An underlying assumption of data assimilation models is that the radiative transfer model used by them can simulate observed radiances with zero bias and small error. For practical reasons a fast parameterized radiative transfer model is used instead of a highly accurate line-by-line model. These fast models usually replace the spectral integration of the product of the transmittance and the Planck function with a monochromatic equivalent, namely, the product of a spectrally averaged transmittance and a spectrally averaged Planck function. The error of using this equivalent form is commonly assumed to be negligible. However, this error is not necessarily negligible and introduces a systematic height-dependent bias to the assimilation scheme. Although the bias could be corrected by a separate bias correction scheme, it is more effective to correct its source, the fast radiative transfer model. I examine the magnitude of error when the monochromatic-equivalent approach is used and demonstrate how a fast parameterized radiative model with Planck-weighted mean transmittances can effectively reduce if not eliminate these errors at source. I focus on channel 12 of the High-Resolution Infrared Radiation Sounder onboard the National Oceanic and Atmospheric Administration (NOAA)-14 satellite that, among all the channels of this instrument, displays the largest error.
© 2000 Optical Society of America
Original Manuscript: February 22, 2000
Revised Manuscript: June 14, 2000
Published: November 1, 2000
David S. Turner, "Systematic errors that are due to the monochromatic-equivalent radiative transfer approximation in thermal emission problems," Appl. Opt. 39, 5663-5670 (2000)