|Title||Assimilation of ATOVS radiances at ECMWF: first year EUMETSAT fellowship report|
|Series/Collection||EUMETSAT/ECMWF Fellowship Programme Research Reports|
|Authors||di Tomaso, E, Bormann, N|
|Event Series/Collection||EUMETSAT/ECMWF Fellowship Programme|
|Place of publication||Shinfield Park, Reading|
This study examines different aspects of the use of ATOVS data at ECMWF: the impact of different constellations of MW-sounders, a review of the bias correction of stratospheric observations, the correction of the absorption coefficient in the AMSU-A observation operator. ATOVS radiances have a significant impact on temperature, wind and humidity, and future changes in the constellation of ATOVS (or equivalent) instruments might affect the skill of NWP forecasts. We evaluate the impact of MW sounding data from three evenly-spaced orbits and from more than three satellites. Observing system experiments show some benefit from having an evenly-spaced orbit constellation of AMSU-A sensors and a clear advantage from assimilating all available ATOVS data. We review the bias correction of high-peaking AMSU-A channels. These channels peak in a region where the forecast model error is particularly significant and they are therefore prone to be erroneously corrected for model error by the variational bias correction VarBC. We show that correcting to a lesser extent AMSU-A channel 12, additionally to channel 14, reduces the bias in the temperature analysis in the upper atmosphere. However, the results of the forecast impact are in some cases slightly negative, and therefore we do not suggest an operational change to the current correction of AMSU-A channels. VarBC also corrects biases resulting from errors in the observation operator. The radiative transfer calculations for the simulation of AMSU-A observations on some of the platforms are currently corrected at ECMWF off-line by scaling the absorption coefficient. We calculate a correction for the absorption coefficient of all the AMSU-A sensors. Assimilation experiments show that the correction reduces significantly air-mass dependent biases in channels 5 to 8. We plan to harmonise in operations the use of the correction over the different platforms when new coefficients will be calculated after the introduction of RTTOV-10.