VHAMP – Optimisation of Aeolus Spatial and Temporal Sampling
Marseille, Gert-Jan1; Stoffelen, Ad1; Schyberg, Harald2; Megner, Linda3; Körnch, Heiner3

Aeolus aims to measure wind profiles from the surface up to about 30 km altitude (see the abstract of Anne Grete Straume-Lindner for more details on the Aeolus mission). In 2010 it was decided to change the Aeolus laser operation from pulsed burst mode (BM) to continuous mode (CM), removing the 7s on / 21s off cycling of the instrument in order to ensure sufficient measurement stability during flight. A further major adjustment is the change in pulse repetition frequency from 100 to ~50 Hz. Based on these changes to the Aeolus measurement sampling, there was a need to reinvestigate which would be the optimized horizontal and vertical sampling and measurement averaging strategy for the Aeolus mission, ensuring maximized impact of the data in Numerical Weather Prediction (NWP) and general atmospheric circulation modeling. VHAMP (Vertical and Horizontal Aeolus Measurement Positioning) addresses this question taking into account atmospheric dynamical and optical characteristics and their interaction with the Aeolus measurement system. The information content of Aeolus (and in fact of all observing systems) for NWP is determined by:

  • Model characteristics; the spread of the observation information in the model domain is determined by the correlation length scale of the background error covariance matrix. Observations separated less than the correlation length scale do not contain independent information and are thus redundant to some extent. Correlation length scales have been determined for the global ECMWF model and the limited area Harmonie model.
  • Observation representativeness error. NWP models are a smoothed simulation of the true atmospheric state, lacking information on small spatial scales that are observed by most observing systems. This discrepancy is called the observation representativeness error that should be accounted for in data assimilation to assign proper weight to the observations in the analysis.
  • Aeolus wind quality. The LIPAS simulation tool is used to simulate Aeolus wind error bias and standard deviation in realistic atmospheric scenes. Hereto, a high resolution (3.5 km horizontal, 125 m vertical) atmospheric database composed of atmospheric scattering and extinction at 355 nm, i.e., the Aeolus laser wavelength, and wind and temperature has been developed based on 4 months of CALIPSO data and co-located ECMWF model data [1]. Simulated Aeolus winds have been used in a theoretical tool, based on data assimilation analysis equations, to estimate Aeolus NWP impact. The tool quantifies Aeolus impact as a function of Aeolus wind error characteristics, the distribution of vertical bins and along track integration strategies.
  • Existing global observing system. Once in orbit, Aeolus observations will be in competition with observations from existing observing systems in operational data assimilation systems. To assess the added value of Aeolus in an operational context, Ensemble Data Assimilation (EDA) experiment have been conducted with the ECMWF system using real existing observations and simulated Aeolus observations from LIPAS. Main conclusions from the study are: 1. Background error correlation length scales for global models are typically in the order of 200 to 400 km. Correlation length scales increase when going from the Poles to the Tropics and increase with altitude. 2. Wind energy density spectra from the ECMWF model start to deviate from observation spectra at spatial scales of 200 km near the ocean surface (based on scatterometer winds) and at 500 km scales in the free troposphere (based on AMDAR/ACARS/Mode-S wind data). This means that the effective model horizontal resolution of global models is 200-500 km, i.e., substantially lower than the typical model grid size of 10-20 km. Along track averaging over 85-100 km of scatterometer and 100-150 km of aircraft data yields spectra similar to model spectra, i.e., wind observations in agreement with model turbulence and thus negligible representativeness error. The effective vertical resolution of the ECMWF model is 1.7 km based on an intercomparison study with radiosonde winds [2]. 3. Along track integration of 85 km or more is needed to meet the mission requirement for wind error standard deviation. Reducing the laser power from 110 mJ to 80 mJ violates the mission requirement in the upper troposphere (2 ms-1). 4. Wind error biases exceeding 0.5 ms-1 are detrimental for Aeolus impact in NWP. Proper calibration is thus crucial for a successful mission. 5. Aeolus wind error correlations should be less than 0.1 which corresponds to a random error increase of 0.2 ms-1. Larger error correlations have a detrimental impact on NWP impact. 6. It is advantageous to change the vertical sampling along track, i.e., positioning Mie bins at higher altitudes (up to 18 km) in the tropics to sample tropical cirrus. 7. Aeolus will have a substantial beneficial impact for NWP, comparable to radiosondes. Reducing laser power from 110 mJ to 80 mJ only marginally reduces Aeolus impact. However, this assumes that laser power reduction does not affect calibration procedures and quality control and classification procedures of the level-2 processor. Based on VHAMP results, it is recommended to integrate Aeolus measurements along tracks of 85-100 km. The bin sizes should be at least 1 km for the Rayleigh channel in the lower troposphere increasing to 2 km in the stratosphere. The Mie channel is expected to be most relevant for zero wind calibration with 250 m bins near the surface. The vertical sampling should be adapted along the orbit with Mie bins up to 11 km over the Poles and up to 18 km in the tropics. Wind biases should be below 0.4 ms-1 for a successful mission. Calibration procedures need to be further evaluated on this requirement, taking into account a possible laser power reduction during the mission. In addition pulse-to-pulse variation of the laser frequency due to hardware micro-vibrations (laser jitter) potentially hampers calibration in particular in heterogeneous atmospheres and has not been considered so far. [1] Marseille, G.J., Houchi. K., de Kloe J., Stoffelen, A., The definition of an atmospheric database for Aeolus, Atmospheric Measurement Techniques, 4 , 2011., pp. 67-88, doi: 10.5194/amt-4-67-2011 [2] Houchi, K., Stoffelen, A., Marseille, G. J., and de Kloe, J., Comparison of Wind and Wind-Shear Climatologies Derived from High-Resolution Radiosondes and the ECMWF Model, J. Geophys. Res., 115 , D22123, 2010, doi: 10.1029/2009JD013196