The ability to understand and predict the effects of routine observations on weather forecasts has only been fleetingly explored to date. One potential implication of such a study is a cost-effective augmentation of the global observational network that leads to reduced forecast errors.

Using methods related to those of Szunyogh et al. (MWR 2000, 2002), the importance of (i) the east Asian rawinsonde network on 3-7 day N.American forecasts, and (ii) the N.American rawinsonde network on 3-7 day European forecasts is analyzed systematically using the NCEP Global Forecast System (GFS). Preliminary insights into the evolution of the influence of the observations ("signals") in various flow regimes will be presented, and the utility of ensemble forecasts in being able to forecast the variance of these signals will be evaluated.