Feature-track winds from geostationary satellites have long been recognized as an important data set for numerical weather prediction, but they are limited to latitudes equatorward of approximately 55 degrees. However, the Cooperative Institute for Meteorological Satellite Studies (CIMSS) at the University of Wisconsin-Madison has in recent years applied their feature-tracking algorithms to data from the MODIS (Moderate Resolution Imaging Spectroradiometer) instrument aboard the EOS Aqua and Terra polar-orbiting satellites to derive feature-track winds in polar regions. These MODIS winds have been shown to yield a significant increase in model skill by ECMWF. This presentation details the use of MODIS winds in NAVDAS (NRL Atmospheric Variational Data Assimilation System), difficulties encountered in implementing them in NAVDAS, and their impact on NOGAPS (Navy Operational Global Atmospheric Prediction System) forecasts.

MODIS winds are formed into superobs for NAVDAS using the same strategy as other feature-track wind datasets. Observations are first partitioned into two degree "prisms"--regions that are approximately equal in size to two degree latitude by two degree longitude boxes at the equator. The longitudinal extent of prisms is allowed to vary to approximately retain this size with the additional constraint that an integer number of prisms be present in a latitude band. Partitioning data into prisms works well at high latitudes since it avoids problems with converging meridians. Winds from different satellites and channels (water vapor or infrared) are superobbed independently, with winds required to be reasonably homogeneous to form a superob.