Aeolus + Processes, Technical Note 5

We analyzed the scale-dependent effects of Aeolus winds in the ECMWF system in
relation to flow properties by evaluating analysis increments, defined as the differences
between Aeolus analyses and associated short-range forecasts, in observing system experi-
ments (OSEs) used in the project. Additionally, we assessed the impact of Aeolus data on
predictability improvements in comparison with COSMIC2 data.
Our results show that data assimilation leads to similar systematic increments in the
CTRL, Aeolus, and COSMIC2 experiments. Near the equator, tropospheric biases are
largely associated with balanced zonal flow, whereas in the outer tropical latitudes they
are primarily related to unbalanced meridional flow. The assimilation consistently enhances
the vertical shear of the zonal wind in CTRL, an effect further amplified by Aeolus assim-
ilation, particularly during the QBO disruption of 2019/2020. Based on these findings, we
recommend that Aeolus2 maximizes the vertical resolution of its measurements.
On average, the assimilation of Aeolus winds produces the largest analysis increments
in planetary-scale zonal winds, especially in Kelvin waves and the n = 1 Rossby waves.
The 4D-Var system generates comparable effects on meridional and zonal winds from both
Aeolus and COSMIC2 data. However, the increments exhibit different balance properties;
Aeolus assimilation decreases unbalanced meridional wind in the stratosphere, whereas
COSMIC2 assimilation increases it.

Forecast errors, defined as the differences between OSE forecasts and analyses, exhibit
distinct growth patterns in the zonal and meridional winds across different layers. In the
early stages, error growth is primarily influenced by model error. After approximately three
days, subsynoptic-scale errors become saturated, and the growth is increasingly dominated
by planetary-scale errors in the zonal winds, particularly in the stratosphere. Predictability
gains from the assimilation of Aeolus winds were estimated at approximately 2.7 hours in
the tropopause layer and 5 hours in the lower stratosphere for 3-day forecasts (Fig. 1). For
comparison, assimilation of COSMIC2 data in the same layers yielded predictability gains
of 7 and 13 hours, respectively, during the studied WP4 period (1–14 December 2022).