Targeted use of rapid-scan geostationary satellite wind data

Satellite observation data now provide the large majority of information used in data assimilation for numerical weather prediction (NWP). The wide variety of new satellite data has great potential to provide observation information in dynamically-sensitive “target regions” where more-accurate atmospheric analyses are required to improve predictive skill of high-impact weather events. The essential problem is to optimize the selection of satellite data for NWP by retaining observations that provide significant information content in target regions, and to eliminate (via intelligent pre-processing) the large amounts of observation data that are redundant or add little value to the atmospheric analysis. In this study, an adjoint-based method is used to identify target regions in which the selection of satellite data can be optimized through channel selection and data thinning procedures. The goal is to provide atmospheric observations in target regions as often as possible and at the maximum horizontal and vertical resolution that increases the value of the atmospheric analyses for NWP. The optimal set of observations is thus a function of the data assimilation procedure, model resolution, and other factors. In this study, the data targeting technique is demonstrated by selectively thinning AMSU-A brightness temperature observations and geostationary satellite wind observations in target (high-sensitivity) and null (low-sensitivity) regions. The study uses the Navy Operational Global Atmospheric Prediction System (NOGAPS) and the NRL Atmospheric Variational Data Assimilation System (NAVDAS).