The Aeolus Data Science and Innovation Cluster (DISC), led by the DLR Institute of Atmospheric Physics, has achieved a major breakthrough in improving Aeolus data quality. Due to this achievement, altitude-resolved wind measurements of the European Space Agency (ESA)’s Earth Explorer Satellite Aeolus are now provided to all interested weather centres and scientific users within three hours of collection. This public data release, which is in place since 12 May 2020, has been met by a broad interest among international weather services: The German Weather Service (DWD) started using the Aeolus data in the daily weather forecast on 19 May, while the British UK MetOffice and the French weather service Météo France plan to follow this year. The European Centre for Medium-Range Weather Forecasts (ECMWF) is already using Aeolus data in weather forecasting since the beginning of January 2020, supplemented by a specially developed method for correcting systematic errors in wind measurements.
The operational availability of these novel, height-resolved global wind data provided by Aeolus in near-real time is the successful result of an intensive 20-month validation and optimisation phase, which began with the launch of the satellite in August 2018. During this time, the Aeolus data and the on-board lidar instrument (lidar: light detection and ranging) were systematically analysed and characterised to improve the derivation of wind profiles from the detector signals.
The Aeolus satellite carries the first wind lidar in space at an altitude of 320 km and was launched on 22 August 2018 to provide altitude-resolved wind information. It was a great success for the Aeolus team that the first vertical profiles of wind speed were measured only 3 weeks after the satellite launch. The measurements showed the typical maxima of wind speed in the upper troposphere of the jet streams at 8-12 km altitude (see also Figure 1). However, after only a few weeks first systematic differences to weather forecast models could be seen. These differences were verified in November 2018 by comparative measurements with the specially equipped DLR Falcon research aircraft (A2D lidar, 2-µm Doppler wind lidar).
Figure 1: Aeolus wind measurement of 19 May 2020, showing vertical profiles of wind speed along the horizontal line-of-sight of Aeolus up to a height of 21 km. The visualisation software VirES for Aeolus - developed by the company EOX in cooperation with DLR and DoRIT - is now available for everyone; HLOS (horizontal line-of-sight) wind speed is the wind speed projected on the horizontal line-of-sight of Aeolus (Picture: DLR, CC-BY3.0).
The correction of such errors by appropriate calibration and algorithms is one of the tasks of the Aeolus DISC. Aeolus DISC is an international consortium led by the DLR Institute of Atmospheric Physics with the participation of the DLR Remote Sensing Technology Institute, the companies DoRIT, ABB, S&T and Serco, as well as several European weather services (ECMWF, Météo-France, KNMI). On behalf of ESA, DISC is responsible for monitoring Aeolus data quality, further developing algorithms, implementing them in operational processors and investigating the influence of Aeolus measurements on weather forecasting.
A few weeks after launch, horizontal stripes were noticed in the measured wind curtains (see Figure 2). These obvious measurement errors could be attributed to single pixels on the CCD detector (CCD: charge-coupled device) which showed increased dark current rates, so-called "hot" pixels. This systematically increased dark current for single pixels of the CCD leads to a small increase of the constant background of the lidar signals, which has to be measured by a calibration in order to be corrected afterwards. Already in June 2019 the influence of these "hot" pixels could be corrected with a method, which was developed at the DLR Institute of Atmospheric Physics.
Figure 2: A correction of the "hot" pixels was introduced on 14 June 2019 for the operational Aeolus processing. The wind speed profiles (colour coded in m/s) of Aeolus without correction (left of the green line) show significant systematic increases at some altitudes (about 3, 11 and 20 kilometres), which disappear with the introduced correction (right of the green line) (Picture: DLR, CC-BY3.0).
After this error had been corrected, further comparison of the Aeolus data with the numerical model at ECMWF in summer 2019 revealed systematic deviations between measured data and model values, which varied both along one satellite orbit and between successive orbits. These deviations showed a clear correlation with the temperature variations of the primary mirror of the Aeolus satellite telescope. This mirror is responsible for collecting the backscattered signals from the atmosphere and thus directly influences the detected measurement signal.
Short- and long-wave radiation of the Earth’s atmosphere into space constantly disturbs the temperature balance of the satellite and the thermal control of the primary mirror can unfortunately balance these disturbances only to a limited extent. However, already small temperature differences across the mirror may cause small changes in the angle of incidence of the laser light reflected back from the atmosphere on the optical spectrometers. Since these spectrometers, which measure the reflected laser light and its wavelength, are very sensitive to angle changes, small temperature variations can lead to erroneous wind measurements.
To correct these measurement errors of several m/s wind speed, a correction method was developed at the DLR Institute of Atmospheric Physics in close cooperation with ECMWF. This correction method is based on a multiple regression approach, which uses a certain number of telescope temperature sensors as predictors for the measurement error. It has been in operation at the ECMWF since mid-April 2020 for error correction. Figure 3 shows the correlation between measured and regression error values for one day of data. The high coefficient of determination (R^2) of 0.81 shows that the described approach is well suited for error correction and that even fine-scale temperature variations can be corrected (see Figure 3 right).
