The NWP impact of Aeolus Level-2B winds at ECMWF

This report documents the research done at ECMWF over the period 2019-2023 on the data assimilation, monitoring and NWP (Numerical Weather Prediction) impact of the wind retrievals from ESA’s Earth Explorer mission Aeolus. This work was done as part of ESA’s Aeolus DISC (Data Innovation Science Cluster) project.

Aeolus, the world’s first Doppler wind lidar in space, produced wind profile retrievals with sufficient quality to improve global Numerical Weather Prediction forecasts by a useful magnitude. This has been demonstrated with an assessment of data quality via comparisons of the Aeolus Level-2B (L2B) HLOS (horizontal line-of-sight) wind data with the ECMWF model equivalents and by NWP impact assessment with the OSE (Observing System Experiments) and the FSOI (Forecast Sensitivity Observation Impact) methods. Aeolus was operationally assimilated at ECMWF from 9 January 2020 until 30 April 2023 (due to end of mission).

The estimated precision (1 standard deviation) of the HLOS winds varied considerably during the mission and with geolocation, season, processing software version and range-bin settings, particularly for the Rayleigh-clear HLOS winds varying from 4-7.5 m/s in the troposphere. The precision was more stable for the Mie-cloudy varying from 2.5-3.6 m/s in the troposphere. After L2B processor corrections the systematic errors were typically within ±1 m/s (daily averages). The HLOS winds have sufficient information content to improve the ECMWF forecasts despite the random and systematic errors being larger than pre-launch expectations.

Positive impact was demonstrated by OSEs for most periods of the mission, however the magnitude of impact was greater during periods of relatively strong signal levels, which varied a lot during the mission: due to laser energy variations and signal loss within the instrument unique to the FM-B laser.

The largest positive impact of the HLOS winds occurs in the tropical upper troposphere; particularly just after the FM-B laser was switched on in July-December 2019. This period showed statistically significant and good magnitude positive impact on vector wind, temperature, geopotential and humidity forecasts in the tropical and polar troposphere and lower stratosphere by 0.5-2% in root mean square error up to 3-4 days. The impact in polar tropospheric regions tends to fade after 4 days, whilst the positive impact extends to 9-10 days in tropics and Southern Hemisphere extratropics at 100 and 50 hPa (15-20 km). The largest tropical impact at 100 hPa occurs in the east Pacific – a region with the largest wind background errors. The next best impact was shown using the FM-A laser (2018/2019 and 2022/2023), with similar spatial patterns of positive impact to that shown with FM-B. The impact also improved as the ground processing algorithms improved during the mission e.g. going from 1st to 2nd reprocessing campaign for FM-B.

When signals were relatively strong and the noise was reasonable, the Rayleigh-clear winds provided a greater proportion of the positive impact than the Mie-cloudy, presumably due their much greater spatial coverage than Mie-cloudy winds. Mie-cloudy winds tend to provide more of the polar region impact. Mie-cloudy impact was improved using a more realistic assigned observation error modelling (including representativeness error).

Given good signal levels, Aeolus impact is of comparable magnitude to several other important (operational) satellite observing systems at ECMWF which is a positive result for a demonstration mission; considering Aeolus accounted for fewer than 0.5% of the observations assimilated i.e. there is a strong impact per observation.

The relative FSOI for Aeolus was 5% in early FM-B reprocessed dataset in 2019 but reduced to 2% in mid-2022 due to the then very poor signal levels, with the Mie-cloudy impact exceeding the Rayleigh-clear in 2022. This relative FSOI increased to 3-4% in late 2022/early 2023 with the switch back to the FM-A laser. At its peak for the early FM-B period, Aeolus’s relative FSOI is similar in magnitude to a MetOp IASI instrument, ranking amongst the highest FSOI per satellite instrument and it had a similar impact to the radiosonde network.

Consistent spatial patterns in how Aeolus modifies the analysis fields were found: the non-systematic changes tend to be largest in convective areas at ~200-70 hPa; mostly correcting the larger random background forecast errors associated with convection (as seen in the EDA spread). Systematic changes are consistently largest in the tropical upper troposphere and lower stratosphere, and there are also some consistent patterns towards the poles (perhaps fighting other biases in the analysis/model); comparisons to radiosondes suggest Aeolus is not degrading biases.

The magnitude of Aeolus’ positive impact was found to be strongly dependent on the implementation of a bias correction of the HLOS winds derived using the ECMWF model as a reference. The biases have a complicated behaviour (with geolocation and time), meaning it took a great effort by the DISC team to understand and correct them. In Autumn 2019 an explanation for a significant contributor to the bias was found, via careful assessments of the ECMWF O-B statistics with reference to the satellite’s housekeeping information. The Rayleigh biases are predominantly dependent on gradients in temperature across the instrument’s primary mirror. The temperature gradients influence the receiver response via angular changes of the light onto the spectrometers, due to varying focus from the parabolic mirror. The variations in temperature gradients are driven by varying top of atmosphere radiation and the mirror’s on-board thermal control mechanism. A bias correction using the NRT primary mirror thermistor reading as predictors was implemented in the operational ground processing chain since 20 April 2020 (and in reprocessed data), providing an accurate bias correction.

The NWP impact of the demonstration mission Aeolus is promising for obtaining a strong and consistent impact from the proposed operational EUMETSAT follow-on mission (EPS-Aeolus) which aims for significantly improved precision HLOS winds compared to Aeolus and operational robustness