Fast and accurate radiative transfer in the microwave with optimum spectral sampling

Author: Alan E. Lipton, Jean-Luc Moncet, S.-A. Boukabara, Gennady Uymin and Katherine J. Quinn
Date: 
January 1, 2009
Type: 
Article
Venue: 
IEEE Transactions on Geoscience and Remote Sensing, 47, 1909–1917, doi:10.1109/TGRS.2008.2010933.
Citation: 

Lipton, A. E., J.-L. Moncet, S.-A. Boukabara, G. Uymin, and K. J. Quinn: 2009: Fast and accurate radiative transfer in the microwave with optimum spectral sampling. IEEE Trans. Geosci. Remote Sens., 47, 1909–1917, doi:10.1109/TGRS.2008.2010933.

Optimal spectral sampling (OSS) is a computationally fast and accurate method for modeling sensor-band transmittances and radiances. The spectral response of a sensor channel is approximated by an optimally weighted average of monochromatic radiative transfer calculations at optimally selected points. The absorption coefficients for the selected points are obtained from prestored lookup tables. Analytical Jacobians are produced in conjunction with the radiances with very little added computational burden. A microwave version of the OSS training algorithm and forward model has been developed in parallel with the infrared version. The microwave model treats O2 and N2 as fixed gases and H2O and O3 as variable gases. Several reference line-by-line (LBL) models are available for training. The method of tabulating and interpolating absorption coefficients has been optimized for execution speed.

Results are shown for OSS application to several current and future microwave sounders. With a selected requirement of 0.05-K rms error (with respect to the reference LBL model), the number of monochromatic points required varies from one, for most window channels, to about four for channels embedded in the 60-GHz O2 line complex and along the 183-GHz H2O line. Even in the cases where a single point is adequate, the optimal point does not necessarily coincide with the center frequency of the channel.