Correction of Satellite Temperature Retrieval Errors Due to Errors in Atmospheric Transmittances
Open Access
- 1 June 1986
- journal article
- Published by American Meteorological Society in Journal of Climate and Applied Meteorology
- Vol. 25 (6) , 869-882
- https://doi.org/10.1175/1520-0450(1986)025<0869:costre>2.0.co;2
Abstract
It is generally assumed that atmospheric transmittance functions are known to an accuracy of no better than about five to ten percent. Consequently, one can expect a major impact of these errors on temperature retrieves based on the inversion of the radiative transfer equation, as opposed to regression methods that do not explicitly use transmittance functions. A numerical simulation study of the sensitivity of the retrieved temperature profiles to errors in transmittance is described. The study shows that most of the transmittance error is propagated into the retrieved profiles in the form of a bias error. A technique for removing this large bias component of the error is given. Furthermore, it is shown how the improper use of regularization transforms sonic of the bias error into an unremovable component of random error. Finally, we show results that indicate how well the bias-error removal technique works in practice using real data. It is found that, despite errors of measurement and errors i... Abstract It is generally assumed that atmospheric transmittance functions are known to an accuracy of no better than about five to ten percent. Consequently, one can expect a major impact of these errors on temperature retrieves based on the inversion of the radiative transfer equation, as opposed to regression methods that do not explicitly use transmittance functions. A numerical simulation study of the sensitivity of the retrieved temperature profiles to errors in transmittance is described. The study shows that most of the transmittance error is propagated into the retrieved profiles in the form of a bias error. A technique for removing this large bias component of the error is given. Furthermore, it is shown how the improper use of regularization transforms sonic of the bias error into an unremovable component of random error. Finally, we show results that indicate how well the bias-error removal technique works in practice using real data. It is found that, despite errors of measurement and errors i...Keywords
This publication has 0 references indexed in Scilit: