A predictability study using geostationary satellite wind observations during NORPEX

High-density geostationary satellite wind observations have become an impor tant new contributor to the observing network over oceanic regions. During the 1998 North Pacific Experiment (NORPEX), assimilation of these data in t he Navy Operational Global Atmospheric Prediction System (NOGAPS) provided substantial improvements in 48-h forecast skill over the northeast Pacific and western North America. The current study shows that the large positive impact of the geostationary satellite winds results mainly from the reducti on of analysis errors that project onto the leading singular vectors derive d from the linearized forecast model. These errors account for only a small fraction of the total analysis error and, during NORPEX, were confined mos tly to the middle and lower troposphere with maxima over the central Pacifi c. These errors do not necessarily coincide with the locations of the large st analysis errors. Experiments in which the satellite information is retai ned only at prescribed vertical levels in the analysis confirm that the inc rements in the middle and lower troposphere account for most of the forecas t impact. Implications for the design of future observing systems, including strategi es for targeted observing, are discussed. It is argued that the results sup port the key underlying principles of targeted observing, namely, that the early stages of error growth in most numerical weather forecasts are domina ted by a relatively small number of unstable structures, and that preferent ially reducing analysis errors that project onto these structures can produ ce significant improvements in forecast skill.