The usefulness of ADM-Aeolus and any successors depends on several factors that influence information content, not least the spatial resolution and coverage of the data, the ability to characterize both random and systematic (correlated) errors, and the smallness of these errors. In the case of ADM-Aeolus, the DWL will provide layer averaged wind measurements with a 1000 m vertical resolution through most of the atmosphere (i.e. from 2 to 16 km), 500 m below 2 km, and 2000 m between 16 and 27 km. The vertical resolution is tunable during flight, down to 250 m, and is expected to be examined further during in-orbit commissioning. The horizontal sampling strategy will produce from each receiver channel one wind profile every 200 km along-track and each profile will be the result of a 50 km along-track integration. This is achieved by operating the DWL in ``burst-mode'' and, for a satellite ground-speed of 7 km per second, the basic repeat cycle consists of firing 700 laser shots in a 7 second interval followed by 21 seconds ``dormant''. In the data assimilation context, background error correlations for wind fields are generally small at horizontal separations of 200 km or more. Consequently, provided observation error correlation is also small, the sampling strategy ensures a high degree of independence between successive observation profiles. The mission observation requirements on random error are 1 m/s below 2 km and 2 m/s above.
This paper gives details of preparations to assimilate Aeolus data in the ECMWF Integrated Forecast System, including a new method for assessing the expected impact of future observational data on numerical weather prediction (NWP). The method is based on comparing assimilation ensembles, and the impact of a particular data type is assessed by its ability to reduce the spread in the ensemble analyses and in their associated forecasts. The method is applied to simulated Aeolus data and, for calibration purposes, the impact of real radiosonde and wind-profiler data is assessed in the same way. The results are corroborated further by comparing the simulated and real data in terms of their contributions to the information content of global atmospheric analyses. Throughout this range of measures the simulated Aeolus single-component wind profiles compare favourably with the vector wind profiles currently used in routine operational forecasting, i.e. those supplied by the global (but largely land-based) radiosonde and wind-profiling network. The results suggest that Aeolus profiles will be highly complementary to existing wind profile observations and that the main benefits from Aeolus for analysed wind fields will be found over ocean regions in both hemispheres and in the Tropics. These regions have been identified previously as priority areas for improvement. The benefits seen in analysed fields lead to benefits in forecast fields in regions downstream of the improved analysis regions.
The relevance of these results depends on the realism of the simulated data and so close attention has been given to the estimates of observation errors that are assigned to such data. The accuracy of Aeolus measurements will depend primarily on the intensity/attenuation of the backscattered laser light, which in turn depends on the presence and thickness of clouds, and the concentration of aerosol. The yield and accuracy of ADM-Aeolus wind profiles has been investigated through detailed simulations with LIPAS, the LIdar Performance Analysis Simulator developed at KNMI (Royal Dutch Meteorological Institute), using realistic cloud distributions and climatological estimates of aerosol concentration. Examples will be provided to show that the simulated Aeolus observation errors compare favourably with the errors assigned to radiosonde wind observations in an operational weather prediction environment. This supports previous expectations that, by meeting the mission observation requirements as currently specified, Aeolus will have a significant impact on operational NWP. The Aeolus simulations are found to be robust to large variations in cloud cover, thus reducing previous uncertainties in anticipated performance.
Supplementary URL: