Fast computation of monochromatic infrared atmospheric transmittances using compressed look-up tables

We have developed a new information-content based look-up table technique for the fast computation of near-monochromatic atmospheric transmittances in the infrared that is well suited for nadir viewing satellite and airplane observations. It allows a user to quickly compute near-monochromatic radiances using a very simple algorithm that is easily ported to many machine architectures. Radiative transfer based on look-up tables of monochromatic absorption coefficients could speed calculations, but they are impractical due to their large size and the need to interpolate long wavenumber vectors in temperature and pressure. We use a Singular Value Decomposition (SVD) to transform monochromatic look-up tables of absorption coefficients into a compressed representation that is almost 100 times smaller. Moreover, temperature and pressure interpolations can be performed in this compressed representation, resulting in significant savings in computation times and computer I/O. We start with the Line-by-line computation of a set of tables of absorption coefficients for each relevant gas. Each 25 wavenumber table has 10,000 wavenumber points and 1,100 temperature/pressure layers (100 pressure layers by 11 temperature profiles). For water vapor we add an extra dimension to these tables that spans 5 water vapor profiles to provide variability in the self-broadening of water vapor spectral lines. On average we need 37 basis vectors for water, 12 for carbon dioxide, and 6 for each of the other required gases in order to reproduce the absorption coefficient tables to an accuracy equivalent to a nadir-viewing monochromatic brightness temperature error of 0.1K.