<title>Uncertainties in FTIR data due to a nonlinear response</title>

Abstract

Apodization functions used with FTIR instruments cause the response of the instruments to be nonlinear over the entire range of gas concentrations that are encountered in the atmosphere. FTIR remote sensors generally use the mathematical technique of classical least squares, along with a set of reference spectra to analyze the spectra. These reference spectra are normally supplied with the instrument and are taken at a single, specific concentration path-length product. At times, the atmospheric concentration path-length product is far removed from the concentration path-length product of the reference spectra, and this causes an error in the data. The error occurs because the mathematical process used in the final analysis of the data is linear but the instrument response is not. Classical least-squares, commonly used for this analysis, is a linear process that essentially multiplies the entire reference spectrum by a single multiplier over the wave number region used in the analysis. An important part of the analysis is how well the absorbance due to water vapor in the field spectrum is matched by the water vapor reference used in the analysis, and quite often this specific matching is not done very well. This paper explores the magnitude of the errors obtained when the reference concentration path- length product is not the same as that of the field spectrum - - for methane, which is of some environmental interest, and for water vapor specifically. This is done for the range of water vapor concentrations normally encountered in the atmosphere and for a range of temperatures.