Experiments with an Optimal Spectral Sampling (OSS) Method of Assimilating Hyperspectral Satellite Data

The optimal spectral sampling (OSS) method has been demonstrated to be a fast and accurate way to model radiometric observations and their Jacobians as a linear combination of monochromatic radiative transfer calculations. For current hyperspectral instruments with thousands of channels (e.g., CrIS, IASI), OSS models the complete set of instrument channels with calculations at a few hundred OSS spectral “nodes”. The OSS method uniquely allows a method of assimilation in which the measurements are represented in terms of nodes instead of in terms of channels. With this method, observations from all channels are projected onto the OSS nodes and the forward modeling of radiances and Jacobians is done strictly on monochromatic nodes. With this node-based approach, the information content of the full channel set can be assimilated with a computational cost comparable to or better than retrieving only a small subset of the channels using the conventional channel-based approach. We will present results of application to IASI measurements with the NOAA GSI assimilation system. The application addresses adaptive bias correction and static and dynamic filtering of observations, as adapted from treatment of channels to treatment of nodes. Representation of the observation error covariance in terms of node radiance errors and assimilation performance metrics will be discussed.