Satellite observation impact studies with the Hirlam 4D-VAR assimilation algorithm

Abstract

We have carried out a number of observation impact studies with the aim to determine the individual and joint contributions of different space-based observing systems on the skill of short-range forecasts (up to 48 hours) over the Northern Atlantic and European regions. These studies are an essential part of the current HIRLAM data assimilation plan to include in the near future a bigger set of satellite observations than currently used in the operational runs of the different meteorological services within the HIRLAM consortium. In our experiments with data from up to 8 different satellites and the HIRLAM 4D-Var scheme running with a 6-hours long assimilation window, we have detected a significant positive impact on the forecast skill. The satellite observations not only help to better characterize the initial state, they reduce the mean size of increments and so contribute to filter numerical noise generated by the advection scheme in the assimilation algorithm. Analysing the geographical distribution of this impact, we have found some dependency on the type of observation in line with expectations, but we have ascertained too a good level of synergism among them. We recognise the need to consider more experiments with longer periods before more sound conclusions can be drawn and we are working on them.