AbstractAn observing systems experiment to assess the impact of aircraft observations on analyses and forecasts from a numerical model is described. Parallel runs with and without aircraft observations show significant analysis differences in the upper troposphere over northern hemisphere oceans, the aircraft giving a better representation of jet strengths and less reliance on poorer quality observations. This positive impact extends to the shape and position of upper tropospheric features in forecasts from these analyses though the magnitude of the improvements is somewhat less.These results are in broad agreement with those from a similar experiment carried out at the European Centre for Medium Range Weather Forecasts. Some differences attributable to the response of the different analysis schemes used are discussed in the light of geostrophic adjustment theory and are studied further using model forecasts from perturbed initial conditions. These studies suggest that beneficial effects on forecasts can be achieved by distributing single‐level data throughout several model levels and using wind observations to update the rotational component of the wind field.A simple linearized theoretical model supports these conclusions and is used to investigate the effects of the repeated insertion analysis procedure used in our experiments. This effectively suppresses the generation of high frequency waves but shows undesirable effects in the presence of a large mean advective velocit