Assimilation of Hyperspectral Satellite Data Projected on Optimal Spectral Sampling (OSS) Nodes

The optimal spectral sampling (OSS) method provides a fast and accurate way to model radiometric observations and their Jacobians (required for inversion problems) as a linear combination of monochromatic quantities. With a “global” training option, OSS minimizes the total number of monochromatic points (“nodes”) required to model a complete set of instrument channels. For current hyperspectral instruments (e.g., CrIS, IASI) with thousands of channels, the number of nodes is a few hundred. OSS has been incorporated as an option within the Community Radiative Transfer Model (CRTM). The OSS forward model can be used as an alternative to other fast forward models in a conventional approach of assimilating observations from hyperspectral channel channels, either for every channel or for a selected subset. An alternative approach, which is unique to OSS, is to project the observations from all channels onto the monochromatic OSS nodes and assimilate the node radiances. This alternative allows the information content of the full channel set to be assimilated with a computational cost comparable to or better than retrieving only a small subset of the channels using the conventional approach. This alternative also benefits from advantages of OSS over with respect to accuracy, and adaptability. We will present a set of experiments with the NOAA GSI assimilation system showing OSS node-based assimilation in relation to conventional approaches with CRTM-OSS and other CRTM options.