Global historical eigenvectors of satellite hyperspectral infrared radiance data can be generated by accumulating radiance samples over a given period of time. These eigenvectors can be used in hyperspectral radiance data compression as well as filtering radiance noise. Preliminary results from simulated Atmospheric InfraRed Sounder (AIRS) radiances show that hyperspectral data can be compressed with a ratio above 15:1 resulting in a significant data volume reduction. The accuracy of the compressed radiance data can be within the instrument noise level. Historical eigenvectors of radiances are generated under all sky conditions, which makes real time data compression possible. The historical eigenvector set is updated on a regular basis when new atmospheric information is detected in the AIRS radiances. At least three months of data are needed to obtain the historical eigenvector from simulated near real-time data. Real AIRS data is currently being studied and the results will be presented along with the eigenvector methodology.