The youngest member of ESA's Earth Explorer family, Aeolus was launched in August 2018 and is the first satellite mission to provide profiles of Earth’s wind in cloud-free air globally. The satellite's novel Doppler wind lidar instrument measures the backscatter of laser light from air molecules through its Rayleigh channel and from clouds and aerosols through its Mie channel.
Aeolus was designed to fill the gap in wind-profile measurements in the weather observation network and improve the accuracy of numerical weather prediction (NWP) models. It also plays an important role in the advancement of our understanding of tropical dynamics and processes relevant to climate variability.
In January 2020, the European Centre for Medium-Range Weather Forecasts (ECMWF) was the first meteorological agency to begin assimilating Aeolus wind data operationally after extensive tests showed that the data significantly improved weather forecasts, especially in the southern hemisphere and the tropics. Aeolus’ data were ready for operational use as early as sixteen months after launch, which is an extraordinary achievement for new types of satellite data.
Three of Europe’s biggest meteorological services followed in ECMWF’s footsteps: Germany’s Deutscher Wetterdienst (DWD), Météo-France and the UK’s Meteorological Office (Met Office).
A long-standing partnership
DWD, Germany’s meteorological service, was the first global weather service after ECMWF to assimilate Aeolus data into their NWP models in May 2020, just one week after Aeolus’ data went public. However, DWD is no stranger to Aeolus. It has been part of the development of the satellite’s wind lidar since it was only a concept on paper in the mid-90s.
"The reason why we were involved for so many years in the development of the Aeolus system, was because we investigated the role of wind and temperature measurements in our data assimilation system and came to the conclusion, as many others, that we need more wind profile data over data void areas, like the oceans, the tropics and the polar regions,” Alexander Cress, senior scientist at DWD, explains.
DWD assimilated Aeolus’s Horizontal Line of Sight (HLOS) wind data, derived by ECMWF and transferred to weather centres globally by EUMETSAT and DWD. DWD’s data assimilation and forecast system consists of a hybrid data assimilation system made up of a Local Ensemble Transform Kalman Filter (LETKF) with 40 members and a three-dimensional variational (3DVar) using the background errors and correlations computed by the LETKF and climatological values. The forecasting model is the Icosahedral Nonhydrostatic (ICON) model developed by DWD and the Max Planck Institute in Hamburg.
DWD uses the same observation operator as ECWMF, and the observation error is derived by using the estimated errors within the data product, separated for Mie and Rayleigh winds. They also developed a bias correction depending on wind speed and height separated for the two channels. DWD found a strong positive impact in their system using Aeolus HLOS winds, especially in the Tropics and the polar areas, but also over the Northern and Southern Hemispheres.
A helping hand during COVID times
In 2020, with the COVID-19 pandemic reducing the amount of meteorological data from commercial aircraft and the availability of bias-corrected HLOS winds in real-time, together with very positive forecast scores in the tropics and in the southern hemisphere, Météo-France decided to assimilate Aeolus wind data operationally in their global model on 30 June 2020.
The French meteorological service, Météo-France, has taken an active interest in the assimilation of atmospheric satellite observations for more than 25 years. They were involved in the development of new techniques and instruments that could improve atmospheric satellite observations and their assimilation into NWP models.
The first assimilation experiments for Aeolus data at Météo-France started in 2019, with positive impacts on forecast skill scores. Data from the Level 2B processor run at ECMWF were provided with an instrumental error that was scaled before assimilation into Météo-France’s NWP Action de Recherche Petite Echelle Grande Echelle (ARPEGE) model, in order to also account for additional errors. The most difficult aspects came from systematic errors and quality controls. If not accounted for properly, they can jeopardise the behaviour of the assimilation system and lead to degraded numerical forecasts.
“An important aspect in data assimilation concerns the quality control of observations and the specification of observation errors,” Jean-François Mahfouf, senior scientist at Météo-France, says. “Here, the roles of ECMWF and ESA were instrumental in providing techniques to correct data biases and remove any data corrupted by systematic errors.”
A clear benefit to forecasts
The United Kingdom’s Meteorological Office (Met Office), began assimilating Aeolus data operationally in December 2020. The assimilation came after six months of extensive testing, during which systems with assimilated Aeolus data ran in parallel with the operational system, so forecasters could assess what changes the assimilation would bring to the system.
“There is a definite improvement in many forecast variables when you use Aeolus,” says Gemma Halloran, research scientist at the Met Office. “Often, when you introduce something new to the model, it might make some things better, but it might also make things worse elsewhere, because you’re changing the whole balance of the system. But with Aeolus, it’s just a clear benefit, I haven’t seen any negative effects of using the data, which are really good for adding in a new observation type.”
At the moment, only Mie channel data have been assimilated into the Met Office’s operational NWP model. Mie channel tests started early, because it showed negligible errors compared to the Rayleigh channel data, which suffered from significant errors due to a mirror temperature-based bias.
Once ESA applied the correction for the bias, prior to the public release of Aeolus data in May 2020, the Met Office could proceed with running assimilation experiments to ensure that the Rayleigh channel data were safe and bring improvements to their forecasts. Assimilation experiments with the Rayleigh winds have shown positive impact and the change has been proposed for the next operational upgrade, which is expected to go operational in early 2022.