertical resolution | 0.5 km (boundary layer, Mie); 1–2 km (free troposphere, Rayleigh) |
| Horizontal averaging length per profile | ~87 km along-track |
| Random wind error (free troposphere) | ~2–5 m/s per profile (Rayleigh channel, before QC aggregation) |
| Orbital revisit | ~7-day repeat cycle; near-global coverage per cycle (sun-synchronous, 320 km altitude) |
| Wind component retrieved | Line-of-sight horizontal only; not a full vector |
| Mission archive | August 2018 to July 2023 (L2B and L2C reanalysis products available) |
| Operational assimilation latency | ~3 hours from observation to ECMWF first-guess (during operational phase) |
| Cloud limitation | Optically thick cloud blocks lidar; Mie channel coverage in lower troposphere substantially reduced under overcast conditions |
Analytics Satellize can run
| Data-void assessment for NWP impact studies | Spatial analysis of Aeolus L2B profile density and QC-pass rates against WMO observing network maps | GIS layer showing wind-observation density by altitude band and region, with flagged gaps relevant to client forecast domains |
| Aeolus reanalysis wind climatology extraction | Statistical aggregation of L2C reanalysis profiles over user-defined regions and time windows, with separation of Mie and Rayleigh channels | Monthly or seasonal wind profile climatology report (CSV and NetCDF) for specified ocean basins or land regions |
| Forecast impact baseline for sovereign NWP investment cases | Literature-grounded assessment using published ECMWF assimilation diagnostics, mapped to client geographic area of interest | Briefing report quantifying the expected NWP skill deficit in the client's region absent a spaceborne wind profiler, with citation to published ECMWF results |
| QBO and stratospheric wind anomaly monitoring | Extraction and time-series analysis of Aeolus Rayleigh profiles in the tropical stratosphere (18–30 km), compared against ERA5 reanalysis | Quarterly anomaly report on equatorial stratospheric wind phase, relevant to seasonal forecast skill and aviation routing |
| Mission gap analysis for Aeolus-2 and successor planning | Coverage modelling using published orbital parameters and instrument specifications for proposed follow-on missions, benchmarked against Aeolus heritage performance | Technical briefing on observing capability timeline, identifying periods of reduced wind-profile coverage between Aeolus end-of-life and successor operations |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.