Influence of assimilation schemes on the impact of adaptive observations

One of the main scientific objectives of FASTEX was to provide so-called adaptive observations, in area of strong sensitivity, in order to control the development of forecast errors. Aircrafts equipped for GPS dropsoundes (NOAA Gulfstream IV and NCEP Learjet36) were used for sampling these dynamically crucial areas (20 flights are studied here). The main goal of this work is to investigate the efficiency of these adaptive observations obtained in real time during FASTEX and to assess the effect of the assimilation system on these results. The model used for this study is the French ARPEGE/IFS model with a resolution of T95C3.5 (3.5 stretching factor) and 27 vertical levels. The assimilation schemes are both incremental 3Dvar and incremental 4Dvar, with a T63 resolution for the increments.

This study shows that the FASTEX adaptive observations could strongly influence the forecasts. However current assimilation systems, such as 3Dvar, are not able to build efficient initial field, i.e. with no significant error in the most unstable directions). The assimilation inaccuracies may hide the impact of adaptive observations, pointing out the lack of optimality of the current assimilation system. The 4Dvar assimilation system seems more accurate and allows to extract more information from adaptive observations. These results prove that the effectiveness of adaptive observations also depends of the assimilation system used.